# Source paper archive

- Source title: Haplotype-resolved long-read transcriptomics uncovers allele-specific heat stress induction and a nucleus-targeted chaperone isoform in potato
- Author(s): Nolte, N., Levak, V., Godec, T., Lukan, T., Stare, K., Gruden, K., Petek, M.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.14.766623v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.14.766623v1.full.pdf
- PDF SHA-256: 45506ba495b444553d283585655ad6daff73e23046859c97ff9c70a2a52a8c76
- Converter: pymupdf4llm 0.0.27
- Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version.

---

&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)

# 1 Haplotype-resolved long-read transcriptomics uncovers allele- 2 specific heat stress induction and a nucleus-targeted chaperone 3 isoform in potato


4


5 Nadja Nolte &#91;1,2*&#93;, Valentina Levak &#91;1&#93;, Tim Godec &#91;1,2&#93;, Tjaša Lukan &#91;1&#93;, Katja Stare &#91;1&#93;, Kristina
6 Gruden &#91;1*†&#93;, Marko Petek &#91;1,3*† &#93;
7
8 1 Department of Biotechnology and Systems Biology, National Institute of Biology, Večna pot 121, 1000
9 Ljubljana, Slovenia
10 2 Jožef Stefan International Postgraduate School, Jamova cesta 39, 1000 Ljubljana, Slovenia
11 3 Biotechnical Faculty, University of Ljubljana, Večna pot 111, 1000 Ljubljana, Slovenia
12
13 - Corresponding authors: Nadja Nolte (Nadja.Franziska.Nolte@nib.si); Marko Petek
14 (Marko.Petek@nib.si); Kristina Gruden (Kristina.Gruden@nib.si)
15 - These authors jointly supervised this work
16


1


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


17 **ABSTRACT**


18 Tetraploid potato is highly heterozygous, and its four haplotypes differ in gene content and


19 regulation of gene and isoform expression, yet responses of individual haplotypes to


20 environmental stresses remain poorly characterized. Here, we used long-read RNA


21 sequencing to profile the haplotype-resolved leaf transcriptome of potato cv. Désirée under


22 prolonged heat. Long-read RNA-seq gene counts correlated with short-read RNA-seq at the


23 gene level but assigned far more reads uniquely to individual haplotypes. Under control


24 conditions, allelic imbalance was widespread, whereas after prolonged heat, most alleles


25 responded in the same way across haplotypes. Allele-specific exceptions point to candidate


26 alleles for heat tolerance. We identified a heat-inducible, redox homeostasis-related N²

27 acetylornithine deacetylase allele and confirmed its allele-specific promoter activity _in planta_ .


28 Heat stress induced alternative isoform usage of 390 genes. Among these, we identified a heat


29 stress upregulated heat shock protein 40 chaperone gene isoform that, unlike its alternative


30 splicing isoform, localizes to the nucleus instead of to the cytoplasm. These results show that


31 in a polyploid crop haplotype-resolved long-read transcriptomics can help in identification of


32 alleles for stress tolerance breeding.


33


34 Keywords: allele-specific expression analysis, heat stress, long-read RNA-seq, polyploids,


35 potato, _Solanum tuberosum_, isoform-specific expression analysis


36


2


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


37 **INTRODUCTION**


38 Heat stress presents a major challenge to global food security, increasingly affecting the


39 productivity of staple crops such as tetraploid potato ( _Solanum tuberosum_ ) &#91;1,2&#93;, the fourth most


40 produced food crop worldwide &#91;3&#93; . Potato is adapted to cool climates and is therefore particularly


41 heat-sensitive. Already moderately elevated temperatures impair tuberization and reduce


42 potato yield &#91;1,4,5&#93; . Currently, differential expression analysis in stress response studies is widely


43 performed at the gene level, which fails to reveal two layers of expression regulation: allele

44 level expression, whereby the individual alleles of a gene could respond differently, and


45 isoform-level expression, whereby alternatively spliced transcripts could change their relative


46 abundance. Resolving gene expression further to allele and isoform level can reveal new


47 mechanisms and identify targets for breeding more resilient cultivars.


48


49 In plants, heat activates a conserved network of heat stress transcription factors (HSFs) that


50 induce heat shock proteins (HSPs), ROS-scavenging enzymes and further HSFs &#91;6,7&#93; . Alternative


51 splicing adds a further layer of complexity, modulating HSF function or targeting transcripts for


52 nonsense-mediated decay &#91;8,9&#93; . The potato heat response has so far been characterized mostly


53 at the gene level, and alternative splicing under heat has so far been profiled only after short

54 term heat treatment &#91;10&#93; . This leaves open the question which differential isoform usages occur


55 in potato under agronomically relevant prolonged heat stress and whether expression differs


56 between the four alleles of heat-responsive genes. Genome sequences of highly inbred


57 varieties, such as the collapsed doubled monoploid (DM) genome in the case of potato, are


58 commonly used as reference for quantification of gene expression &#91;11&#93;, but cannot resolve


59 cultivar-specific genes, structural variation, or individual haplotypes. Haplotype-resolved


60 reference genomes, now becoming available for crop cultivars &#91;12&#93;, capture cultivar and


61 haplotype-level variation and allow all haplotype copies (alleles) of each gene to be quantified


62 individually, enabling allele-specific expression analysis. In a haplotype-resolved assembly,


63 alleles can be identified as syntelogs, i.e., homologous genes located in collinear regions (the


3


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


64 same locus) on different haplotypes; we refer to all alleles of one gene as a syntelog group. A


65 key limitation of phased reference genomes, however, is that the high sequence similarity


66 between haplotypes causes a large fraction of reads from commonly used short-read RNA

67 seq to multimap, meaning they cannot be uniquely assigned to a single haplotype &#91;13&#93; . Long-read


68 RNA-seq overcomes this by sequencing full-length transcripts: reads spanning entire


69 transcripts map accurately to highly similar haplotypes and also allow novel transcripts and


70 genes to be identified and quantified &#91;14&#93; . Correct read assignment is especially important


71 because many stress-responsive genes belong to large families, so paralogs (gene family


72 members at different loci) and alleles are readily confounded unless reads can be assigned to


73 a defined haplotype.


74


75 Allelic imbalance, where the alleles of a gene do not contribute equally to its total expression,


76 can reveal the effects on gene expression caused by allelic variation in _cis_ -regulatory regions,


77 such as single nucleotide polymorphisms (SNPs) in transcription factor binding sites or larger


78 structural variants (SVs) &#91;15&#93; . For example, in potato, a quantitative trait locus (QTL) for tuber


79 yield at elevated temperature was resolved by QTL mapping to the heat shock protein _HSC70_


80 allele carrying a TA repeat in its promoter. When expressed under its native promoter in potato


81 cv. Désirée, this allele raised _HSC70_ expression and heat tolerance &#91;16&#93; . This shows that


82 individual _cis_ -regulatory alleles can have substantial phenotypic consequences. Unlike


83 allopolyploids, which combine genomes from different species and often show subgenome


84 expression bias &#91;17,18&#93;, autotetraploid potato lacks such subgenome dominance because all four


85 haplotypes derive from the same species. Nevertheless, local allelic imbalance can still arise,


86 but has so far been studied only in three potato studies: first in cv. Atlantic, against the


87 collapsed DM reference &#91;19&#93;, and the remaining two using the haplotype-resolved cv. Atlantic


88 reference and short-read RNA-seq &#91;20,21&#93; . Long-read RNA-seq has recently been applied for


89 allele-specific expression analysis in hexaploid sugarcane &#91;22&#93;, but high-depth long-read data


90 remain scarce, and it has not yet been used to resolve allele- and isoform-level expression


91 under stress in any crop. Long-read quantification alone does not explain why expression of


4


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


92 alleles differs. Thus, to trace allelic expression differences back to _cis_ -regulatory variation,


93 allele-specific expression analysis needs to be combined with haplotype-resolved promoter


94 sequence and chromatin accessibility analyses.


95


96 Here, we aim to resolve the response of tetraploid potato to prolonged heat stress at the level


97 of individual alleles and isoforms, and to identify the _cis_ -regulatory variation that causes alleles


98 of the same gene to be expressed differently. We hypothesize that (i) long-read RNA-seq


99 mapped to a haplotype-resolved reference genome enables reliable allele-specific


100 quantification; (ii) allelic imbalance is widespread and is at least partly explained by sequence


101 and structural variation in allele promoters, and is therefore shared between cultivars carrying


102 the same haplotypes; and (iii) prolonged heat alters the relative expression of alleles and the


103 usage of isoforms, with consequences for the function of the encoded proteins. To test these


104 hypotheses, we combined long-read RNA-seq of Désirée under 7 days of heat with a


105 haplotype-resolved genome, promoter variation, chromatin accessibility and long-read RNA

106 seq of a second cultivar, Atlantic, and validated selected predictions experimentally.


107


5


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


108 **RESULTS**

109 **Long-read RNA-seq reliably quantifies global allele-specific gene expression in potato**
110 To study the expression response of potato to prolonged heat stress at the level of alleles and


111 isoforms, we sequenced leaf samples from control and 7-day heat-treated potato Désirée with


112 PacBio Kinnex long-read RNA-seq. To evaluate quantification accuracy, we first compared our


113 long-read RNA-seq to standard Illumina RNA-seq performed on the same samples (Fig. 1a,


114 Supplementary Table S1). Reads per sample ranged from 37.9M to 50.0M for short-read RNA

115 seq and from 24.9M to 38.2M for long-read RNA-seq. Mapping long-read RNA-seq to the


116 haplotype-resolved Désirée reference genome identified 25,639 novel isoforms and 6,054


117 novel genes missing from the UniTato DM annotation &#91;23&#93; (Supplementary Fig. S1a). 92% of


118 these novel genes were already annotated in Désirée-specific genome annotation &#91;24&#93;,


119 suggesting they represent cultivar-specific genes absent from the DM reference. Long-read


120 data also corrected the intron–exon structure of several mis-annotated genes (Supplementary


121 Fig. S1b). Next, using only single-isoform genes, we assessed how read length and the


122 reference genome affect the number of reads mapped uniquely to one locus. With both


123 technologies, most genes in the collapsed DM reference had a unique mapping rate above


124 90%, i.e. more than 90% of their reads mapped uniquely (Fig. 1b). Against the haplotype

125 resolved Désirée genome, however, only 7% of alleles had a unique mapping rate above 90%


126 with short-read RNA-seq, and most alleles had more than 90% multimapping reads, as reads


127 could not be distinguished between highly similar alleles. In contrast, with long-read RNA-seq,


128 44% of alleles still had a unique mapping rate above 90% on the haplotype-resolved reference


129 (Fig. 1b).


6


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


**a**



**Plant material** › **Sequencing** › **Reference genome** › **Quantification** › **Validation**



control



control heat





short-read RNA-seq

with Illumina


long-read RNA-seq
with PacBio Kinnex



DM genome


haplotype-resolved

Désirée genome



1


2


3


4







heat



isoform localization



**b** **c** **e**



**DM genome** **DM genome** **DM genome**

5.0



**long-read**


**haplotype-resolved**

**Désirée genome**


**long-read**





short-read



Pearson r = 0.77

n = 9170


count


1000


100


10


1


-5.0 -2.5 0.0 2.5 5.0
short-read RNA-seq l o g 2 FC


**haplotype-resolved**

**Désirée genome**


Pearson r = 0.72

n = 14625









2.5


0.0


-2.5


-5.0


5.0


2.5


0.0


-2.5


-5.0



**Désirée genome**


30000

20000

10000


4000


3000


2000


1000


0

0.0 0.5 1.0



-5.0 -2.5 0.0 2.5 5.0
short-read RNA-seq l o g 2 FC



**d**



**per-locus unique mapping rate**


**0–0.05** **0.05–0.1** **0.1–0.5** **0.5–1.0**



5.0


2.5


0.0


-2.5


-5.0



count


1000


100


10


1


-5.0 -2.5 0.0 2.5 5.0 -5.0 -2.5 0.0 2.5 5.0 -5.0 -2.5 0.0 2.5 5.0 -5.0 -2.5 0.0 2.5 5.0

short-read RNA-seq l o g 2 FC



Pearson r = 0.749

n = 3951



Pearson r = 0.675

n = 7209



Pearson r = 0.729

n = 546



Pearson r = 0.781

n = 2919



130



131 **Fig. 1:** **Experimental design and comparison of long-read and short-read RNA-seq in potato Désirée using**
132 **the collapsed doubled monoploid (DM) and a haplotype-resolved reference.** **(a)** RNA samples from potato
133 Désirée under control and 7-day heat-treated conditions were sequenced with Illumina short-read RNA-seq and
134 PacBio Kinnex long-read RNA-seq, and reads were mapped to the collapsed DM reference genome and to the
135 haplotype-resolved Désirée reference genome. Expression is quantified at the gene (top right) or allele and isoform
136 level (bottom right) and can be interpreted with chromatin accessibility data (ATAC-seq coverage in blue) and
137 transcription factor binding (shown in falcate shape). **(b)** Barplots of per-locus unique mapping rate of single-isoform
138 genes for long- (grey) and short-read RNA-seq (yellow), mapped to the collapsed (top) and haplotype-resolved


7


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


139 reference (bottom). Higher rates indicate that more reads were assigned uniquely to a gene. **(c)** Correlation of
140 heat/control log2 fold changes (log 2 FC) between short- and long-read RNA-seq, quantified using collapsed (DM,
141 top) and haplotype-resolved (Désirée, bottom) reference genome. In all scatterplots (c, d), only genes detected by
142 both technologies were included; n = number of single-isoform genes. **(d)** Correlation of log2FC between long- and
143 short-read RNA-seq mapped to the haplotype-resolved reference for genes grouped by short-read RNA-seq per144 locus unique mapping rate into four groups. **(e)** Venn diagrams of genes differentially expressed between heat and
145 control conditions, detected exclusively by long-read RNA-seq, exclusively by short-read RNA-seq, or by both
146 technologies mapping to the DM reference genome (top) or haplotype-resolved Désirée genome (bottom).
147


148 Log 2 FC correlation between technologies was influenced by transcript length: genes shorter


149 than 750 nt showed poor inter-technology correlation, while genes longer than 750 nt showed


150 comparably strong agreement (Supplementary Fig. S2). On the haplotype-resolved Désirée


151 reference genome, correlation between technologies was lower than when using a collapsed


152 genome as reference (Fig. 1c). This drop can be explained by mapping behavior: per-locus


153 unique mapping rate strongly influenced inter-technology correlation, with lower unique


154 mapping rates of short-read RNA-seq corresponding to lower correlation (Fig. 1d). Because


155 the four Désirée haplotypes are highly similar, short reads often multimapped across


156 haplotypes, whereas full-length long reads remained distinguishable. Thus, the lower


157 correlation on the phased reference reflects haplotype-level multimapping of short reads, a


158 limitation that long-read RNA-seq overcomes.


159


160 Following this stringent quality control, we identified heat-responsive genes by differential


161 expression analysis and asked whether the sequencing technology affected which genes were


162 detected. No genes were regulated in opposite directions between technologies (Fig. 1e;


163 Supplementary Tables S2–S5), and the functional ontology terms enriched among genes


164 differentially expressed under prolonged heat were largely shared across technologies and


165 reference genomes (Supplementary Table S6). Thus, the sequencing technology only affected


166 the number of genes detected as differentially expressed, with long-read RNA-seq detecting


167 more.


168


169 **Allelic imbalance is widespread under control conditions**
170 To evaluate the effect of different haplotypes on gene expression, we compared expression


171 levels of alleles across haplotypes. In the Désirée assembly, 61% of protein-coding genes


8


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


172 were found in syntelogous relationships on at least two haplotypes without copy number


173 variation, with 41% present once in all four haplotypes (Fig. 2a, Supplementary Table S7).


174 Syntelog groups present once on each of the four haplotypes were further filtered to keep only


175 sufficiently expressed groups (see Methods) and to remove groups in which two or more alleles


176 have identical sequences. Reads from identical alleles cannot be assigned to a single allele,


177 so allelic imbalance cannot be estimated for them. The 1,915 filtered syntelog groups were


178 tested for allelic imbalance and 803 showed imbalanced expression. No expression bias


179 toward a particular haplotype was observed for any chromosome (Fig. 2b, Supplementary


180 Table S8), which is in line with what is known for autopolyploids. Allele-specific expression


181 results were largely concordant between long- and short-read RNA-seq; at the same false


182 discovery rate (FDR) both detected a similar number of imbalanced syntelog groups. However,


183 allelic ratio differences were systematically smaller with short reads, so more syntelog groups


184 passed the effect-size threshold with long reads (Supplementary Fig. S3, Supplementary Note


185 S1, Supplementary Table S9).


9


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


balanced

imbalanced







**c**



**d** _CYP89A_ **e** **f**


2339nt



TE class


SV type


allele i


13-LOX type II _DWARF1_ _RIN4_





186



hap1/2/ 3



187 **Fig. 2: Allelic expression in potato leaves grown under control conditions** . **(a)** Pie chart showing the number
188 of genes in the haplotype-resolved Désirée gene annotation classified by their presence/absence and copy number
189 per haplotype in a syntelogous or non-syntelogous relationship (hatched). 1+, the same number of copies (more
190 than one) on each haplotype; CNV, copy number variation, i.e. different numbers of gene copies on different
191 haplotypes. **(b)** Allelic expression ratios of syntelog groups with one allele on each of the four haplotypes with
192 balanced (grey) and imbalanced (orange) expression under control conditions across all potato chromosomes. Red
193 line represents the cutoff for allelic expression ratios defined as imbalanced. **(c)** Difference in expression of alleles
194 (allelic log 2 FC) plotted against the distance of promoter structural variants (SVs) to the transcription start site (TSS,
195 dotted line), type of SV is indicated by color (green: miniature inverted-repeat transposable element (MITE), dark
196 blue: DNA transposon, pink: long terminal repeat (LTR) retrotransposon, orange: long interspersed nuclear element
197 (LINE), grey: not identified as a transposable element (TE), light blue: short interspersed nuclear element (SINE),
198 red: rolling-circle transposon (RC; Helitron), yellow: unknown TE class) and shape (x: deletion, Δ: insertion, °: highly
199 divergent region (HDR)). Circled dots: deletion in promoters of gene shown in panel (d). **(d)** Example of a gene with
200 allelic imbalance - _CYP89A_ (Soltu.DM.06G029570). Barplot of allele-specific expression and illustration showing
201 insertion of long terminal repeat (LTR) in the promoter of haplotype 4 (hap4). **(e)** Correlation of allelic log 2 FC
202 between Désirée and Atlantic for genes in which two alleles, i and j, have promoters that are identical between the
203 two cultivars (the promoter of allele i is identical in Désirée and Atlantic, and the same applies to allele j); allelic
204 log 2 FC = log2(CPM of allele i / CPM of allele j), calculated in each cultivar under control conditions; the six genes


10


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


205 with allelic imbalance conserved in the same direction in both cultivars are highlighted in orange; r,
206 Pearson correlation coefficient; p, p-value of the correlation. **(f)** CPMs for Atlantic and Désirée of three genes
207 highlighted in (e): _13-LOX type II_ (Soltu.DM.01G002150), _DWARF1/DIMINUTO_ (Soltu.DM.02G003240, _DIM_ ),
208 RPM1-interacting protein _RIN4_ (Soltu.DM.06G033690). Alleles sharing the same promoter sequence across
209 cultivars are plotted in the same color (allele i, brown; allele j, white). CPM – counts per million.
210


211 **Promoter structural variants are frequent but can rarely be linked to allelic imbalance**

212 **under control conditions**

213 To link differences in promoter sequences to differences in allelic expression, promoter


214 sequences of allele pairs were annotated for larger structural variants (SVs; ≥ 50 bp) and


215 smaller variants (SNPs and indels; &lt; 50 bp). We identified 9,155 promoter pairs with divergent


216 sequences, while 123 promoter pairs were identical (Supplementary Fig. S4a). Of these, 4,340


217 contained no SVs, 3,318 contained exactly one SV with a median length inside the promoter


218 of 347 bp, and 1,497 contained more than one SV (Supplementary Fig. S4b, Supplementary


219 Table S10). SVs were then classified against a Désirée transposable element (TE) library (Fig.


220 2c). Most SVs were derived from DNA transposons, followed by long terminal repeat (LTR)


221 retrotransposons, while 42% did not match any TE class (Supplementary Fig. S5). Manual


222 inspection of selected imbalanced allele pairs revealed cases in which an SV in the proximal


223 promoter likely explains the imbalance. For example, the cytochrome P450 gene _CYP89A_


224 (Soltu.DM.06G029570) showed lower expression of haplotype 4 than the other haplotypes


225 (Fig. 2d). Promoter comparison revealed a 2,339 bp insertion 339 bp downstream of the


226 transcription start site (TSS) on haplotype 4, classified as an LTR (Fig. 2d). Across all allele


227 pairs, imbalanced promoter pairs had slightly more SVs than balanced pairs (mean 0.92 vs.


228 0.80, p = 1.97 × 10⁻⁷), and their SVs lay closer to the TSS (mean 612 vs. 676 bp, p = 3.57 ×


229 10⁻⁵). Among promoter pairs without SVs, imbalanced pairs also had slightly more SNPs and


230 indels than balanced pairs (mean 55.1 vs. 50.7, p = 2.69 × 10⁻³), and their first SNP or indel


231 lay closer to the TSS (mean 72 vs. 85 bp, p = 0.018; Supplementary Fig. S4c). Thus, although


232 promoter sequences of alleles diverge widely in potato, these differences were only weakly


233 associated with allelic imbalance under control conditions, and only in a few cases could a


234 specific variant be linked to it.


235


11


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


236 **Alleles with identical promoters show correlated imbalance in Désirée and Atlantic**
237 To assess whether allelic expression patterns are maintained across cultivars, allelic


238 expression was compared between cvs. Désirée and Atlantic, reusing publicly available leaf


239 Oxford Nanopore Technologies (ONT) data from cv. Atlantic grown under comparable non

240 stressed conditions &#91;25&#93; . Out of 939 genes tested pairwise, also testing alleles present not on all


241 four haplotypes, for allelic imbalance in Atlantic and 2,990 in Désirée, 185 genes were shared.


242 Out of these, 43 had an identical pair of promoter sequences in both cultivars, allowing us to


243 compare the expression ratios of alleles with identical promoters in different potato cultivars.


244 The high correlation of log 2 FC for these allele pairs between the two cultivars suggests that


245 alleles with identical promoters exhibit similar relative expression patterns across potato


246 cultivars (Pearson's r = 0.83, p-value = 9.33 × 10⁻¹²) (Fig. 2e, Supplementary Table S11). Six


247 genes showed significant allelic imbalance in the same direction in both cultivars. Among these


248 were lipoxygenase ( _13-LOX type II_ ) (Soltu.DM.01G002150), cell elongation protein


249 _DWARF1/DIMINUTO_ (Soltu.DM.02G003240, _DIM_ ) and RPM1-interacting protein _RIN4_


250 (Soltu.DM.06G033690) (Fig. 2f). In 4 of the 6 allele pairs with identical promoters whose log 2 FC


251 diverged strongly between cultivars (log 2 FC &gt;1), at least one allele of the gene lacked a gene


252 model in one of the two reference annotations, so its reads were likely counted toward the


253 annotated allele, which may bias the expression ratio (Supplementary Table S12).


254


255 **Prolonged heat alters the expression of 1,258 alleles, and a distal promoter region**
256 **drives stronger heat induction of a NAOD allele**
257 We identified 1,258 alleles, belonging to 734 genes, with changed expression when plants


258 were exposed to prolonged heat stress (Fig. 3a). Functional enrichment analysis of


259 differentially expressed genes identified secondary metabolism (phenolics, terpenoids), cell


260 wall lignin biosynthesis, cytokinin degradation, protease inhibitors and proteolysis, pathogen


261 defense, phytosterol metabolism, NRT1/PTR anion transport, and the enzyme classes


262 oxidoreductases, glycosyltransferases and hydrolases as functionally enriched


263 (Supplementary Table S6).


12


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)



**a** **b**






|TF Class Repr mes ee mn bta ertive Motif|Col2|
|---|---|
|Basic leucine zipper&lt;br&gt;O2||
|Homeo domain&lt;br&gt;ZHD10||
|Tryptophan cluster&lt;br&gt;RVE7&lt;br&gt;Heat shock&lt;br&gt;HSFB2b&lt;br&gt;TCP&lt;br&gt;TCP4&lt;br&gt;|Tryptophan cluster&lt;br&gt;RVE7&lt;br&gt;Heat shock&lt;br&gt;HSFB2b&lt;br&gt;TCP&lt;br&gt;TCP4&lt;br&gt;|
|C3H(C),C2HC zinc-fingers like&lt;br&gt;UB3&lt;br&gt;Other C4 zinc finger-type&lt;br&gt;CDF3&lt;br&gt;B3&lt;br&gt;RAV2&lt;br&gt;HC3 zinc ribbon factors&lt;br&gt;LSH1|C3H(C),C2HC zinc-fingers like&lt;br&gt;UB3&lt;br&gt;Other C4 zinc finger-type&lt;br&gt;CDF3&lt;br&gt;B3&lt;br&gt;RAV2&lt;br&gt;HC3 zinc ribbon factors&lt;br&gt;LSH1|
|MADS box factors&lt;br&gt;PI||
|Basic helix-loop-helix&lt;br&gt;MYC2||
|GCM domain factors&lt;br&gt;NAC037|GCM domain factors&lt;br&gt;NAC037|



**c** **d**



13-LOX type II
Soltu.DM.01G002150



DEAD(D/H)-box
Soltu.DM.05G007650



_NAOD_

Soltu.DM.08G022800



no change allele i higher j
n=260



allele i higher



allele j higher



n=15234



n=238





haplotype haplotype haplotype


**control** **heat**


Promoter



**e**


3.5



5.0


2.5


0.0


-2.5


-5.0


i j i j i j
allele



**f**





Hap3


0



3.5



Hap4


0





1.5kb 605b



TSS ATG



0 10 20 0 10 20



264 time (hours)
265 **Fig. 3:** **Changes in expression during heat at allele level** . **(a)** Volcano plot of differential expression analysis at
266 the allele level, comparing heat versus control conditions. Alleles with |log 2 FC| &gt; 3 are labeled. **(b)** Enriched motifs
267 (Q ≤ 0.05) in promoters of strongly upregulated alleles during heat treatment (log 2 FC &gt; 2). **(c)** Pairwise comparison
268 of the heat response of two alleles of the same gene. Each allele pair (i, j) is shown once, with allele i on the lower269 numbered haplotype (hap1–hap3) and allele j on the higher-numbered haplotype (hap2–hap4); log 2 FC is
270 heat/control for each allele, and Δlog 2 FC = log 2 FC(i) − log 2 FC(j). Panels are split by type of expression change. In
271 grey: alleles with a similar response (|Δlog 2 FC| ≤ 1 or FDR ≥ 0.05). In red and blue: alleles with a significantly
272 different response to heat (|Δlog 2 FC| &gt; 1, FDR &lt; 0.05) **. (d)** Example genes with differential allelic expression in
273 heat: a lipoxygenase (13-LOX type II) (Soltu.DM.01G002150), an N²-acetylornithine deacetylase ( _NAOD_ )
274 (Soltu.DM.08G022800), and a DEA(D/H)-box RNA helicase (Soltu.DM.05G007650). Scatterplot of CPMs per allele
275 in control (grey) and heat (orange); the x-axis indicates the haplotype on which each allele is located. Lines indicate
276 mean of biological replicates. **(e)** Pairwise alignment of promoter sequences for N²-acetylornithine deacetylase
277 ( _NAOD_ ) (Soltu.DM.08G022800) of haplotype 3 (grey), haplotype 4 (orange) and a shorter (blue; 605 bp long) variant
278 of haplotype 4. Identical promoter regions in grey, mismatches in red and small insertions/deletions in black. ATAC279 seq coverage of heat-treated replicates for the promoters of haplotypes 3 and 4. **(f)** Luciferase assays under control
280 and heat conditions for N²-acetylornithine deacetylase ( _NAOD_ ) (Soltu.DM.08G022800) promoters of haplotype 3
281 (grey), haplotype 4 (orange) and a shortened variant of haplotype 4 (blue), which completely loses its inducibility in
282 heat. Average ± SE is shown. Peak luminescence per well was compared between hap3 and hap4 with a two-sided
283 Wilcoxon rank-sum test, and p-values were FDR-adjusted (Benjamini–Hochberg); control, p.adj = 3.7 × 10⁻¹²; heat,
284 p.adj = 0.0055. ** p.adj &lt; 0.01, *** p.adj &lt; 0.001.
285


286 To better understand gene regulation during heat, we performed a motif enrichment analysis


287 of promoters of differentially upregulated alleles. Promoters of 88 strongly (log 2 FC &gt; 2, FDR &lt;


13


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


288 0.05) upregulated alleles vs. 2,294 non-responding alleles (|log 2 FC| &lt; 0.1) were tested for motif


289 enrichment. We identified 31 enriched motifs (Q ≤ 0.05) from 12 transcription factor classes,


290 including binding motifs of HSFs and basic helix-loop-helix factors (e.g., MYC2), both known


291 to be involved in the heat response (Fig. 3b, Supplementary Table S13).


292 To identify alleles that respond differently to heat, we compared allelic expression between


293 control and heat conditions. Syntelog groups with exactly one copy on at least two haplotypes


294 were tested. 498 (3%) of 15,732 tested pairs showed a differential response to heat (FDR &lt;


295 0.05 and |Δlog 2 FC| &gt; 1) (Fig. 3c, Supplementary Table S14). This indicates that for most genes


296 all alleles responded similarly to heat, i.e., they were all either not affected, down-, or


297 upregulated.


298


299 To confirm the allelic imbalance identified by long-read RNA-seq in heat response, we selected


300 the N²-acetylornithine deacetylase ( _NAOD_ ) gene (Soltu.DM.08G022800) in which haplotype 4


301 allele was more strongly induced by heat than the other two alleles, while the corresponding


302 allele is absent from haplotype 2 (Fig. 3d). Due to the high variability of the promoters between


303 the two haplotypes, laboratory experiments were required to map the motifs causing the


304 difference. We cloned selected promoters (1,500 bp upstream of the start codon) into vectors


305 upstream of a luciferase reporter, transiently expressed this promoter-reporter fusion in


306 _Nicotiana benthamiana_, and measured luminescence under control and heat conditions. The


307 luciferase assays confirmed the stronger heat response of the haplotype 4 promoter compared


308 with haplotype 3, as predicted by long-read RNA-seq (Fig. 3f). Interestingly, heat inducibility


309 was lost when the haplotype 4 promoter was shortened to 605 bp upstream of the start codon,


310 indicating that the deleted region (605–1,500 bp upstream of the start codon) is required for


311 heat-responsive expression. As this region differs at only a few positions between the


312 haplotype 4 and haplotype 3 promoters, these sequence differences represent strong


313 candidates for the enhanced heat responsiveness of the haplotype 4 allele (Fig. 3e).


314 Consistent with this finding, assay for transposase-accessible chromatin with sequencing


315 (ATAC-seq) revealed a major peak of chromatin accessibility in heat-treated samples for both


14


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


316 alleles, located approximately 1,000–1,300 bp upstream of the start codon (Fig. 3e). This


317 region overlaps the deleted segment, further supporting its role in heat induction.


318


319 **Long-read RNA-seq resolves the heat response of individual members of multigene**

320 **families**

321 We next examined individual members of multigene families, which short-read RNA-seq


322 resolves poorly. One such family, the heat shock proteins, is regulated by HSFs, so their


323 promoters are expected to carry HSF binding motifs. To study their regulation, we combined


324 chromatin accessibility data, haplotype-resolved promoter sequences and expression profiles.


325 We observed that the heat shock protein genes ( _HSP90_, _HSP70_, _HSP101_ and DNAJ

326 containing proteins) that were significantly upregulated in heat-treated samples all had ATAC

327 seq coverage colocalized with motifs of HSFs (Fig. 4a). Notably, the motif shared by two HSFs


328 from tomato, SolycHSFB2b and SolycHSFA7, was found in most ATAC-seq coverage peaks


329 close to the TSS and was also found to be enriched in strongly heat-upregulated alleles (Fig.


330 3b). The alleles of each gene responded similarly to heat (comparable log 2 FC heat/control),


331 the only exception being a DNAJ-containing protein (Soltu.DM.01G039980). Haplotype 2 allele


332 of this DNAJ-containing protein has a 5,536 bp deletion compared to the other haplotypes,


333 potentially affecting heat inducibility (Fig. 4a).


334


335 Next, we looked at the Kunitz-type protease inhibitor gene family located on chromosome 3,


336 members of which were differentially expressed during prolonged heat stress. Their copy


337 number varies between haplotypes: haplotype 1 lacks two of the protease inhibitor genes,


338 whereas in haplotype 4 two genes are translocated and duplicated. Haplotypes 2 and 4 share


339 all but one allele, whereas haplotype 3 carries unique alleles (Fig. 4b,c). Among the Kunitz

340 type protease inhibitors, Cysteine protease inhibitor 8 ( _CPI8,_ Soltu.DM.03G018520) was the


341 most highly expressed and had the highest log 2 FC during heat across all haplotypes. It is


342 known to be induced by jasmonic acid during biotic stress &#91;26&#93;, and we observed a _MYC2_ binding


343 site 714 bp upstream of the TSS within an ATAC-seq peak (Supplementary Fig. S6).


15


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)



DM


Hap1


Hap2


Hap3


Hap4

|Soltu.DM. Soltu.DM. Soltu|Col2|Soltu.DM. Soltu|Soltu. M. oltu.DM.|
|---|---|---|---|
|||||
|||||
|||||



Hap1


Hap2


Hap3


Hap4



**a**


_**HSP90**_

Soltu.DM.03G023440


_**HSP70**_

Soltu.DM.04G007430


_**DNAJ**_

Soltu.DM.01G039980


_**HSP70**_

Soltu.DM.03G022200


_**HSP90**_

Soltu.DM.07G015790


_**HSP70**_

Soltu.DM.11G020440


_**HSP101**_

Soltu.DM.03G029350



Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4


Hap1


Hap2


Hap3


Hap4



**b**


**c**



**long-read**

**RNA-seq**

log2FC

4


2


0


CPM


200


150


100


50


0


**ATAC-seq**

(normalized per-gene)


1.00


0.75


0.50


0.25


0.00





**Motif**


HSFA6B


HSFB2A


HSFB4


SolycHSFA7


SolycHSFB2b


SolycHSFC1


**Strand**


   
## +



log 2 FC

4


2


0












|X|X|
|---|---|
|X|X|
|X|X|
|||


|Col1|Col2|
|---|---|
|||
|5,536|bp|
|||
|||
|||
|||



344



-2000 TSS TSS+500
Position (bp)



345 **Fig. 4:** **Heat-responsive gene families** . **(a)** Log 2 FC, average CPM expression in control and heat, ATAC-seq
346 coverage normalized by gene and six known HSF binding motif matches shown 2,000 bp upstream and 500 bp
347 downstream of the TSS. Motifs identified on the forward (triangle) and reverse (circle) strands of HSFA6B (blue),
348 HSFB2A (red), HSFB4 (purple), SolycHSFA7 (pink), SolycHSFB2b (green) and SolycHSFC1 (yellow). Where motifs
349 overlapped, the one with the highest position weight matrix (PWM) score is shown. For the DNAJ protein
350 (Soltu.DM.01G039980), a 5,536 bp deletion in the promoter of haplotype 2 is marked by a black triangle. **(b)**
351 Haplotype-resolved locations of Kunitz-type protease inhibitor genes on chromosome 3 with differential expression
352 during heat. The same genes are connected by lines and identical alleles share color. Arrow directions correspond
353 to orientation in the genome. **(c)** Heatmap showing log 2 FC and presence/absence (“X”) of Kunitz-type protease
354 inhibitor genes. Lowly expressed genes marked with “0”.
355


356


357 **Prolonged heat stress leads to differential isoform usage for 3** **90** **genes**
358 We also explored isoform diversity in potato leaves at the allele level. To identify genes whose


359 major isoform (mRNA splicing variant) differed between haplotypes, we compared expressed


360 isoform structures across haplotypes under control conditions. Of 4,605 genes tested, 154


361 differed significantly in their major isoform, mostly within the untranslated regions (UTRs) (Fig.


362 5a, Supplementary Table S15).


363


364 For example, the gene Soltu.DM.02G008950, coding for an isoprenylcysteine methylesterase


365 (ICME), has an allele-specific isoform 2 predominantly expressed on haplotype 3 differing in


366 two splice sites from the other haplotypes (Fig. 5b), one of these is the alternative 5′ splice site


16


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


367 in the first exon. Haplotypes 1 and 4 had low expression of this isoform and haplotype 2 did


368 not express it at all. There is no splice site mutation in the first intron in the second isoform to


369 explain its alternative splicing difference in haplotype 3, but it has a 168 bp deletion in the 5′


370 UTR located 19 bp upstream of the start codon. In addition, haplotype 3 expressed transcripts


371 with a shortened exon 9 due to a variation in the 3′ splice site of exon 9 (Fig. 5b). Similarly, in


372 Atlantic, haplotype 4 expressed the same alternative isoform (Supplementary Fig. S7).


373


374 To identify isoform usage changes induced by heat, we compared isoform usage ratios across


375 all expressed genes, identifying 390 of 13,115 tested genes with differential usage (change &gt;


376 10%, FDR &lt; 0.05; Fig. 5c, Supplementary Table S16). For each of these genes, isoform usage


377 shifted in the same direction in every expressed haplotype. We examined genes with known


378 heat-response functions or previously reported alternative splicing, as described below. In


379 _HSFA6B_ (Soltu.DM.09G020350), the two major isoforms were upregulated under heat, with


380 increased use of the isoform carrying an additional C-terminal exon relative to the control

381 dominant isoform (Fig. 5d); this splicing replaced the last 9 residues with 23 residues predicted


382 to form an amphipathic α-helix (Supplementary Fig. S8). For the jasmonate-ZIM-domain


383 protein _JAZ3_ (Soltu.DM.01G000760), heat increased the abundance of isoform 1


384 (Soltu.DM.01G000760.1), which uses an alternative 3′ splice site of the third exon that shortens


385 the exon by 96 nt, while expression of isoform 2 decreased; the TIFY and ZIM domain regions


386 are unaffected, and the protein is an ortholog of jasmonate-ZIM-domain protein 3 in


387 Arabidopsis (Fig. 5e). _SPA1_ (Suppressor of phyA-105) _-RELATED 3_ (Soltu.DM.12G029670),


388 an ortholog of Arabidopsis _SPA3_, showed increased expression of an alternative isoform


389 (Soltu.DM.12G029670.1) in haplotype 2 under heat stress; this isoform uses an alternative 5′


390 splice site within the fourth intron, potentially introducing a premature stop codon, and the


391 predicted truncated protein retains three of the seven WD40 repeats present in the longer


392 isoform (Fig. 5f). In _COP1_ (Constitutive Photomorphogenesis 1; Soltu.DM.12G023350),


393 increased use of an alternative 3′ splice site in the second exon under heat extends this exon


394 from 78 to 114 nt, introducing a premature stop codon in Soltu.DM.12G023350.2 (Fig. 5g).


17


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


395 Finally, in beta-carbonic anhydrase (Soltu.DM.05G008530), a novel isoform (BambuTx2249)


396 arising from an alternative transcription initiation site accounted for the majority of transcripts


397 on haplotype 2 under heat, compared with roughly equal isoform proportions under control


398 conditions, and this alternative TSS produces a coding sequence (CDS) with a different N

399 terminus (Fig. 5h).


400


18


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)



c

























**i**




|b 1|Col2|Col3|Col4|
|---|---|---|---|
|sion|sion|sion|sion|
|~~0.5~~&lt;br&gt; oform expres|~~0.5~~&lt;br&gt; oform expres|~~0.5~~&lt;br&gt; oform expres|~~0.5~~&lt;br&gt; oform expres|
|~~**d**~~&lt;br&gt;~~**_HSFA6b_**~~&lt;br&gt;~~Soltu.DM.09G020350.1~~&lt;br&gt;~~Soltu.DM.09G020350.3~~&lt;br&gt;~~between haplotype~~s~~)~~&lt;br&gt;Soltu.DM.02G008950.2&lt;br&gt;Soltu.DM.02G008950.1&lt;br&gt;~~0~~&lt;br&gt;2&lt;br&gt;~~3~~&lt;br&gt;~~4~~&lt;br&gt;~~168 bp DEL~~&lt;br&gt;~~ICME~~&lt;br&gt;relative is&lt;br&gt;~~**e**~~|~~**d**~~&lt;br&gt;~~**_HSFA6b_**~~&lt;br&gt;~~Soltu.DM.09G020350.1~~&lt;br&gt;~~Soltu.DM.09G020350.3~~&lt;br&gt;~~between haplotype~~s~~)~~&lt;br&gt;Soltu.DM.02G008950.2&lt;br&gt;Soltu.DM.02G008950.1&lt;br&gt;~~0~~&lt;br&gt;2&lt;br&gt;~~3~~&lt;br&gt;~~4~~&lt;br&gt;~~168 bp DEL~~&lt;br&gt;~~ICME~~&lt;br&gt;relative is&lt;br&gt;~~**e**~~|~~**d**~~&lt;br&gt;~~**_HSFA6b_**~~&lt;br&gt;~~Soltu.DM.09G020350.1~~&lt;br&gt;~~Soltu.DM.09G020350.3~~&lt;br&gt;~~between haplotype~~s~~)~~&lt;br&gt;Soltu.DM.02G008950.2&lt;br&gt;Soltu.DM.02G008950.1&lt;br&gt;~~0~~&lt;br&gt;2&lt;br&gt;~~3~~&lt;br&gt;~~4~~&lt;br&gt;~~168 bp DEL~~&lt;br&gt;~~ICME~~&lt;br&gt;relative is&lt;br&gt;~~**e**~~|~~**d**~~&lt;br&gt;~~**_HSFA6b_**~~&lt;br&gt;~~Soltu.DM.09G020350.1~~&lt;br&gt;~~Soltu.DM.09G020350.3~~&lt;br&gt;~~between haplotype~~s~~)~~&lt;br&gt;Soltu.DM.02G008950.2&lt;br&gt;Soltu.DM.02G008950.1&lt;br&gt;~~0~~&lt;br&gt;2&lt;br&gt;~~3~~&lt;br&gt;~~4~~&lt;br&gt;~~168 bp DEL~~&lt;br&gt;~~ICME~~&lt;br&gt;relative is&lt;br&gt;~~**e**~~|
|~~**_TED 3_**~~&lt;br&gt;rol&lt;br&gt;eat&lt;br&gt;~~**g**~~&lt;br&gt;~~**h**~~&lt;br&gt;~~Soltu.DM.12G023350.2~~&lt;br&gt;~~Soltu.DM.12G023350.3~~&lt;br&gt;~~atio (heat vs contro~~l~~)~~&lt;br&gt;control&lt;br&gt;heat&lt;br&gt;~~**_COP1_**~~&lt;br&gt;rol&lt;br&gt;eat&lt;br&gt;~~WD40-repeats~~&lt;br&gt;~~Coil~~&lt;br&gt;~~STOP~~&lt;br&gt;~~RING-finger~~&lt;br&gt;~~WD40-repeats~~&lt;br&gt;~~12G029670.1~~&lt;br&gt;12G029670.2|~~**_TED 3_**~~&lt;br&gt;rol&lt;br&gt;eat&lt;br&gt;~~**g**~~&lt;br&gt;~~**h**~~&lt;br&gt;~~Soltu.DM.12G023350.2~~&lt;br&gt;~~Soltu.DM.12G023350.3~~&lt;br&gt;~~atio (heat vs contro~~l~~)~~&lt;br&gt;control&lt;br&gt;heat&lt;br&gt;~~**_COP1_**~~&lt;br&gt;rol&lt;br&gt;eat&lt;br&gt;~~WD40-repeats~~&lt;br&gt;~~Coil~~&lt;br&gt;~~STOP~~&lt;br&gt;~~RING-finger~~&lt;br&gt;~~WD40-repeats~~&lt;br&gt;~~12G029670.1~~&lt;br&gt;12G029670.2|~~**_Carbonicanhydrase_**~~&lt;br&gt;control&lt;br&gt;heat&lt;br&gt;ol&lt;br&gt;at||
|nt&lt;br&gt;h&lt;br&gt;nt&lt;br&gt;h|nt&lt;br&gt;h&lt;br&gt;nt&lt;br&gt;h|tr&lt;br&gt;he|tr&lt;br&gt;he|
|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|on&lt;br&gt;|on&lt;br&gt;|
|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|c|c|
|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|co&lt;br&gt;&lt;br&gt;co&lt;br&gt;|||
|||||
|||||
|||||
|||||
|||||



**j**
















&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


406 sites specific to haplotype 3 marked in red. For clarity, 4,000 bp of exonic sequence with identical splicing across
407 all haplotypes were omitted from the visualization. DEL: 168 bp deletion in the 5′ UTR of haplotype 3. **(c)** Volcano
408 plot of differential isoform usage between heat and control conditions. **(d–i)** Individual genes with heat-responsive
409 differential isoform usage, each shown as a gene model with the alternative splice event, isoform CPMs for one
410 haplotype and gene CPMs in control versus heat. Black: CDS, grey: UTRs; **(d)** _HSFA6B_ (Soltu.DM.09G020350)
411 where isoform 1 has an additional coding exon; **(e)** _JAZ3_ (Soltu.DM.01G000760), with an alternative isoform with a
412 shortened third exon; **(f)** _SPA1-RELATED 3_ ( _SPA3_ ; Soltu.DM.12G029670), with a truncated isoform retaining three
413 of seven WD40 repeats; **(g)** _COP1_ (Soltu.DM.12G023350), with an alternative 3′ splice site producing a shorter
414 isoform with a premature stop codon; the CDS is shown from the start codon shared with the full-length isoform,
415 the likely translation start, rather than the annotated longest CDS. **(h)** carbonic anhydrase (Soltu.DM.05G008530)
416 with an isoform with alternative TSS; **(i)** _DNAJ/HSP40_ chaperone (Soltu.DM.03G037350), with an alternative 3′
417 splice site introducing a frameshifted, extended C-terminus; Lysines/K and Arginines/R are marked in red. **(j)**
418 Confocal microscopy images of _N. benthamiana_ leaves transiently expressing _HSP40_ isoform 1
419 (Soltu.DM.03G037350.1, left) or isoform 2 (Soltu.DM.03G037350.2, right) fused to mTagBFP2. Overlay of the
420 fluorescent protein channel (green) and chlorophyll autofluorescence (purple). All channels are shown in
421 pseudocolor. Separate channels are shown in Supplementary Fig. S9; raw images are deposited in Zenodo.
422


423 **Heat-induced splicing relocalizes a DNAJ/HSP40 chaperone to the nucleus**
424 Heat also changed isoform usage of the DNAJ/HSP40 chaperone Soltu.DM.03G037350. Its


425 two main isoforms differ at the 3′ splice site of the fifth intron: isoform 2


426 (Soltu.DM.03G037350.2) uses an alternative site 4 nt downstream of the site used by isoform


427 1 (Soltu.DM.03G037350.1). Under heat, the usage of isoform 2 increased in all four haplotypes


428 (FDR ≤ 0.003), on average from 20% to 35% of transcripts, whereas that of isoform 1


429 decreased on average from 72% to 60%; isoform 1 nevertheless remained the main isoform


430 in both conditions. The alternative splice site causes a frameshift that gives isoform 2 a longer,


431 different C-terminus (Fig. 5i). This difference was predicted to alter subcellular localization:


432 isoform 1 was predicted to localize about equally to the nucleus and cytoplasm, whereas


433 isoform 2 was predicted to localize mainly to the nucleus (Supplementary Table S17).


434 To test the hypothesis that the two expressed isoforms lead to different localization of HSP40,


435 we fused the coding sequences of both isoforms to the fluorescent tag mTagBFP2 under the


436 control of the CaMV 35S promoter. The constructs were transiently expressed in _N._


437 _benthamiana_ leaves, and their subcellular localization was examined using confocal


438 microscopy under control conditions. The two isoforms localized to distinct compartments:


439 isoform 1 to the cytoplasm and isoform 2 to the nucleus (Fig. 5j, Supplementary Fig. S9).


440


20


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


441 **DISCUSSION**

442 Prolonged, moderate heat is the form of heat stress most relevant to potato cultivation, yet its


443 molecular effects have so far been studied almost exclusively at the whole-gene level &#91;4,27,28&#93; .


444 Zagorščak et al. performed integrative multi-omics profiling of heat, drought and waterlogging


445 in Désirée and identified stress-specific transcriptional, proteomic and metabolic signatures


446 and their hub regulators, but quantified each gene as a single unit, without distinguishing its


447 alleles or isoforms &#91;27&#93; . More recent work has begun to resolve stress responses at the isoform


448 level &#91;10&#93;, yet how the individual alleles of a gene contribute to heat response in potato remained


449 unexplored.


450


451 Here, we combined a haplotype-resolved Désirée reference genome with long-read RNA-seq


452 to quantify expression at the allele and isoform level under control conditions and after 7 days


453 of heat (Fig. 1a). Because long reads can be assigned to individual alleles more reliably than


454 short reads, long-read RNA-seq is particularly beneficial for allele-level quantification in


455 polyploids and in diploids with duplicated genes (Fig. 1b) &#91;13,29&#93; . In our data, short reads


456 systematically underestimated allelic ratio differences (Supplementary Note S1), likely


457 because reads that cannot be uniquely assigned dilute allele-specific signal.


458


459 Haplotype-resolved reference genomes, combined with long-read RNA-seq, in principle allow


460 promoter variation between alleles to be linked to expression differences. In organisms with


461 little sequence variation between haplotypes, alleles differ at only a few positions, so


462 expression differences between them can be more easily attributed to specific _cis_ -regulatory


463 variants &#91;30&#93; . However, in highly heterozygous organisms such as commercial potato cultivars,


464 alleles typically differ at many promoter positions at once (Supplementary Fig. S4a,


465 Supplementary Table S10), and for most genes it is unknown which transcription factors


466 regulate them and where they bind. It is therefore difficult to determine which of the many


467 sequence differences, if any, causes an observed expression difference. Consistent with this,


21


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


468 many alleles with promoter SVs were expressed at similar levels (Fig. 2c), showing that an SV


469 alone does not predict allelic imbalance unless it affects functional _cis-_ elements.


470


471 While _cis_ -elements in promoters influence expression level, _cis_ -elements (e.g. splice sites)


472 within exons or introns can regulate which isoform is expressed. We identified genes whose


473 major isoform differed between haplotypes under control conditions and, in some cases, linked


474 this to specific sequence differences (Fig. 5a, Supplementary Table S15). In ICME


475 (Soltu.DM.02G008950), the haplotype 3 isoform differed from the other haplotypes at two


476 splice sites (Fig. 5b). One difference likely resulted from a C&gt;T variant at position −3 of the 3′


477 splice site (CAG&gt;TAG) (Fig. 5b), weakening the splice site and leading to the selection of a


478 cryptic splice site &#91;31&#93; . If this isoform is non-functional, the other three alleles would still provide


479 the canonical isoform, lowering but not abolishing ICME activity. As Arabidopsis ICME


480 (AT5G15860) promotes abscisic acid (ABA) signaling &#91;32&#93;, a lower dosage of potato ICME might


481 subtly dampen ABA responsiveness.


482


483 If promoter sequence contributes to allelic imbalance, alleles that share an identical promoter


484 should show similar imbalance in different cultivars. Such shared promoters are expected in


485 potato, because historical European cultivars have low haplotype diversity &#91;33&#93; . Indeed, we


486 identified identical promoters between Atlantic and Désirée, and the relative expression of


487 alleles sharing a promoter was highly correlated between the two cultivars (Fig. 2e,


488 Supplementary Table S11), indicating that allelic imbalance is at least partly encoded in _cis_


489 and is reproducible across cultivars. Genes imbalanced in both cultivars included the RPM1

490 interacting protein _RIN4_ (Soltu.DM.06G033690), where one haplotype was approximately


491 twice as highly expressed (Fig. 2f). Allelic dosage of such a gene may matter for disease


492 resistance, as _AtRIN4_ (AT3G25070) regulates both PAMP-triggered and effector-triggered


493 immunity &#91;34,35&#93; .


494


22


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


495 Interestingly, only 3% (498) of the allele pairs tested showed a differential allelic response to


496 heat (Fig. 3c, Supplementary Table S14), suggesting that the _cis_ -regulatory regions of alleles


497 of a gene either share most of their heat-responsive elements or lack such elements altogether.


498 We restricted this analysis to syntelog groups in which alleles differ in sequence and are


499 present as exactly one copy on at least two haplotypes, because only for these can reads be


500 assigned unambiguously to individual alleles. This conservative selection may enrich for


501 conserved genes, so the proportion of divergent allele pairs genome-wide could be higher. In


502 a potato study across a range of abiotic and biotic stresses, alleles of the same gene were


503 assigned to different coexpression network modules, pointing to divergent allelic regulation


504 across conditions &#91;20&#93; . Beyond potato, divergent stress responses between gene copies have so


505 far been described mainly for homoeologs in allopolyploids. In hexaploid wheat, homoeologs


506 from the three subgenomes showed extensive expression bias and expression partitioning


507 during heat and drought acclimation &#91;36&#93; . In the allopolyploid grass _Brachypodium hybridum_, the


508 early transcriptional response to prolonged heat was driven by homoeologs inherited from the


509 heat-tolerant progenitor _B. stacei_ &#91;37&#93; . In allotetraploid _Brassica napus_, homoeologs differed in


510 expression under heat, cold and drought, with expression biased towards the ‘A’ subgenome


511 under all conditions &#91;38&#93; . Unlike homoeologs, which derive from the genomes of distinct


512 progenitor species, the haplotypes of autotetraploid potato are alleles of a single genome and


513 are less divergent, which may explain why most allele pairs in Désirée shared their heat


514 response. However, not all haplotypes of cultivated potato are equally divergent, as some stem


515 from introgressions of wild Solanum species, many introduced through breeding for disease


516 resistance &#91;39&#93; . Such wild-derived haplotypes may carry _cis_ -regulatory elements that differ from


517 those of the cultivated gene pool and could account for some of the few allele pairs with a


518 divergent heat response.


519


520 Identifying alleles that respond differently to environmental stress can reveal favorable


521 variants. Using luciferase assays, we confirmed the differential heat response of one such


522 allele pair in a potato N²-acetylornithine deacetylase homolog most similar to Arabidopsis


23


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


523 _NAOD/CGP1_ (AT4G17830) (Fig. 3f). AT4G17830 was originally reported to function in


524 ornithine biosynthesis &#91;40&#93; and was subsequently identified as a cysteinylglycine (Cys-Gly)


525 dipeptidase involved in the cytosolic glutathione (GSH) degradation pathway &#91;41&#93; . The promoter


526 alleles differed in their heat inducibility relative to their basal activity under control conditions.


527 Given that Arabidopsis mutants with reduced glutathione levels show impaired acclimation to


528 high temperatures &#91;42&#93;, haplotype-dependent regulation of this gene could contribute to


529 differences in GSH turnover and redox homeostasis under heat stress in potato. These findings


530 therefore highlight _NAOD_ alleles as candidate loci for further investigation of the genetic basis


531 of heat tolerance and their potential utility in breeding heat-resilient cultivars.


532


533 Long-read RNA-seq and phased reference genomes also resolve expression changes within


534 gene families, such as the two strongly heat-responsive families examined here: heat shock


535 proteins and Kunitz-type protease inhibitors. The heat shock motif TTCTAGAA, which was also


536 enriched among strongly heat-upregulated alleles (Fig. 3b) and previously identified as an HSF


537 motif in tomato &#91;43&#93;, was present within an ATAC-seq peak in almost all HSP promoters (Fig. 4a).


538 This motif corresponds to a gapless variant of the canonical heat shock element


539 (AGAAnnTTCT), lacking the spacer nucleotides between the two inverted repeats. This


540 suggests that the motif also drives heat-induced expression in potato under prolonged heat.


541


542 Kunitz-type protease inhibitors are among the most abundant defense proteins in potato,


543 where they inhibit serine and cysteine proteases and accumulate in response to wounding and


544 pathogen attack &#91;44–46&#93; . Copy number of this family varies between the four Désirée haplotypes,


545 and several haplotypes share identical alleles, yet each gene copy could still be quantified


546 separately (Fig. 4b,c). Several, but not all, members of this family were strongly upregulated


547 under prolonged heat. Cysteine protease inhibitor 8 ( _CPI8_, Soltu.DM.03G018520) was the


548 most highly expressed member and showed the strongest heat induction, and we identified a


549 _MYC2_ binding site in an accessible region of its promoter (Supplementary Fig. S6), consistent


24


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


550 with its reported jasmonate-dependent regulation &#91;26&#93; . Short-read RNA-seq could not resolve


551 these copies, as most of their reads map to multiple alleles.


552


553 Alternative splicing is an important mechanism for regulating the heat stress response in


554 plants &#91;9&#93; . For genes with differential isoform usage, all haplotypes responded similarly


555 (Supplementary Table S16). Heat-induced isoform switching was therefore largely shared


556 among alleles, mirroring the uniform allelic response at the expression level. This contrasts


557 with allopolyploid crops, where homoeologs often show partitioned splicing responses to heat,


558 as in wheat &#91;47,48&#93; and Brassica napus &#91;38&#93; .


559


560 In wheat, alternative splicing of the ortholog _TaHSFA6e_ under heat stress produces two main


561 transcripts; the longer, carrying an additional 14-aa C-terminal amphipathic helix, shows


562 enhanced transcriptional activity on _HSP70_ &#91;49&#93; . We similarly observed a heat-induced increase


563 in an _HSFA6B_ isoform with a different C-terminus predicted to form an amphipathic α-helix


564 (Fig. 5d, Supplementary Fig. S8), suggesting that this splicing event may be conserved.


565 In tomato, temperature-dependent splicing of _HSFA2_ yields a cytosolic isoform involved in


566 thermomemory and a nuclear isoform driving the early heat response, and intronic variants


567 affecting this splicing co-segregate with thermotolerance &#91;50&#93; . Similarly, the two isoforms of the


568 DNAJ/HSP40 chaperone Soltu.DM.03G037350 were predicted to localize to different


569 subcellular compartments (Supplementary Table S17). Confocal microscopy confirmed this:


570 isoform 1, with a shorter C-terminus, localized to the cytoplasm; isoform 2, induced in heat


571 (Fig. 5i), localized to the nucleus (Fig. 5j, Supplementary Fig. S9). In humans, the same pattern


572 occurs in a DNAJ/HSP40: the longer DNAJB6a isoform is nuclear, whereas DNAJB6b, which


573 lacks an NLS, is cytoplasmic &#91;51&#93; . _DNAJB6a_ overexpression also increases nuclear import of


574 another protein &#91;52&#93; . In rice, salt stress induces nuclear import of the chaperone OsDNAJ15,


575 where, together with OsBAG4, it enhances DNA binding of the transcription factor OsMYB106


576 and thereby promotes salt tolerance &#91;53&#93; . We therefore hypothesize that the nuclear isoform of


25


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


577 potato HSP40 may similarly facilitate nuclear import of other proteins or transcription factor


578 binding, and thereby contribute to gene regulation.


579


580 **CONCLUSIONS**

581 In summary, long-read RNA-seq mapped to a haplotype-resolved Désirée genome resolves


582 allele- and isoform-level expression that short-read sequencing fails to do accurately. Allele

583 resolved expression data can therefore help identify haplotypes that confer heat tolerance for


584 potato breeding.


585


26


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


586 **SUPPLEMENTARY FILES**


587 **Supplementary Tables**
588 Supplementary Table S1: Sequencing and mapping statistics.


589 Supplementary Table S2: Heat-responsive genes, short-read RNA-seq, DM reference.


590 Supplementary Table S3: Heat-responsive genes, long-read RNA-seq, DM reference.


591 Supplementary Table S4: Heat-responsive genes, short-read RNA-seq, haplotype-resolved


592 reference.


593 Supplementary Table S5: Heat-responsive genes, long-read RNA-seq, haplotype-resolved


594 reference.


595 Supplementary Table S6: Functional enrichment of heat-responsive genes.


596 Supplementary Table S7: Synteny classification of genes.


597 Supplementary Table S8: Allelic imbalance under control conditions.


598 Supplementary Table S9: Comparison of allelic imbalance between sequencing


599 technologies.


600 Supplementary Table S10: Promoter structural variants between haplotypes.


601 Supplementary Table S11: Allelic expression of shared promoters in Désirée and Atlantic.


602 Supplementary Table S12: Missing haplotypes in Désirée and Atlantic for alleles with


603 identical promoters.


604 Supplementary Table S13: Motif enrichment in promoters of heat-induced alleles.


605 Supplementary Table S14: Allele-level heat response.


606 Supplementary Table S15: Allele-specific major isoforms.


607 Supplementary Table S16: Heat-induced differential isoform usage.


608 Supplementary Table S17: Predicted subcellular localization of isoforms.


609 Supplementary Table S18: gBlock sequences.


610


611


27


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


612 **Supplementary Figures**
613 Supplementary Fig. S1: Novel genes and isoforms and corrected gene models identified by


614 long-read RNA-seq.


615 Supplementary Fig. S2: Correlation of log2FC between long- and short-read RNA-seq for


616 genes grouped by transcript length.


617 Supplementary Fig. S3: Comparison of long- and short-read allelic quantification.


618 Supplementary Fig. S4: Promoter divergence of balanced and imbalanced allele pairs.


619 Supplementary Fig. S5: Transposable element composition of promoter structural variants.


620 Supplementary Fig. S6: MYC2 binding site within an ATAC-seq peak upstream of a Kunitz

621 type protease inhibitor gene.


622 Supplementary Fig. S7: Allele-specific isoform expression of ICME (Soltu.DM.02G008950) in


623 Atlantic.


624 Supplementary Fig. S8: Predicted protein structures of Désirée HSFA6B isoforms.


625 Supplementary Fig. S9: Separate channels of the confocal microscopy images in Fig. 5j.


626 Supplementary Fig. S10: RNA quality and batch-effect correction of long- and short-read


627 RNA-seq.


628


629 **Supplementary Notes**
630 Supplementary Note S1: Comparison of allele-specific expression between long- and short

631 read RNA-seq.


632


633 **AUTHOR CONTRIBUTIONS**

634 Conceptualization: N.N., K.G., M.P. Methodology: N.N., V.L., M.P. Software: N.N. Formal


635 analysis: N.N. Investigation: N.N., V.L., T.L., K.S. Resources: T.G. Data curation: N.N., V.L.


636 Visualization: N.N. Writing – original draft: N.N. Writing – review &amp; editing: all authors.


637 Supervision: K.G., M.P. Project administration: K.G., M.P. Funding acquisition: K.G., M.P. All


638 authors read and approved the final manuscript version.


639


28


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


640 **COMPETING INTERESTS**

641 The authors declare no competing interests.


642


643 **ACKNOWLEDGEMENTS**

644 The authors wish to thank Marijke Woudsma and Doretta Boomsma from HZPC Research


645 for providing the Désirée plantlets for this study and Klára Panzarová and Lamis Abdelhakim


646 for conducting the experiments at Photon Systems Instruments (PSI) Research Center


647 (Drásov, Czech Republic). We acknowledge the research infrastructures ELIXIR Slovenia


648 and IC Planta (National Institute of Biology) for providing computational resources and plant

649 growth facilities. The authors would like to acknowledge support of the National Genomics


650 Infrastructure (NGI) for providing assistance in massively parallel sequencing and


651 computational infrastructure. Work performed at NGI has been funded by RFI and Science


652 for Life Laboratory, Sweden.


653 Moreover, we thank Sophia Sonnewald, Markus Teige and the extended LongTREC network


654 for fruitful discussions.


655


656 **FUNDING**

657 This work has received funding from the European Union’s Horizon 2020 research and


658 innovation programme under the Marie Skłodowska-Curie Actions Doctoral Network


659 “LongTREC” grant agreement no. 101072892. This work was further supported by the


660 European Union’s Horizon 2020 research and innovation programme through the project


661 “ADAPT” grant agreement no. 862858 and by the Slovenian Research and Innovation


662 Agency (ARIS) through the research grants P4-0463 and J4-70165.


663


664 **METHODS**

665 _**Plant material and growth conditions**_


666 Potato plants ( _Solanum tuberosum_ cv. Désirée) were grown as described in Zagorščak et al.


667 (2025). Briefly, in vitro cuttings were planted in 250 ml pots with peat substrate (Klasmann 2)


668 and, after two weeks, transplanted into 3 L pots with peat substrate mixed with sand (3:1).


29


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


669 After 32 days of cultivation, plants were assigned to control and heat groups, and leaves


670 were sampled on day 7 of treatment, i.e. 39 days after the in vitro cuttings were planted in


671 substrate. Control plants were cultivated at 22/19 °C (day/night) and heat-treated plants at


672 30/28 °C (day/night). In both groups, watering was maintained up to 60% field capacity, light


673 intensity was set at 330 µmol m⁻² s⁻¹ PPFD with a 16 h photoperiod, and relative humidity


674 was maintained at 55%. Four biological replicates per condition were collected. For each


675 replicate, the second and third fully developed leaves were harvested and flash-frozen in


676 liquid nitrogen, and leaf tissue was homogenized and aliquoted.


677 _Nicotiana benthamiana_ plants were grown from seeds, and individual seedlings were


678 transferred to pots after two weeks. Throughout their growth, plants were maintained at


679 22/19 °C (day/night), 55% relative humidity, and an irradiance of 85 µmol m⁻² s⁻¹ PPFD


680 (400–700 nm). For heat treatment, plants were maintained at 28/28 °C (day/night) and 55%


681 relative humidity for 5 days prior to the luciferase assay.


682


683 _**RNA extraction**_


684 Total RNA was extracted and DNase-treated from 80 mg of frozen, homogenized leaf tissue


685 using MagMAX Plant RNA Isolation Kit (Thermo Fisher Scientific) on a KingFisher Apex


686 purification system (Thermo Fisher Scientific). The same RNA samples were previously


687 sequenced with Illumina in 2 × 150 bp paired-end mode &#91;28&#93; . For long-read RNA-seq, full-length


688 cDNA libraries were prepared with the Kinnex full-length RNA kit (PacBio) according to the


689 Iso-Seq protocol and sequenced on three SMRT Cells on a PacBio Revio system at the


690 National Genomics Infrastructure, Sweden, yielding full-length transcript reads.


691


692 _**Molecular cloning**_


693 Genomic DNA (gDNA) was isolated from potato cv. Désirée leaves using the DNeasy Plant


694 Mini Kit (Qiagen) per the manufacturer’s instructions. The _NAOD_ (Soltu.DM.08G022800)


695 promoter fragment (1,500 bp upstream of the start codon) was amplified from gDNA by


696 touchdown nested PCR with Phusion High-Fidelity DNA Polymerase (New England Biolabs)


30


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


697 using primer pairs Pept-H3-F (5′-caccTACACTGAGGCAAACAAACATCT-3′; forward for


698 haplotype 3) or Pept-H4-F (5′-caccGAACATGTGTAACTACACTGAG-3′; forward for


699 haplotype 4) and Pept-CDS-R (5′-GATAACCAGTGGTCCGCCA-3′; reverse for both


700 haplotypes) followed by PCR using primer pairs Pept-H3-F or Pept-H4-F and Pept-H3-R (5′

701 GTTAATAATATGGGAATGAAACTAGT-3′; reverse for haplotype 3) or Pept-H4-R (5′

702 GTTAATAATATGGGAATGAAAGTGC-3′; reverse for haplotype 4). A shorter variant of the


703 haplotype 4 promoter (605 bp upstream of the start codon) was amplified using forward


704 primer 5′-CACCAAAGGGTTAGTTTTTCAAACTGAAC-3′ and reverse primer Pept-H4-R.


705 Coding regions of _HSP40_ isoforms 1 and 2 and coding sequence of mTagBFP2 were


706 ordered as gBlocks from IDT (Supplementary Table S18) with overlapping ends and


707 recombined and amplified in a single PCR reaction with primers HSP40-gene-F (5′

708 CACCATGGTAGACAAAAGCAATGATTA-3′) and mTagBFP2-stop-R (5′

709 TTAATTAAGCTTGTGCCCCAG-3′). Amplicons were gel-purified using Wizard SV Gel and


710 PCR Clean-Up System (Promega) and cloned into pJET1.2 blunt cloning vector (Thermo


711 Fisher Scientific) according to the manufacturer’s protocol. Chemically competent _E. coli_


712 were transformed with the constructs, plasmids were isolated using the GenElute Plasmid


713 Miniprep Kit (Sigma) and sequenced using the above-listed primers (GATC Biotech). The


714 promoter regions were subsequently amplified from pJET1.2 blunt cloning vector using the


715 above-listed primers and ligated into pENTR/D-TOPO (Thermo Fisher Scientific) following


716 the manufacturer’s protocol. The promoter was recombined from pENTR/D-TOPO into the


717 binary destination vector pGWB435 &#91;54&#93; by LR reaction (Gateway LR Clonase II Enzyme Mix,


718 Thermo Fisher Scientific) following the manufacturer’s protocol. HSP40-mTagBFP2 was


719 recombined from pENTR/D-TOPO into the binary destination vector pK7WG2 (VIB)


720 downstream of CaMV 35S promoter.


721


722 _**Transient transformation**_


723 Homemade electrocompetent _Agrobacterium tumefaciens_ LBA4404 cells were transformed


724 with the prepared constructs by electroporation as described before &#91;55&#93; . Overnight cultures of


31


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


725 the transformed Agrobacterium were centrifuged for 10 min at 4,000 rpm at room


726 temperature, and the resulting pellets were resuspended in agroinfiltration solution (10 mM


727 MgCl₂, 10 mM MES, and 0.15 mM acetosyringone) to an OD₆₀₀ of 0.3. The prepared


728 Agrobacterium suspensions were incubated with gentle shaking in the dark at room


729 temperature for 2–4 h and subsequently used for agroinfiltration of fully developed leaves of


730 3–5-week-old _N. benthamiana_ plants.


731


732 _**Luciferase assay**_


733 Luciferase assays were performed according to a modified version of the transactivation


734 assay protocol described previously &#91;56&#93; . Briefly, _N. benthamiana_ plants agroinfiltrated with


735 Agrobacterium cultures carrying the _NAOD_ promoter variants upstream of the luciferase


736 coding sequence were sampled 5 days after heat treatment or control conditions and 4 days


737 after agroinfiltration. Leaf discs were collected using a 0.6-cm-diameter cork borer and


738 placed abaxial side up in white 96-well microtiter plates (Corning) containing MS liquid


739 medium supplemented with 35 µM beetle luciferin (Promega). Luminescence was measured


740 at 30-min intervals using a Centro LB 963 luminometer (Berthold), and the time course of


741 luminescence was plotted for each construct and condition in relative light units normalized


742 to the maximum measured per experiment. For each construct and condition, 24 leaf discs


743 were analyzed. The experiment was performed twice independently. For each construct, the


744 peak luminescence of each of the replicate wells (maximum over the time course) was


745 compared between the haplotype 3 and haplotype 4 promoters with a two-sided Wilcoxon


746 rank-sum test, separately for control and heat, and p-values were adjusted for multiple


747 testing with the Benjamini–Hochberg method &#91;57&#93; .


748


749 _**Confocal imaging**_


750 _N. benthamiana_ plants carrying transiently expressed HSP40-mTagBFP2 constructs


751 (isoforms 1 and 2 of _HSP40_ ) were imaged 3 days after infiltration using the TCS LSI confocal


752 macroscope (Leica) equipped with an ACS APO 20×/0.60 IMM objective. HSP40-mTagBFP2


32


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


753 fluorescence was detected using the 405 nm laser line at 95% power, with an emission


754 window of 440–500 nm and a detector gain of 650 V. Chlorophyll fluorescence was detected


755 using the 488 nm laser line at 90% power, with an emission window of 670–730 nm and a


756 detector gain of 600 V. Images were acquired as Z-stacks encompassing epidermal cells at a


757 voxel size of 5 µm, with the pinhole set to 1 Airy unit, and subsequently presented as


758 maximum-intensity projections. Detailed information on image acquisition is deposited on


759 Zenodo. HSP40 localization was assessed in two leaves from three plants, each set of plants


760 infiltrated with two independent Agrobacterium colonies, and consistent localization patterns


761 were observed.


762


763 _**Reference genome annotation preprocessing and read mapping**_


764 For alignment to a linear reference genome, the reference genome v6.1 of the doubled


765 monoploid potato DM 1-3 516 R44 &#91;11&#93; with the UniTato annotation &#91;23&#93; was used. Haplotype

766 resolved Désirée reference genome v1 &#91;24&#93; and haplotype-resolved Atlantic reference genome


767 v3 &#91;25&#93; with gene annotations transferred from UniTato using Liftoff &#91;58&#93; were used for haplotype

768 resolved analyses. We used the Liftoff annotation rather than the native Désirée annotation


769 because the native annotation was generated independently for each haplotype, resulting in


770 larger differences in gene models and UTR lengths between alleles; such differences would


771 confound comparisons of allelic expression. Transferring a single annotation to all haplotypes


772 provides consistent gene models across alleles. Liftoff was run separately for each


773 haplotype, providing a chromosome mapping file (-chroms), minimum alignment coverage


774 and sequence identity thresholds of 0.9 (-a 0.9, -s 0.9), and the -copies option to annotate


775 additional gene copies.


776


777 Short-read RNA-seq data were processed with the nf-core/rnaseq pipeline v3.27.0 &#91;59,60&#93;


778 (parameters --pseudo_aligner salmon and --keepDuplicates) and mapped to the DM and


779 Désirée reference genomes.


780


33


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


781 Long reads were mapped to the DM reference genome and to the haplotype-resolved


782 Désirée reference genome with the longrnaseq pipeline v1.0.1 &#91;61&#93; run in Nextflow v24.10.6


783 with the parameter --technology PacBio. The pipeline performs alignment to the genome and


784 transcriptome with minimap2 v2.29-r1283, novel isoform discovery with bambu v3.4.0 &#91;62&#93;, and


785 transcript quantification with oarfish v0.9 &#91;63&#93; . For comparisons of allele-specific expression


786 under control conditions, reads were mapped to transcripts with UTRs removed to avoid


787 quantification bias arising from differences in UTR lengths between alleles &#91;13&#93; . ONT long-read


788 cDNA data from cultivar Atlantic were processed in the same way, using --technology ONT


789 with the corresponding reference genome &#91;25&#93; . Aligned reads were tagged with haplotypes


790 based on primary alignment and visualized with JBrowse 2 &#91;64&#93; .


791


792 _**Batch correction &amp; differential gene expression analysis**_


793 Salmon &#91;65&#93; and oarfish transcript counts were imported into R v4.4.3 using tximport &#91;66&#93; with


794 transcript-to-gene mappings. Long-read RNA-seq counts aligned to the DM and haplotype

795 resolved reference genomes were batch-corrected with ComBat_seq from the sva package


796 v3.54.0 &#91;67&#93; using default parameters; the batches used for correction were those visible in RNA


797 concentration (Supplementary Fig. S10a). Batch correction increased the correlation of


798 log 2 FC between long- and short-read RNA-seq (Supplementary Fig. S10b). Short-read RNA

799 seq counts were not batch-corrected.


800 For the DM reference genome alignment, genes with fewer than 3 counts per million (CPM)


801 in at least 5 samples were removed prior to analysis. For the haplotype-resolved reference


802 genome, alleles with fewer than 4 CPM in at least 6 samples were removed. Differential


803 expression between heat and control conditions was assessed using the edgeR::voomLmFit


804 function from edgeR v4.4.0 &#91;68,69&#93; with a design matrix modeling the two conditions. An


805 empirical Bayes approach (eBayes) was applied, and differentially expressed genes were


806 defined by an adjusted p-value &lt; 0.05 and |log 2 FC| &gt; 1. Correlation of the edgeR log 2 FC


807 between long-read and short-read samples was calculated with cor(method = "pearson") for


808 batch-corrected and uncorrected counts.


34


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


809


810 Functional enrichment of the resulting differentially expressed gene (DEG) sets was


811 performed with clusterProfiler::enricher function &#91;70&#93; using the UniTato MapMan v4 ontology


812 gene set file (GSEA-compatible, gene level) as the annotation term set, with the background


813 set to all genes tested in the DE analysis. For DEGs using the haplotype-resolved reference,


814 genes that had at least one allele differentially expressed were used. Enriched terms were


815 considered significant at BH-adjusted p-value &lt; 0.05.


816


817 _**Motif enrichment analysis**_


818 Motif enrichment analysis was performed using SEA from the MEME Suite v5.5.9 &#91;71&#93; .


819 Promoter sequences were extracted 2,000 bp upstream of the annotated transcription start


820 site (TSS) for two allele sets: heat-induced alleles (log 2 FC &gt; 2, FDR &lt; 0.05) and non

821 responding alleles (−0.1 &lt; log 2 FC &lt; 0.1). Differential motif enrichment between the two sets


822 was assessed using SEA with the JASPAR CORE (2026) plants database &#91;43&#93; using default


823 parameters.


824


825 _**Syntelog identification**_


826 Syntelog genes were identified using GENESPACE v1.3.1 &#91;72&#93; within the SyntelogFinder


827 pipeline (v1.0.0) &#91;73&#93; . In brief, genes were first grouped into orthogroups by OrthoFinder2


828 v3.1.5 &#91;74&#93;, and collinear syntenic blocks were identified with MCScanX &#91;75&#93; . Genes were defined


829 as syntelogs when they were both orthologous (assigned to the same orthogroup) and


830 syntenic (located within a shared collinear block); the syntelogs of one gene across the four


831 haplotypes form a syntelog group.


832


833 _**Allelic expression**_


834 Allelic and isoform expression were analyzed with PolyASE v1.3.6 &#91;29&#93; ; PolyASE functions used


835 for each step are given in parentheses below. For gene-level allele-specific expression


836 analysis under control conditions, syntelog groups were retained


35


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


837 (filter_low_expressed_genes) if they had at least 40 raw unique read counts per syntelog


838 group across all replicates under control conditions, were present on all four haplotypes


839 without copy number variation, had a multimapping rate &lt; 0.15 (number of multimapping


840 counts per syntelog group/total counts per syntelog group; calculate_multi_ratios), and had


841 the same transcript length and the same number of transcripts per allele. Allelic ratios were


842 calculated from unique counts (calculate_allelic_ratios, calculate_per_allele_ratios) and


843 tested for imbalance (test_allelic_ratios_within_conditions). Syntelog groups were considered


844 imbalanced if the FDR-corrected p-value was below 0.05 and the ratio difference above 0.1


845 from the expected 0.25. Short-read RNA-seq counts were processed in the same manner,


846 except that syntelog groups with fewer than 400 salmon em counts were excluded and


847 instead of unique counts, em (expectation maximized) counts were tested for imbalance. For


848 pairwise comparisons (test_allelic_ratios_pairwise), syntelog groups with exactly one copy


849 on at least two haplotypes were tested for imbalance, and alleles with FDR &lt; 0.05 and


850 |log 2 FC| &gt; 1 were considered significantly different.


851 For cultivar Atlantic, pairwise comparison of allelic expression within leaf control conditions


852 was conducted in the same way as for Désirée for syntelog groups in which all members had


853 more than 40 uniquely mapping reads. For genes tested for pairwise allelic imbalance in both


854 cultivars, promoter sequences 1,500 bp upstream of the TSS were extracted and identical


855 promoters (100% sequence identity and 1,500 bp alignment length) between Atlantic and


856 Désirée were identified by BLASTn v2.12.0 &#91;76&#93; . Pearson correlation of log 2 FC was calculated


857 between expression of shared promoter pairs.


858


859 _**Unique mapping ratios**_


860 Unique and ambiguous read counts per transcript were obtained from the oarfish and salmon


861 outputs for one representative sample each, mapped against both the collapsed DM


862 reference and the haplotype-resolved tetraploid Désirée Liftoff annotation. Analysis was


863 restricted to single-isoform genes, and genes with &lt;100 total supporting reads were


864 excluded. For each gene, the unique read fraction was calculated as unique / (unique +


36


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


865 ambiguous) reads, and its distribution was compared between long- and short-read RNA

866 seq, separately for the collapsed and haplotype-resolved references.


867


868 _**Structural variant identification and promoter analysis**_


869 Individual haplotype sequences of the Désirée v1 assembly were aligned pairwise using


870 minimap2 v2.29-r1283 (asm20 preset, -k15 -s50 -z1000 -r3k,100k -O3,20 --eqx), and SVs


871 were called with SyRI v1.7.0 (-F B) &#91;77&#93; . For this analysis, promoter regions were defined as


872 2,000 bp upstream of the TSS on the reference as well as query haplotype. SVs were


873 retained if their reference coordinates overlapped the promoter of one gene and their query


874 coordinates overlapped the promoter of a different gene on the same chromosome; cross

875 chromosomal SVs, SVs lacking query coordinates, and inversions &gt;2,000 bp were excluded.


876 SVs (≥ 50 bp) comprised insertions, deletions, and highly divergent regions (HDRs), i.e.,


877 aligned regions with high sequence dissimilarity, as annotated by SyRI. Small variants (&lt; 50


878 bp; SNPs and indels) were taken from the same SyRI output and assigned to promoter pairs


879 using the same coordinate criteria. Promoter pairs without any called SV, SNP, or indel were


880 considered identical. Differences in the number of SVs and in the distance of SVs to the TSS


881 between balanced and imbalanced promoter pairs were assessed; for promoter pairs without


882 SVs, the number of SNPs and indels and the distance of the first indel to the TSS were


883 compared in the same way. All comparisons were assessed using a one-sided Mann–


884 Whitney U test, with the alternative hypothesis that imbalanced promoter pairs carry more


885 variants than balanced pairs and that their variants lie closer to the TSS.


886


887 _**Transposable element analysis**_


888 For each promoter containing exactly one SV longer than 49 bp, the SV sequence was


889 extracted from SyRI VCF output and queried via BLASTn (e-value ≤ 1×10⁻⁵) against a


890 combined EDTA TE library &#91;24&#93; built from all four Désirée v1 haplotypes to identify SVs with


891 homology to known transposable elements.


892


37


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


893 _**Differential isoform usage**_


894 For isoform-level analysis, genes with a minimum of 100 counts across replicates were


895 retained (filter_low_expressed_genes). Oarfish em counts were used instead of unique


896 counts so that genes with identical sequences on all haplotypes were not excluded. Testing


897 was performed separately for each haplotype: for each isoform, the usage ratio (isoform


898 count / gene total count) was compared between heat and control conditions using a beta

899 binomial likelihood-ratio test (test_isoform_DIU_between_conditions, which uses


900 betabinom_lr_test from isotools2 &#91;78,79&#93;, and p-values were Benjamini–Hochberg corrected. A


901 gene was considered to show differential isoform usage on a given haplotype if at least one


902 of its isoforms had an FDR-corrected p-value below 0.05 and an absolute difference in usage


903 ratio between heat and control conditions greater than 0.1.


904


905 _**Differential isoform structure between haplotypes**_


906 For each syntelog group, the major (or, as fallback, minor) isoform of the most highly


907 expressed haplotype was matched to its structural counterpart on each other haplotype by


908 exon/intron length similarity, and isoform usage (isoform count / haplotype total count) was


909 compared between haplotypes using a beta-binomial likelihood-ratio test


910 (test_differential_isoform_structure). Syntelog groups with FDR &lt; 0.05 and a usage-ratio


911 difference &gt; 0.2 were called significant.


912


913 _**Pairwise allelic response to heat**_


914 For each pair of alleles within a syntelog group, we tested whether their relative expression


915 shifted between control and heat by modeling the ratio of one allele's count to the pair's


916 combined count (i / (i + j)) in each condition, using unique long-read counts, with a beta

917 binomial likelihood-ratio test (test_allelic_ratios_between_conditions, which uses


918 betabinom_lr_test from isotools2 &#91;78,79&#93; ). Pairs with zero combined coverage in either condition


919 were excluded. P-values were Benjamini–Hochberg corrected across all tested pairs.


920 Δlog 2 FC was defined as the difference between the heat response of the two alleles,


38


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


921 Δlog 2 FC = log 2 FC(i) − log 2 FC(j), where log 2 FC is the heat/control fold change of each allele;


922 this is equivalent to the change in the allelic log 2 ratio (i/j) between heat and control. An allele


923 pair was called as showing a divergent heat response when FDR &lt; 0.05 and |Δlog 2 FC| &gt; 1.


924


925 _**Prediction of differences between isoforms**_


926 Isoform subcellular localization was predicted with DeepLoc 2.1 &#91;80&#93; and functional domains


927 with InterProScan 110.0 &#91;81,82&#93; via their web servers. Structures were compared with AlphaFold


928 Server 2026.08.19 &#91;83&#93; . Multiple sequence alignments of promoters and transcripts were done


929 with ClustalW &#91;84&#93; and DiGAlign &#91;85&#93; .


930


931 _**ATAC-seq**_


932 ATAC-seq samples of control and heat-treated Désirée leaves are publicly available and


933 described in &#91;28&#93; .Paired-end 150 bp reads were processed with the nf-core/atacseq pipeline


934 v2.1.2 &#91;59,86&#93; against the Désirée v1 assembly (scaffolds and organellar sequences removed),


935 retaining multimapping reads with --keep_multi_map to preserve signal in regions identical


936 between haplotypes; peaks were called with MACS2 &#91;87&#93; and normalized coverage tracks were


937 used for promoter-accessibility analyses.


938


939 _**Promoter motif scanning and chromatin accessibility profiling**_


940 To visualize the relationship between promoter architecture, chromatin accessibility, and heat


941 responsiveness, we selected genes annotated as heat shock proteins (HSPs), chaperones,


942 or heat shock-related transcription factors, retaining those for which at least one haplotype


943 showed a log 2 FC greater than 1 in response to heat. Promoter sequences spanning 2,000 bp


944 upstream and 500 bp downstream of the transcription start site were extracted and scanned


945 for transcription factor binding motifs using position weight matrices derived from JASPAR


946 CORE (2026) plants &#91;43&#93; entries for SolycHSFA7 (MA2727.1), HSFA6B (MA1664.2), HSFB2A


947 (MA1665.2), SolycHSFC1 (MA2728.1), HSFB4 (MA2020.1), SolycHSFB2b (MA2711.1) and


948 MYC2 (Myc2/la2, MA2736.1). Motif matching was performed with Biostrings &#91;88&#93; by scanning


39


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


949 both the forward and reverse-complement strands at a minimum score threshold of 90% of


950 the maximum PWM score, and the start position, strand, and score of each hit were


951 recorded. In parallel, ATAC-seq signal was extracted from a BigWig track generated from


952 combined heat-sample reads, summarized in 25 bp bins, and normalized per promoter. Motif


953 positions and ATAC-seq signal were then displayed together with allele-specific log 2 FC for


954 the selected genes and haplotypes.


955 **DATA AVAILABILITY**

956 All sequencing data of Désirée leaves under control and prolonged heat conditions are


957 available in the ENA under BioProject PRJEB127277. The long-read Kinnex RNA-seq data


958 were generated in this study (run accessions ERR18072167–ERR18072174). Short-read


959 RNA-seq data and ATAC-seq data of the same conditions, generated previously &#91;28&#93; are


960 available under run accessions ERR18076900–ERR18076907 and ERR18072205–


961 ERR18072212.


962 Publicly available ONT long-read RNA-seq data of Atlantic were obtained from the SRA


963 (SRR14993894, SRR14993895, SRR14996168).


964 Subsamples of the long-read RNA-seq, the GFF containing novel isoforms and bigwigs of


965 ATAC-seq have been added to the Désirée genome browser (https://desiree.nib.si).


966 The updated annotation (GFF) of the haplotype-resolved Désirée genome, including novel


967 genes and isoforms identified by long-read RNA-seq, and the raw confocal microscopy


968 images of HSP40-mTagBFP2 localization in _N. benthamiana_, together with the full image


969 acquisition settings, are available on Zenodo 10.5281/zenodo.23594697.


970


971


972 **CODE AVAILABILITY**

973 Analysis scripts are available on GitHub: https://github.com/NIB-SI/LR-DESIREE

974 &#91;reproducibility.&#93;(https://github.com/NIB-SI/LR-DESIREE-reproducibility)


975


976


40


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


977 **REFERENCES**

978 1. Koch, L., Lehretz, G. G., Sonnewald, U. &amp; Sonnewald, S. Yield reduction caused by


979 elevated temperatures and high nitrogen fertilization is mitigated by SP6A overexpression


980 in potato ( _Solanum tuberosum_ L.). _The Plant Journal_ **117**, 1702–1715 (2024).


981 2. Rivero, R. M., Mittler, R., Blumwald, E. &amp; Zandalinas, S. I. Developing climate‐resilient


982 crops: improving plant tolerance to stress combination. _The Plant Journal_ **109**, 373–389


983 (2022).


984 3. FAO. FAOSTAT: Crops. https://www.fao.org/faostat/en/#data/QCL (2024).


985 4. Hancock, R. D. _et al._ Physiological, biochemical and molecular responses of the potato (


986 _S olanum tuberosum_ L .) plant to moderately elevated temperature. _Plant Cell &amp;_


987 _Environment_ **37**, 439–450 (2014).


988 5. Levy, D. &amp; Veilleux, R. E. Adaptation of potato to high temperatures and salinity-a


989 review. _Amer J of Potato Res_ **84**, 487–506 (2007).


990 6. Guo, M. _et al._ The Plant Heat Stress Transcription Factors (HSFs): Structure, Regulation,


991 and Function in Response to Abiotic Stresses. _Front. Plant Sci._ **7**, (2016).


992 7. Kan, Y., Mu, X.-R., Gao, J., Lin, H.-X. &amp; Lin, Y. The molecular basis of heat stress


993 responses in plants. _Molecular Plant_ **16**, 1612–1634 (2023).


994 8. Bakery, A. _et al._ Heat stress transcription factors as the central molecular rheostat to


995 optimize plant survival and recovery from heat stress. _New Phytologist_ **244**, 51–64


996 (2024).


997 9. Ling, Y., Mahfouz, M. M. &amp; Zhou, S. Pre-mRNA alternative splicing as a modulator for


998 heat stress response in plants. _Trends in Plant Science_ **26**, 1153–1170 (2021).


999 10. Shang, L. _et al._ Construction of heat stress regulation networks based on Illumina and


1000 SMRT sequencing data in potato. _Front. Plant Sci._ **14**, 1271084 (2023).


1001 11. Pham, G. M. _et al._ Construction of a chromosome-scale long-read reference genome


1002 assembly for potato. _GigaScience_ **9**, giaa100 (2020).


41


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1003 12. Yang, H. _et al._ Haplotype-resolved genomics in horticultural crops: advances, challenges,


1004 and perspectives. _Horticulture Research_ uhag326 (2026) doi:10.1093/hr/uhag326.


1005 13. Nolte, N. _et al._ The promise of long-read RNA-seq: reducing bias in analyses of allele


1006 imbalance. _NAR Genomics and Bioinformatics_ **8**, lqag071 (2026).


1007 14. Chen, Y. _et al._ A systematic benchmark of Nanopore long-read RNA sequencing for


1008 transcript-level analysis in human cell lines. _Nat Methods_ **22**, 801–812 (2025).


1009 15. Fang, C. _et al._ Dynamics of accessible chromatin regions and subgenome dominance in


1010 octoploid strawberry. _Nat Commun_ **15**, 2491 (2024).


1011 16. Trapero‐Mozos, A. _et al._ Engineering heat tolerance in potato by temperature‐dependent


1012 expression of a specific allele of _HEAT‐SHOCK COGNATE 70_ . _Plant Biotechnology_


1013 _Journal_ **16**, 197–207 (2018).


1014 17. Alger, E. I. &amp; Edger, P. P. One subgenome to rule them all: underlying mechanisms of


1015 subgenome dominance. _Current Opinion in Plant Biology_ **54**, 108–113 (2020).


1016 18. Bird, K. A., VanBuren, R., Puzey, J. R. &amp; Edger, P. P. The causes and consequences of


1017 subgenome dominance in hybrids and recent polyploids. _New Phytologist_ **220**, 87–93


1018 (2018).


1019 19. Pham, G. M. _et al._ Extensive genome heterogeneity leads to preferential allele expression


1020 and copy number‐dependent expression in cultivated potato. _The Plant Journal_ **92**, 624–


1021 637 (2017).


1022 20. Brose, J. _et al._ An allelic resolution gene atlas for tetraploid potato provides insights into


1023 tuberization and stress resilience. _The Plant Journal_ **124**, e70557 (2025).


1024 21. Feke, A. _et al._ High resolution diel transcriptomes of autotetraploid potato reveal


1025 expression and sequence conservation among rhythmic genes. _BMC Genomics_ **26**, 925


1026 (2025).


42


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1027 22. Wang, J. _et al._ Genetic architecture of sugarcane traits in a polyploid genomics


1028 framework. _Nature_ https://doi.org/10.1038/s41586-026-10576-7 (2026)


1029 doi:10.1038/s41586-026-10576-7.


1030 23. Zagorščak, M. _et al._ Evidence-based unification of potato gene models with the UniTato


1031 collaborative genome browser. _Front. Plant Sci._ **15**, 1352253 (2024).


1032 24. Godec, T. _et al._ Haplotype-resolved genome assembly of the tetraploid potato cultivar


1033 Désirée. _Sci Data_ **12**, 1044 (2025).


1034 25. Hoopes, G. _et al._ Phased, chromosome-scale genome assemblies of tetraploid potato


1035 reveal a complex genome, transcriptome, and predicted proteome landscape underpinning


1036 genetic diversity. _Molecular Plant_ **15**, 520–536 (2022).


1037 26. Levak, V. Strategija razvoja biosenzorjev imunskega odziva v krompirju. (2020).


1038 27. Zagorščak, M. _et al._ Integration of multi-omics data and deep phenotyping provides


1039 insights into responses to single and combined abiotic stress in potato. _Plant Physiology_


1040 **197**, kiaf126 (2025).


1041 28. Godec, T. _et al._ Combined heat and drought stress alters chromatin accessibility and


1042 primes potato abscisic acid biosynthesis. Preprint at


1043 https://doi.org/10.64898/2026.11.11.765576 (2026).


1044 29. Nolte, N., Gruden, K. &amp; Petek, M. LongPolyASE: An end-to-end framework for allele

1045 specific gene and isoform analysis in polyploids using long-read RNA-seq. Preprint at


1046 https://doi.org/10.64898/2026.02.03.703480 (2026).


1047 30. Larson, N. B. _et al._ Comprehensively Evaluating cis-Regulatory Variation in the Human


1048 Prostate Transcriptome by Using Gene-Level Allele-Specific Expression. _The American_


1049 _Journal of Human Genetics_ **96**, 869–882 (2015).


1050 31. Vořechovský, I. From benign to pathogenic variants and vice versa: pyrimidine transitions


1051 at position -3 of TAG and CAG 3’ splice sites. _J Hum Genet_ **70**, 125–133 (2025).


43


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1052 32. Huizinga, D. H., Omosegbon, O., Omery, B. &amp; Crowell, D. N. Isoprenylcysteine


1053 Methylation and Demethylation Regulate Abscisic Acid Signaling in _Arabidopsis_ . _The_


1054 _Plant Cell_ **20**, 2714–2728 (2008).


1055 33. Sun, H. _et al._ The phased pan-genome of tetraploid European potato. _Nature_ **642**, 389–


1056 397 (2025).


1057 34. Liu, J. _et al._ RIN4 Functions with Plasma Membrane H+-ATPases to Regulate Stomatal


1058 Apertures during Pathogen Attack. _PLoS Biol_ **7**, e1000139 (2009).


1059 35. Mackey, D., Holt, B. F., Wiig, A. &amp; Dangl, J. L. RIN4 Interacts with Pseudomonas


1060 syringae Type III Effector Molecules and Is Required for RPM1-Mediated Resistance in


1061 Arabidopsis. _Cell_ **108**, 743–754 (2002).


1062 36. Liu, Z. _et al._ Temporal transcriptome profiling reveals expression partitioning of


1063 homeologous genes contributing to heat and drought acclimation in wheat (Triticum


1064 aestivum L.). _BMC Plant Biol_ **15**, 152 (2015).


1065 37. Takahagi, K. _et al._ Homoeolog-specific activation of genes for heat acclimation in the


1066 allopolyploid grass _Brachypodium hybridum_ . _GigaScience_ **7**, giy020 (2018).


1067 38. Lee, J. S. &amp; Adams, K. L. Global insights into duplicated gene expression and alternative


1068 splicing in polyploid _Brassica napus_ under heat, cold, and drought stress. _The Plant_


1069 _Genome_ **13**, e20057 (2020).


1070 39. Hardigan, M. A. _et al._ Genome diversity of tuber-bearing _Solanum_ uncovers complex


1071 evolutionary history and targets of domestication in the cultivated potato. _Proc. Natl._


1072 _Acad. Sci. U.S.A._ **114**, (2017).


1073 40. Molesini, B. _et al._ The Arabidopsis _N_ -Acetylornithine Deacetylase Controls Ornithine


1074 Biosynthesis via a Linear Pathway with Downstream Effects on Polyamine Levels. _Plant_


1075 _Cell Physiol_ pcw167 (2016) doi:10.1093/pcp/pcw167.


44


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1076 41. Miyaji, S. _et al._ _N_ _&#91;2&#93;_ ‐Acetylornithine deacetylase functions as a C YS ‐G LY dipeptidase in


1077 the cytosolic glutathione degradation pathway in _Arabidopsis thaliana_ . _The Plant Journal_


1078 **118**, 1603–1618 (2024).


1079 42. Dard, A. _et al._ Glutathione-mediated thermomorphogenesis and heat stress responses in


1080 _Arabidopsis thaliana_ . _Journal of Experimental Botany_ **74**, 2707–2725 (2023).


1081 43. Ovek Baydar, D. _et al._ JASPAR 2026: expansion of transcription factor binding profiles


1082 and integration of deep learning models. _Nucleic Acids Research_ **54**, D184–D193 (2026).


1083 44. Gruden, K. _et al._ Potato cysteine proteinase inhibitor gene family: molecular cloning,


1084 characterisation and immunocytochemical localisation studies. _Plant Mol Biol_ **34**, 317–


1085 323 (1997).


1086 45. Odeny, D. A., Stich, B. &amp; Gebhardt, C. Physical organization of mixed protease inhibitor


1087 gene clusters, coordinated expression and association with resistance to late blight at the


1088 _StKI_ locus on potato chromosome III. _Plant Cell &amp; Environment_ **33**, 2149–2161 (2010).


1089 46. Valueva, T. A., Revina, T. A., Kladnitskaya, G. V. &amp; Mosolov, V. V. Kunitz‐type


1090 proteinase inhibitors from intact and _Phytophthora_ ‐infected potato tubers. _FEBS Letters_


1091 **426**, 131–134 (1998).


1092 47. Liu, Z. _et al._ Global profiling of alternative splicing landscape responsive to drought, heat


1093 and their combination in wheat ( _Triticum aestivum_ L.). _Plant Biotechnology Journal_ **16**,


1094 714–726 (2018).


1095 48. Lu, Y. _et al._ Alternative Splicing Diversified the Heat Response and Evolutionary


1096 Strategy of Conserved Heat Shock Protein 90s in Hexaploid Wheat (Triticum aestivum


1097 L.). _Front. Genet._ **11**, 577897 (2020).


1098 49. Wen, J. _et al._ Alternative splicing of _T_ _A_ _HSFA6_ _E_ modulates heat shock protein–mediated


1099 translational regulation in response to heat stress in wheat. _New Phytologist_ **239**, 2235–


1100 2247 (2023).


45


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1101 50. Hu, Y. _et al._ Natural variation in HsfA2 pre‐mRNA splicing is associated with changes in


1102 thermotolerance during tomato domestication. _New Phytologist_ **225**, 1297–1310 (2020).


1103 51. Hanai, R. &amp; Mashima, K. Characterization of two isoforms of a human DnaJ homologue,


1104 HSJ2. _Mol Biol Rep_ **30**, 149–153 (2003).


1105 52. Pei, Y. _et al._ A Hsp40 Chaperone Protein Interacts with and Modulates the Cellular


1106 Distribution of the Primase Protein of Human Cytomegalovirus. _PLoS Pathog_ **8**,


1107 e1002968 (2012).


1108 53. Liu, Y. _et al._ Plasma membrane‐localized Hsp40/DNAJ chaperone protein facilitates


1109 OsSUVH7‐OsBAG4‐OsMYB106 transcriptional complex formation for _OsHKT1;5_


1110 activation. _JIPB_ **65**, 265–279 (2023).


1111 54. Nakagawa, T. _et al._ Improved Gateway Binary Vectors: High-Performance Vectors for


1112 Creation of Fusion Constructs in Transgenic Analysis of Plants. _Bioscience,_


1113 _Biotechnology, and Biochemistry_ **71**, 2095–2100 (2007).


1114 55. Lukan, T. _et al._ Chloroplast redox state changes mark cell‐to‐cell signaling in the


1115 hypersensitive response. _New Phytologist_ **237**, 548–562 (2023).


1116 56. Lasierra, P. &amp; Prat, S. Transient Transactivation Studies in Nicotiana benthamiana


1117 Leaves. in _Two-Hybrid Systems_ (ed. Oñate-Sánchez, L.) vol. 1794 311–322 (Springer


1118 New York, New York, NY, 2018).


1119 57. Benjamini, Y. &amp; Hochberg, Y. Controlling the False Discovery Rate: A Practical and


1120 Powerful Approach to Multiple Testing. _Journal of the Royal Statistical Society Series B:_


1121 _Statistical Methodology_ **57**, 289–300 (1995).


1122 58. Shumate, A. &amp; Salzberg, S. L. Liftoff: accurate mapping of gene annotations.


1123 _Bioinformatics_ **37**, 1639–1643 (2021).


1124 59. Ewels, P. A. _et al._ The nf-core framework for community-curated bioinformatics


1125 pipelines. _Nat Biotechnol_ **38**, 276–278 (2020).


46


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1126 60. Harshil Patel _et al._ nf-core/rnaseq: nf-core/rnaseq v3.27.0 - Vanadium Vulture. Zenodo


1127 https://doi.org/10.5281/ZENODO.1400710 (2026).


1128 61. Nolte, N. NIB-SI/longrnaseq: v1.0.1. Zenodo https://doi.org/10.5281/ZENODO.17631221


1129 (2025).


1130 62. Chen, Y. _et al._ Context-aware transcript quantification from long-read RNA-seq data with


1131 Bambu. _Nat Methods_ **20**, 1187–1195 (2023).


1132 63. Zare Jousheghani, Z., Singh, N. P. &amp; Patro, R. Oarfish : enhanced probabilistic modeling


1133 leads to improved accuracy in long read transcriptome quantification. _Bioinformatics_ **41**,


1134 i304–i313 (2025).


1135 64. Diesh, C. _et al._ JBrowse 2: a modular genome browser with views of synteny and


1136 structural variation. _Genome Biol_ **24**, 74 (2023).


1137 65. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. &amp; Kingsford, C. Salmon provides fast


1138 and bias-aware quantification of transcript expression. _Nat Methods_ **14**, 417–419 (2017).


1139 66. Soneson, C., Love, M. I. &amp; Robinson, M. D. Differential analyses for RNA-seq:


1140 transcript-level estimates improve gene-level inferences. _F1000Res_ **4**, 1521 (2015).


1141 67. Zhang, Y., Parmigiani, G. &amp; Johnson, W. E. ComBat-seq: batch effect adjustment for


1142 RNA-seq count data. _NAR Genomics and Bioinformatics_ **2**, lqaa078 (2020).


1143 68. Chen, Y., Chen, L., Lun, A. T. L., Baldoni, P. L. &amp; Smyth, G. K. edgeR v4: powerful


1144 differential analysis of sequencing data with expanded functionality and improved support


1145 for small counts and larger datasets. _Nucleic Acids Research_ **53**, gkaf018 (2025).


1146 69. Law, C. W., Chen, Y., Shi, W. &amp; Smyth, G. K. voom: precision weights unlock linear


1147 model analysis tools for RNA-seq read counts. _Genome Biol_ **15**, R29 (2014).


1148 70. Yu, G. Thirteen years of clusterProfiler. _The Innovation_ **5**, 100722 (2024).


1149 71. Bailey, T. L. &amp; Grant, C. E. SEA: Simple Enrichment Analysis of motifs. Preprint at


1150 https://doi.org/10.1101/2021.08.23.457422 (2021).


47


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1151 72. Lovell, J. T. _et al._ GENESPACE tracks regions of interest and gene copy number


1152 variation across multiple genomes. _eLife_ **11**, e78526 (2022).


1153 73. Nolte, N. NIB-SI/syntelogfinder: Inital release. Zenodo


1154 https://doi.org/10.5281/ZENODO.17627949 (2025).


1155 74. Emms, D. M. &amp; Kelly, S. OrthoFinder: phylogenetic orthology inference for comparative


1156 genomics. _Genome Biol_ **20**, 238 (2019).


1157 75. Wang, Y. _et al._ MCScanX: a toolkit for detection and evolutionary analysis of gene


1158 synteny and collinearity. _Nucleic Acids Research_ **40**, e49–e49 (2012).


1159 76. Ye, J., McGinnis, S. &amp; Madden, T. L. BLAST: improvements for better sequence


1160 analysis. _Nucleic Acids Research_ **34**, W6–W9 (2006).


1161 77. Goel, M., Sun, H., Jiao, W.-B. &amp; Schneeberger, K. SyRI: finding genomic rearrangements


1162 and local sequence differences from whole-genome assemblies. _Genome Biol_ **20**, 277


1163 (2019).


1164 78. Bi, Y., Lankenau, T. L., Lienhard, M. &amp; Herwig, R. IsoTools 2.0: Software for


1165 Comprehensive Analysis of Long-read Transcriptome Sequencing Data. _Journal of_


1166 _Molecular Biology_ **437**, 169049 (2025).


1167 79. Lienhard, M. _et al._ IsoTools: a flexible workflow for long-read transcriptome sequencing


1168 analysis. _Bioinformatics_ **39**, btad364 (2023).


1169 80. Ødum, M. T. _et al._ DeepLoc 2.1: multi-label membrane protein type prediction using


1170 protein language models. _Nucleic Acids Research_ **52**, W215–W220 (2024).


1171 81. Blum, M. _et al._ InterPro: the protein sequence classification resource in 2025. _Nucleic_


1172 _Acids Res_ **53**, D444–D456 (2025).


1173 82. Blum, M., Hobbs, E., Florentino, L. &amp; Bateman, A. InterProScan 6: a modern large-scale


1174 protein function annotation pipeline. _Bioinform Adv_ **6**, vbag141 (2026).


48


&#91;bioRxiv preprint doi: https://doi.org/10.64898/2026.11.14.766623; this version posted November 16, 2026. The copyright holder for this&#93;(https://doi.org/10.64898/2026.11.14.766623)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
&#91;perpetuity. It is made available under a CC-BY 4.0 International license.&#93;(http://creativecommons.org/licenses/by/4.0/)


1175 83. Abramson, J. _et al._ Accurate structure prediction of biomolecular interactions with


1176 AlphaFold 3. _Nature_ **630**, 493–500 (2024).


1177 84. Larkin, M. A. _et al._ Clustal W and Clustal X version 2.0. _Bioinformatics_ **23**, 2947–2948


1178 (2007).


1179 85. Nishimura, Y., Yamada, K., Okazaki, Y. &amp; Ogata, H. DiGAlign: Versatile and Interactive


1180 Visualization of Sequence Alignment for Comparative Genomics. _Microb. Environ._ **39**,


1181 n/a (2024).


1182 86. Harshil Patel _et al._ nf-core/atacseq: &#91;2.1.2&#93; - 2022-08-07. Zenodo


1183 https://doi.org/10.5281/ZENODO.2634132 (2023).


1184 87. Zhang, Y. _et al._ Model-based Analysis of ChIP-Seq (MACS). _Genome Biol_ **9**, R137


1185 (2008).


1186 88. H. Pagès, P. A. Biostrings. Bioconductor


1187 https://doi.org/10.18129/B9.BIOC.BIOSTRINGS (2017).


1188


49


