# Source paper archive

- Source title: Accumulation of two distinct alleles causes astringency loss in cultivated polyploid persimmon
- Author(s): Fujiwara, Y., Nishiyama, S., Onoue, N., Matsuzaki, R., Kono, A., Azuma, A., Saito, T., Shimizu, T., Sato, A., Ichino, T., Yamane, H., Yonemori, K., Tao, R.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.10.31.762479v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.10.31.762479v1.full.pdf
- PDF SHA-256: 55b268a16221d1ae6e8eca99eb9de192ac7ccbdc655abbdccceae11d52e99b26
- 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.

---

[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in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

# **Accumulation of two distinct alleles causes astringency loss in** **cultivated polyploid persimmon**


**Yosuke Fujiwara** **[1 ]** **†‡, Soichiro Nishiyama** **[1, 2 ]** ***†, Noriyuki Onoue** **[3]** **, Ryusuke Matsuzaki** **[3]** **,**
**Atsushi Kono** **[3]** **, Akifumi Azuma** **[3]** **, Toshihiro Saito** **[3]** **, Takeo Shimizu** **[3]** **, Akihiko Sato** **[3]** **§,**
**Takuji Ichino** **[4]** **, Hisayo Yamane** **[1]** **, Keizo Yonemori** **[5]** **and Ryutaro Tao** **[1 ]**

1 Graduate School of Agriculture, Kyoto University, Kyoto, Kyoto, Japan
2 Research Center for Agricultural Information Technology, National Agriculture and Food
Research Organization, Tsukuba, Ibaraki, Japan
3 Grape and Persimmon Station, Institute of Fruit Tree and Tea Science, National Agriculture
and Food Research Organization, Higashi-Hiroshima, Hiroshima, Japan
4 Laboratory of Medicinal Cell Biology, Kobe Pharmaceutical University, Kobe, Hyogo, Japan
5 Ryukoku Extension Center (REC), Ryukoku University, Otsu, Shiga, Japan

 - Correspondence: nishiyama.soichiro.8e@kyoto-u.ac.jp; +81-75-753-6051

  - These authors contributed equally to this work.

 - Present address: Faculty of Agriculture, University of Miyazaki, Miyazaki, Miyazaki, Japan.
**§** Present address: Regional Revitalization and Agricultural Research Institute, Kindai
University, Yuasa, Wakayama, Japan.


**Abstract**


Persimmon fruit accumulates high levels of proanthocyanidins (PAs), causing extreme
astringency. In pollination-constant and non-astringent (PCNA) cultivars, PA accumulation
ceases early in fruit development. Although the PCNA trait is recessively controlled by the
_ASTRINGENCY_ ( _AST_ ) locus, its causal gene has remained unknown. Using haplotype-resolved
forward genetics in a hexaploid genetic background, we identified the previously
uncharacterized _TANNIN_ ( _TAN_ ) as the causal gene. Long-read sequencing revealed that PCNA
originated through the accumulation of independently derived _ast_ alleles that cause
transcriptional deficiency of _TAN_ . These findings resolve the complex genetic basis of natural
astringency loss in persimmon and provide a framework for molecular breeding. Further
investigation of TAN function would uncover mechanisms underlying massive accumulation
of PAs in persimmon fruit and insight into specialized metabolite sequestration.


1


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Introduction**


Proanthocyaindins (PAs), also known as condensed tannins, are widespread specialized
metabolites in land plants. PAs strongly influence food quality by causing astringency, while
also providing potential health benefits through their antioxidant properties and microbiota
modulation [1, 2] . PAs are polymers of flavan-3-ols and are highly reactive and potentially
cytotoxic; therefore, their intracellular sequestration is critical for maintaining cellular integrity.
Although there have been much progress in understanding the molecular basis of PA
biosynthesis, the mechanisms underlying PA polymerization and intracellular trafficking into
the vacuole remain unresolved [3, 4] . This gap limits our understanding of how plants achieve the
quantitative control of PA accumulation.

Persimmon ( _Diospyros kaki_ Thunb.; 2 _n_ = 6 _x_ = 90) is a fruit tree crop that originated in
east Asia. A unique characteristic of persimmon is the massive accumulation of PAs, which
causes strong astringency of the fruit. Persimmon cultivars have long been classified into
pollination-constant and astringent (PCA), pollination-variant and astringent (PVA),
pollination-variant and non-astringent (PVNA) and pollination-constant and non-astringent
(PCNA) [5, 6] . In non-PCNA cultivars, astringency loss in mature fruit is associated with seedderived volatile compounds [7] and is quantitatively regulated. On the other hand, PCNA cultivars
are unique in that PA accumulation ceases early during fruit development, and genes related to
PA biosynthesis decrease coinciding with the cessation of PA accumulation [5, 8, 9] . PCNA
phenotype is conferred by a single locus called _ASTRINGENCY_ ( _AST_ ), and PCNA is recessive
homozygous ( _aaaaaa_ ) [10, 11] . This natural loss of astringency is a commercially essential trait in
modern persimmon breeding. However, PCNA germplasm remains rare and genetically narrow,
consistent with their presumed recent origin from a limited founder background [12] . Elucidation
of the causal gene will therefore provide both a foundation for molecular breeding and a unique
framework for dissecting mechanisms underlying continuous and massive PA accumulation.
An _ast_ -linked insertion, Indel-3, has been found and intensively used for marker-assisted
selection of PCNA offspring [13, 14] . However, the causal gene of _AST_ locus has not been
identified due to the complexity of polyploid persimmon genome. We previously delimited the
_AST_ region using a segregating population of hexaploid _D. kaki_ and found that the
corresponding interval spans 915 kb, including Indel-3, in the diploid _D. lotus_ (2 _n_ = 2 _x_ =30)
genome [15] . However, no gene within this interval showed clear association with PA
biosynthesis or differential expression between PCNA and non-PCNA genotypes [15, 16] . These
results suggested that a haplotype-resolved analysis of the hexaploid _D. kaki_ genome is
required to fully characterize the candidate genes and allelic variations at the _AST_ locus.


**Results**


                                              Identification of a single candidate responsible for PCNA trait by haplotype resolved hexaploid
genome analysis

We first established haplotype-resolved sequences covering the delimited _AST_ region
in the non-PCNA cultivar ‘Taiten’. Previous analyses revealed that ‘Taiten’ carries six distinct
haplotypes ( _A_ 224, _A_ 240, _a_ 349, _a_ 353_2a, _a_ 353_2b, and _a_ 355) at the Indel-3 region [17] . To enable
accurate haplotype reconstruction, we obtained 43.5X coverage of Pacbio HiFi reads from


2


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


‘Taiten’ and applied a trio-binning strategy, in which HiFi reads are partitioned into each
parental group prior to assembly [18] (Supplementary Fig. 1, Supplementary Table 1). In our study,
based on partitioning accuracy test of each parent-specific k-mer length (Supplementary Table
2), whole-genome reads were partitioned into three groups: hapA (derived from ‘Kurokuma’;
_A_ 224, _A_ 240 and _a_ 353_2a), hapB (derived from ‘Taishu’; _a_ 349, _a_ 353_2b and _a_ 355), and hapU
(unclassified).

Using k = 25 and 31 partitioning results, de novo assemblies were performed separately
for each parental group (Supplementary Table 3, Supplementary Table 4). As a result, assembly
in k = 25 failed to generate complete haplotype sequences (Supplementary Fig. 2). In k = 31,
five contigs covering the region were produced, and _A_ 224, _A_ 240, and _a_ 353_2a were generated
from the ‘Kurokuma’ assembly, whereas _a_ 349, _a_ 353_2a, and _a_ 353_2b were generated from
the ‘Taishu’ assembly (Fig. 1A). Notably, polymorphisms used for the _A_ 224- and _A_ 240specific markers [15] were confirmed in the corresponding contigs and were consistent across the
delimited region (Fig. 1A). However, _a_ 353_2a was assembled from both parental read groups,
and _a_ 355 was not assembled, indicating incomplete partitioning of reads specific to this
haplotype under the current parameters. Previous studies showed that _a_ 355 is highly similar to
_a_ 353_2b, differing only in short sequence repeat (SSR) length [17] . Consistent with this, no
sequence polymorphisms were detected between _a_ 353_2b and _a_ 355 within the HiFi read span
(~18 kb), except for the SSR (Supplementary Fig. 3). Therefore, we classified _a_ 353 and _a_ 355
into the same haplotype group for the following analysis. Nevertheless, the k = 31 assembly
resolved all sequence-distinguishable _AST_ haplotypes as independent contigs spanning the
delimited region. Moreover, no large structural variant was detected in these contigs compared
to the diploid _D. oleifera_ sequence [19] .

Using the haplotype-resolved sequences as references, we aimed to identify
polymorphisms that cosegregate with the presence of dominant _AST_ allele(s) in hexaploid _D._
_kaki_ . Whole genome sequencing data were obtained from 11 cultivars and from the 652-line
breeding population (‘Fuyu’ x ‘Yoshidagosho’). The paternal parent of the 652-line population,
‘Yoshidagosho’, carries the _A_ 353 haplotype at the _AST_ locus, which retains dominant _AST_
function despite carrying the Indel-3 insertion typically associated with recessive _ast_
haplotypes [20] . Because all reported _ast_ haplotypes carry Indel-3 insertion, _A_ 353 is considered
closely related to the _ast_ haplotypes while retaining _AST_ function (Supplementary Fig. 4).
Therefore, sequence information derived from the 652-line population is highly informative
for further refining candidate polymorphisms associated with PCNA trait.

A mapping-based variant calling approach, using the _A_ 224 and _A_ 240 haplotype
sequences as a reference, identified polymorphisms within the delimited _AST_ region that were
heterozygous in genotypes carrying functional _AST_ haplotype(s) and homozygous in genotypes
lacking them (Supplementary Fig. 5). Gene positions in this region were annotated by fulllength transcript (Iso-seq) mapping (Fig. 1B, Supplementary Table 5). Strikingly, candidate
polymorphisms were strongly concentrated in a genomic region that harbors a single gene
homologous to _CW7_, originally described in _Arabidopsis_ (AT1G59520) [21], although its
biological function has not been characterized. Most polymorphisms were mapped to intron
regions of this gene on the _A_ 240 reference (Supplementary Fig. 6). On the _A_ 224 reference,
polymorphisms were more broadly distributed across the interval, yet 42 of 146 variants also


3


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


concentrated within this gene (Supplementary Fig. 7). A few other genes detected using the
_A_ 240 reference containing candidate polymorphisms showed no differences between PCNA
and non-PCNA types in expression levels, and the detected polymorphisms do not influence
nonsynonymous substitution or the gene splicing (Supplementary Fig. 8A, Supplementary Fig.
9, Supplementary Fig. 10). Therefore, the homolog of _CW7_ was selected for further
investigation as the sole strong candidate underlying the _AST_ locus, and we designated this
gene as _TANNIN_ ( _TAN_ ).


Complex transcript variants in _DkTAN_ caused by two mutations defining distinct _ast_ alleles

Gene expression levels of _DkTAN_ were not associated with the PCNA phenotype
(Supplementary Fig. 8B). However, PCR analysis failed to detect cDNA encoding full-length
DkTAN protein in PCNA samples, suggesting that transcript structure, rather than total
expression level, may be associated with _DkTAN_ dysfunction. We therefore examined whether
_DkTAN_ dysfunction is a consistent feature of PCNA individuals by amplicon sequencing of
full-length cDNA from immature fruit. We first analyzed ‘Taiten’ (non-PCNA; _AAaaaa_ ) and
‘Kanshu’ (PCNA; _aaaaaa_ ). Both cultivars produced multiple transcript variants predicted to
encode truncated proteins. In ‘Taiten’, transcripts encoding full-length functional proteins
(389–392 aa) accounted for ~30% of total _DkTAN_ transcripts. In contrast, ‘Kanshu’ produced
almost no functional transcripts (Fig. 2A; 1.0 %).

To determine whether functional _DkTAN_ transcription corresponds with the
PCNA/non-PCNA phenotypes, we conducted amplicon sequencing across a comprehensive
set of persimmon cultivars and _Diospyros_ relatives (Supplementary Table. 6). Functional
transcripts were detected only at trace levels in all available 25 PCNA landraces and modern
cultivars (mean ± SE, 1.2 ± 0.14 %), whereas at least 25% of reads encoded full-length
functional proteins in 106 non-PCNA cultivars and six related species (mean ± SE, 60.7 ±
0.69 %; Fig. 2B, C). The same pattern was observed in a segregating population (Fig. 2D).
Moreover, the proportion of reads translating into full-length _DkTAN_ increased with the dosage
of the dominant _AST_ alleles [22] (Fig. 2E). Taken together, these data strongly support that the
presence or absence of functional _DkTAN_ transcripts is tightly associated with PCNA and nonPCNA phenotypes.

Comparison of the haplotype-resolved _DkTAN_ sequences with the _D. oleifera_ reference
genome revealed two large indels present in haplotypes _a_ 353_2a and _a_ 353_2b but absent in
_a_ 349 (Supplementary Fig. 11A). These included a 1.5-kb deletion that removes exon 7 and a
17-kb insertion within an intron. However, all representative PCNA cultivars are heterozygous
for the 1.5-kb deletion (Supplementary Fig. 11B), indicating that it cannot be the sole causal

mutation of all _ast_ exists in PCNA cultivars.


To clarify the allelic basis of _DkTAN_ dysfunction, we further analyzed the full-length
amplicon sequencing data from ‘Taiten’ and ‘Kanshu’. Sequence analysis showed that the 1.5kb deletion generates 180 aa, 205 aa, and 315 aa proteins through exon structure disruption
(Fig. 2A, Fig. 3). However, additional truncated variants (76 aa, 98 aa, 366 aa) were detected
in ‘Taiten’ and ‘Kanshu’ but not explained by the 1.5-kb deletion, indicating the presence of
other causal mutation(s) associated with a distinct _ast_ haplotype. From this point, we identified
an _a_ 349-specific SNP (C→G) located at the 5′ splice site of exon 3, which is predicted to induce

4


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


alternative splicing and exon skipping which cause 76 aa and 366 aa variants. The association
between the candidate polymorphisms and _a_ 349 was confirmed using siblings from an _a_ 349segregating population. Exon 3 SNP cosegregated with the _a_ 349, whereas the exon 7 deletion
was absent from the _a_ 349 (Supplementary Fig. 12A, B), consistent with the haplotype-resolved
genomic sequences (Supplementary Fig. 11). The predicted translation was verified by fulllength cDNA amplicon sequencing (Supplementary Fig. 12C). Nonetheless, the 98 aa
transcript variant is generated via exon reduction at the 3′ splice site, demonstrating that this
variant is a natural isoform. Together, these results indicate that at least two independent
mutations, the exon 7 deletion and a splice-site-related mutation around exon 3, lead to lossof-function _DkTAN_ alleles, each associated with a different _ast_ haplotype group represented by
_a_ 353_2a/ _a_ 353_2b/ _a_ 355 and _a_ 349, respectively (Fig. 3). Moreover, the identification of two
independent loss-of-function allelic series, provides strong genetic support that _DkTAN_ is the
causal gene underlying the _AST_ locus, consistent with established principles for inferring gene–
trait causality [23] .


     -     Long range haplotype phased analysis for genotyping the two causal mutations

Although we identified two candidate causal polymorphisms, genotyping them across
the germplasm collection required haplotype-phased approach because each polymorphism
appeared in the heterozygous state when analyzed individually (Supplementary Fig. 13).
However, the two candidate causal polymorphisms are separated by approximately 15 kb,
making their simultaneous phasing highly challenging with conventional approaches. To
achieve this, we adapted a target-enrichment approach combining probe-based hybridization
with long-read sequencing (Supplementary Fig. 14). A total of 21 PCNA and four non-PCNA
cultivars were analyzed, and a high enrichment efficiency was achieved within the _DkTAN_
region selected for probe design in all cultivars (Supplementary Table 7). Following de novo
assembly, haplotype sequences covering the _DkTAN_ region were successfully reconstructed
for each cultivar. Phylogenetic analysis together with the ‘Taiten’ haplotype assembly revealed
that these sequences could be classified into three major groups: _AST_, _a_ 349, and _a_ 353 (Fig.
4A). Examination of the phased haplotype sequences demonstrated that the SNP near exon 3
was exclusively associated with the _a_ 349 group, whereas the deletion spanning exon 7 was
specifically found in the _a_ 355 group (Fig. 4B). Importantly, all _DkTAN_ sequences identified in
PCNA cultivars belonged to either the _a_ 349 or _a_ 353 group. These phased genotyping results

demonstrated that PCNA cultivars carries either of the two causal mutations across six _DkTAN_

alleles, collectively accounting for the loss of functional _DkTAN_ copies. This strongly supports
the conclusion that these mutations are responsible for the functional deficiency of _DkTAN_ .


Independent _DkTAN_ mutations underlying PCNA while maintaining haplotype diversity of the
_AST_ locus

The clear phylogenetic separation between the _a_ 349 and _a_ 353 groups suggests that
these _ast_ haplotypes emerged independently (Fig. 4A). Notably, some _DkTAN_ haplotypes from
non-PCNA cultivars were closely related to the _a_ 349 or _a_ 355 groups despite lacking causal
mutations, suggesting the presence of intermediate haplotypes in non-PCNA. Haplotype


5


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


network analysis further revealed that _TAN_ homolog from _D. oleifera_ were more closely related
to _DkTAN_ than those from _D. lotus_, consistent with previous reports proposing _D. oleifera_ as
one of the closest diploid ancestors of _D. kaki_ [24] (Fig. 4C). In addition, it was found that _a_ 349
haplotype group was closer to _D. lotus_ and _D. oleifera_ sequences, suggesting earlier emergence
of _a_ 349 from _AST_ than _a_ 353. Together, these findings suggest that the PCNA trait became
established through the accumulation of multiple independently arisen _ast_ haplotypes within
the hexaploid genome, while diverse functional _AST_ haplotypes were retained in non-PCNA
cultivars (Supplementary Fig. 15). Consistent with this scenario, no evidence of a selective
sweep in PCNA cultivars was detected across the _AST_ locus, including the _DkTAN_ region (Fig.
4D), suggesting that sequence diversity has been maintained among _ast_ haplotypes. Thus, the
establishment of PCNA appears to have resulted from accumulation of independent _ast_ alleles
rather than fixation of a single mutational event.


**Discussion**

The massive accumulation of PAs in persimmon fruit represents one of the most
extreme examples of specialized metabolite accumulation among fruit crops. Cultivar
differences related to pollination response and fruit astringency have been recognized for more
than a century and remain central to persimmon classification and breeding [25] . Although the
_AST_ locus has long been recognized as the major determinant of the PCNA phenotype, the
underlying gene remained unknown because of the highly heterozygous hexaploid genome of
persimmon. Here, long-read-assisted forward genetics identified _DkTAN_ as the causal gene
underlying _AST_ and revealed that PCNA arose through at least two independently derived lossof-function haplotypes (Fig. 1, Fig. 2, Fig. 3). Because PCNA is a recessive trait in a hexaploid
species, breeding of PCNA cultivars has relied on repeated use of a limited number of elite
founders, leading to a narrow genetic background and inbreeding depression-like constraints
12 . In contrast, our results show that the _AST_ locus itself retains haplotype diversity through
multiple independent _DkTAN_ loss-of-function alleles. This finding suggests that the
unfavorable genetic constraints associated with PCNA breeding may not be explained solely
by reduced diversity at the _AST_ locus, and raises the possibility that loss of TAN function itself,
or genetic factors inherent to PCNA founder backgrounds, contributes to these constraints.
Thus, identification of _DkTAN_ provides both an evolutionary framework for the recent
establishment of PCNA cultivars and a molecular basis for more precise marker-assisted
breeding (Fig. 4) [12, 20] .

Given the distinctive pattern of PA accumulation in persimmon compared with other
plant species and its recently reported unique cellular sequestration system [26], TAN may play a
central role in the exceptionally high accumulation of PA specific to persimmon fruit. As TAN
has not been characterized in any plant species, further investigation of its molecular function
may provide new insight into both the mechanism underlying natural astringency loss in
persimmon and large-scale sequestration of specialized metabolites in plants.


6


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Materials and Methods**

**Plant materials and DNA sequencing**

Young leaves were collected from the persimmon cultivar ‘Taiten’ grown at the Grape
and Persimmon Research Station of the Institute of Fruit Tree and Tea Science, National

Agriculture and Food Research Organization (NIFTS; Higashi-Hiroshima, Japan). Nuclei were
isolated from leaf tissues following a modified version of the previous protocol [27] . Highmolecular-weight DNA was subsequently extracted using the Nanobind DNA Extraction Kit
(Circulomics), and DNA fragments shorter than 5 kb were removed using the Short Read
Eliminator Kit (Circulomics). A total of 70 μg of high-molecular-weight DNA was used for
library preparation and sequenced across seven SMRT Cells on the PacBio Sequel II platform.
HiFi reads (≥Q20) were generated from the resulting subreads using packages available in
pbbioconda (https://github.com/PacificBiosciences/pbbioconda). In total, approximately 185
Gb of HiFi reads were obtained from ‘Taiten’, corresponding to approximately 43× coverage
of the hexaploid persimmon genome.
For polymorphism analysis, 13 cultivars, including the parental cultivars of ‘Taiten’
(‘Kurokuma’ and ‘Taishu’), grown at the Kyoto University Experimental Farm and NIFTS,
were used. In addition, ten progeny individuals homozygous for Indel-3 from the ‘Fuyu’ ×
‘Yoshidagosho’ population (652-line), generated and maintained at NIFTS, were included in
the analysis. The ten progeny individuals were selected based on sensory evaluation data

obtained at NIFTS so that the selected set included five non-PCNA and five PCNA individuals.
Genomic DNA was extracted from young leaves using the CTAB method [28] . Short-read
sequencing data from the cultivar panel and the 652-line population were generated using the
Illumina platform. The resulting reads were subjected to quality filtering before downstream
analyses. The sequencing results and accession numbers were summarized in Supplementary

Table 1.

**Haplotype-resolved assembly of ‘Taiten’ genome**
Trio-binning was performed using a modified version of the publicly available pipeline
(https://github.com/esrice/trio_binning) [18, 29] . Short-read sequencing data from the parental
cultivars ‘Kurokuma’ and ‘Taishu’ were decomposed into k-mers using the find-unique-kmers
function, and parent-specific k-mers were identified. Independent analyses were conducted
using multiple k-mer lengths (k = 17, 19, 21, 23, 25, 27, 31, and 37). To evaluate the optimal
k-mer length, 254 HiFi reads spanning the 5R region [13] were manually assigned to their
corresponding haplotypes based on sequence polymorphisms. Using the classify_by_kmers
function, HiFi reads containing the 5R region were classified into the ‘Kurokuma’ haplotype
group (hapA), the ‘Taishu’ haplotype group (hapB), or an unclassified group (hapU). Reads
were assigned to hapA or hapB when they contained at least 10 parent-specific k-mers and
more than 90% of the parent-specific k-mers were derived from a single parent (Supplementary
Fig. 1). For each k-mer length, the numbers of correctly classified reads in hapA and hapB (a
and b, respectively) were summed, and the partition accuracy was calculated as (a + b)/254
(Supplementary Table 2).

Because k = 25 and k = 31 showed the highest partition accuracy (see main text), these
k-mer lengths were subsequently used for genome-wide classification of HiFi reads. The


7


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


classified read sets corresponding to the ‘Kurokuma’ haplotype group (hapA + hapU) and the
‘Taishu’ haplotype group (hapB + hapU) were independently assembled using hifiasm [30] with
the parameters --n-hap 3 and -l 0. Primary contigs corresponding to the _AST_ locus region were
then identified. The haplotype origin of each contig was determined using marker sequences
located within the 5R region. Finally, sequence similarity between the assembled contigs and
the _AST_ locus region of the _D. oleifera_ reference genome [19] was assessed using NUCmer from
the MUMmer package v3.23 [31] .

**Read mapping, variant calling, and identification of** _**AST**_ **-linked polymorphisms**
The workflow used for the identification of _AST_ -linked polymorphisms is summarized
in Supplementary Fig. 4. Short-read DNA sequencing data obtained from the cultivars and
populations described above were mapped independently to the _A_ 224 and _A_ 240 haplotype
sequences assembled from ‘Taiten’, focusing on the delimited _AST_ region [15] . Read mapping
was performed using BWA-MEM v0.7.17 [32] . BAM files were generated using SAMtools view
33 . For the 652-line population, BAM files from five PCNA individuals and five non-PCNA
individuals were merged separately. Variant calling detecting single nucleotide polymorphisms
(SNPs) and short insertions/deletions (short indels) was performed using SAMtools mpileup.
Variant call datasets from the cultivar panel and the 652-line population were subsequently
merged using custom Python scripts.

To eliminate false-positive variants originating from repetitive regions with excessive
read accumulation, coverage filtering was applied prior to variant screening. For ‘Kanshu’,
‘Taishu’, ‘Taiten’, and ‘Kurokuma’, only sites with read depths between 20× and 300× were
retained. For all other cultivars, a threshold of 50×–300× was applied. Sites outside these ranges
were excluded from further analyses. Among the retained variants, the candidate _AST_ -linked
polymorphisms were defined as sites that were homozygous in all PCNA cultivars (major allele
ratio = 100%) and heterozygous in non-PCNA cultivars (minor allele ratio ≥ 5%). To further
validate linkage with the phenotype, variant calls from the 652-line population were extracted
only at the candidate _AST_ -linked sites detected in the diversity panel. Variants were retained
when the major allele ratio was 100% in both ‘Fuyu’ and the pooled PCNA progeny, while the
minor allele ratio was at least 15% in both ‘Yoshidagosho’ and the pooled non-PCNA progeny.
Finally, variants were further filtered to retain only those for which PCNA cultivars
were homozygous for the alternative allele, defined as an alternative allele frequency of at least
99% relative to the reference sequence. These variants were designated as _AST_ -linked
polymorphisms. The distribution of identified _AST_ -linked polymorphisms across the delimited
_AST_ region in the _A_ 224 and _A_ 240 haplotype sequences was visualized using 1-kb genomic bins.

**Iso-Seq library preparation, sequencing, and analysis**

Mesocarp tissues were collected from ‘Taiten’ fruits harvested in June, July, and
August 2016. Total RNA was extracted using the hot borate method [34] . RNA quality was
assessed using a Qubit 4 Fluorometer (Invitrogen), and only samples with an RNA IQ value
greater than 9 were used for library preparation. Full-length transcript sequencing (Iso-Seq)
was performed on the PacBio Sequel II platform. Raw sequence data were processed using
SMRT Link (PacBio) to generate high-quality Iso-Seq reads for each sampling date.


8


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


Iso-Seq datasets obtained from the three developmental stages were merged and
mapped to the delimited _AST_ regions in the _A_ 224 and _A_ 240 haplotype sequences of ‘Taiten’
using minimap2 [35] . Mapped reads were retrieved using SAMtools view with the -F 2048 option
to remove supplementary alignments. To eliminate spurious alignments, reads were further
filtered by local BLAST analysis, retaining only those for which more than 50% of the
transcript sequence was aligned to the target region. The filtered reads were subsequently
remapped to the delimited _AST_ regions in the _A_ 224 and _A_ 240 haplotypes using minimap2.
Mapping results were visualized using Integrative Genomics Viewer (IGV) [36], and the positions
of expressed genes within the delimited _AST_ region were manually annotated based on
transcript alignments.

**Amplicon sequencing**
A comprehensive germplasm panel consisting of 131 persimmon cultivars, six related
_Diospyros_ accessions, and ten progeny individuals from the 652-line population were used for
amplicon sequencing analysis (Supplementary Table 6). Fruits were collected between late
June and early July, and total RNA was extracted from mesocarp tissues using the ISOSPIN
Plant RNA Kit (Nippon Gene) according to the manufacturer’s instructions with Assist Buffer
for ISOSPIN Plant RNA (Nippon gene). First-strand cDNA was synthesized from oligo (dT)purified RNA (total 1 µg) using SuperScript III Reverse Transcriptase (Invitrogen). The fulllength coding sequence of the candidate gene, including the start and stop codons, was
amplified by a first-round PCR using KOD One® PCR Master Mix (TOYOBO). Samplespecific index sequences were subsequently added in a second-round PCR. Primer sequences
are listed in Supplementary Table 8. To minimize amplification bias, the numbers of PCR
cycles were limited to 25 and 10 cycles for the first and second PCRs, respectively. The
resulting libraries were subjected to full-length amplicon sequencing on the PacBio Sequel
platform. Following demultiplexing based on index sequences, cDNA sequences were
translated into amino acid sequences using gotranseq (https://github.com/feliixx/gotranseq).
Translated sequences were further processed using custom Python scripts to remove amino
acid sequences downstream of premature stop codons. Amino acid sequences encoding fulllength proteins without premature stop codons were classified as functional transcripts.

**Target enrichment sequencing and haplotype analysis**

The workflow is summarized in Supplementary Fig. 15. Young leaves were collected
from four non-PCNA and 21 PCNA cultivars grown at the Kyoto University Experimental
Farm. Approximately 5 g of leaf tissue was used for genomic DNA extraction following the
CTAB method. Purified DNA was mechanically sheared using a Megaruptor [®] system
(Diagenode) according to the manufacturer's instructions. Fragment size distributions were
evaluated for representative samples using a Femto Pulse System (Agilent), confirming that
most DNA fragments were concentrated around 7–8 kb. Sequencing libraries were prepared
from the sheared DNA using the GenNext [®] NGS Library Prep Kit (TOYOBO), followed by
ligation of TruSeq Combinatorial Dual Index adapters (Illumina). Target enrichment was
performed using a custom probe panel, designed by Twist Bioscience, against the candidate
gene region in the ‘Taishu’ genome (Chr5: 16446515- 16509450) [37] . Capture enrichment was


9


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


conducted according to the Long Read Library Preparation and Standard Hyb v2 Enrichment
protocol (Twist Bioscience). Two independently prepared multiplex libraries were pooled and
sequenced on the PacBio Revio platform.
Following sequencing, reads were demultiplexed according to sample-specific index
sequences. HiFi reads longer than 5 kb were assembled de novo using hifiasm with the
parameters -l 0 and --n-hap 6. Primary contigs were aligned to the ‘Taiten’ haplotype sequences
( _A_ 240 and _a_ 353_2b) using minimap2, and contigs corresponding to the candidate gene region
were extracted using custom Python scripts. Contigs longer than 20 kb for the _A_ 240 haplotype
and longer than 40 kb for the _a_ 353_2b haplotype were retained and combined for subsequent
analyses. The selected sequences were aligned using MAFFT v7 [38] with the --auto option. The
resulting multiple sequence alignment was subjected to phylogenetic analysis using IQ-TREE
3 [39], with automatic model selection (-m MFP) and 1,000 ultrafast bootstrap replicates (-B 1000).
Phylogenetic trees were visualized using the Interactive Tree of Life (iTOL;
https://itol.embl.de/).

For haplotype network analysis, the MAFFT alignment was converted to NEXUS
format and imported into DnaSP v6 [40] to define haplotype groups. The resulting data were
exported in RDF format and analyzed using NETWORK v10.2 (Fluxus Technology Ltd.,
https://www.fluxus-engineering.com/index.htm). Haplotype networks were constructed using
the median-joining algorithm.

**Sequence diversity analysis**
BAM files generated for the polymorphism analysis from nine PCNA cultivars (‘Fuyu’,
‘Okugosho’, ‘Kanshu’, ‘Jiro’, ‘Hanagosho’, ‘Fujiwaragosho’, ‘Mikado’, ‘Midai’, and
‘Taishu’) and four non-PCNA cultivars (‘Taiten’, ‘Tenryubo’, ‘Kuromuka’, and
‘Yoshidagosho’) were used for variant calling with SAMtools mpileup and VarScan v2.3 [41, 42],
using the ‘Taiten’ _A_ 224 haplotype sequence spanning the delimited _AST_ region as the reference.
In the resultant Variant Call Format (VCF) file, biallelic SNPs were initially filtered using
VCFtools v0.1.17 [43] with the options (--minDP 20 --max-meanDP 200 --max-missing 0.8 --mac
4 --min-alleles 2 --max-alleles 2 --remove-indels). Because VarScan is not designed for
hexaploid genotype calling, genotype calls were further refined using custom Python scripts
based on allele frequencies. Genotypes with alternative-allele read frequencies below 0.3%
were converted to homozygous reference calls, whereas those above 99.7% were converted to
homozygous alternative calls. Genotypes with alternative-allele read frequencies between
0.3% and 5% or between 95% and 99.7% were masked as missing to reduce ambiguous lowfrequency calls, and those between 5% and 95% were treated as heterozygous. Population
genetic statistics, including _F_ ST and nucleotide diversity (π), were then calculated for the PCNA
and non-PCNA populations using VCFtools in non-overlapping 10-kb windows.

**Data availability**

All sequencing data generated in this study have been deposited in the DDBJ Sequence

Read Archive under accession number PRJDB42921.


10


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**References**

1. P. M. Aron, J. A. Kennedy, Review Flavan-3-ols: Nature, occurrence and biological
activity. _Mol. Nutr. Food Res._ **52**, 79–104 (2008).
2. M. C. Rodríguez-Daza, E. C. Pulido-Mateos, J. Lupien-Meilleur, D. Guyonnet, Y.
Desjardins, D. Roy, Polyphenol-mediated gut Microbiota modulation: Toward prebiotics
and further. _Front. Nutr._ **8**, 689456 (2021).
3. R. A. Dixon, S. Sarnala, Proanthocyanidin Biosynthesis—A Matter of Protection. _Plant_
_Physiol._ **184**, 579–591 (2020).
4. K. Yu, Y. Song, J. Lin, R. A. Dixon, The complexities of proanthocyanidin biosynthesis
and its regulation in plants. _Plant Commun._ **4**, 100498 (2023).
5. K. Yonemori, J. Matsushima, Property of Development of the Tannin Cells in NonAstringent Type Fruits of Japanese Persimmon ( _Diospyros kaki_ ) and Its Relationship to
Natural Deastringency. _J. Japan. Soc. Hotic. Sci._ **54**, 201–208 (1985).
6. K. Yonemori, A. Sugiura, M. Yamada, Persimmon genetics and breeding. _Plant Breed._
_Rev._ **19**, 191–225(2000).
7. A. Sugiura, T. Tomana, Relationships of Ethanol Production by Seeds of Different Types
of Japanese Persimmons and Their Tannin Content. _HortScience_ **18**, 319–321 (1983).
8. T. Akagi, A. Ikegami, T. Tsujimoto, S. Kobayashi, A. Sato, A. Kono, K. Yonemori,
DkMyb4 is a Myb transcription factor involved in proanthocyanidin biosynthesis in
persimmon fruit. _Plant Physiol._ **151**, 2028–2045 (2009).
9. S. Nishiyama, N. Onoue, A. Kono, A. Sato, K. Yonemori, R. Tao, Characterization of a
gene regulatory network underlying astringency loss in persimmon fruit. _Planta_ **247**, 733–
743 (2018).
10. I. Ikeda, M. Yamada, A. Kurihara, T. Nishida, Inheritance of Astringency in Japanese
Persimmon. J. Japan. _Soc. Hotic. Sci._ **54**, 39–45 (1985).
11. M. Yamada, A. Sato, Segregation for fruit astringency type in progenies derived from
crosses of ‘Nishimurawase’ x pollination constant non-astringent genotypes in oriental
persimmon ( _Diospyros kaki_ Thunb.). _Sci. Hortic._ **92**, 107–111 (2002).
12. M. Yamada, A. Sato, Advances in table grape breeding in Japan. _Breed. Sci._ **66**, 34–45
(2016).
13. S. Kanzaki, T. Akagi, T. Masuko, M. Kimura, M. Yamada, A. Sato, N. Mitani, N.
Ustunomiya, K. Yonemori, SCAR markers for practical application of marker-assisted
selection in persimmon ( _Diospyros kaki_ Thunb.) breeding. _J. Japan. Soc. Hotic. Sci._ **79**,

```
    –
```

150 155 (2010).

14. Y. Fujiwara, S. Nishiyama, N. Onoue, A. Kono, A. Sato, K. Yonemori, R. Tao, Phenotypic
characterization of astringent persimmon lines with recessive homozygosity at the

```
                      –
```

_ASTRINGENCY_ marker locus. _Hort. J._ **94**, 148 161 (2025).

15. S. Nishiyama, N. Onoue, A. Kono, A. Sato, K. Ushijima, H. Yamane, R. Tao, K. Yonemori,
Comparative mapping of the _ASTRINGENCY_ locus controlling fruit astringency in
hexaploid persimmon ( _Diospyros kaki_ Thunb.) with the diploid _D. lotus_ reference genome.
_Hort. J._ **87**, 315–323 (2018).
16. Y. Fujiwara, S. Nishiyama, N. Onoue, R. Matsuzaki, K. Yonemori, R. Tao, Candidate
gene analysis for _ASTRINGENCY_ controlling fruit astringency in _Diospyros kaki_ based on


11


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


mRNA- and small RNA-sequencing analyses. _Acta Hortic._ **1338**, 269–276 (2022).
17. N. Onoue, S. Kobayashi, A. Kono, A. Sato, SSR-based molecular profiling of 237
persimmon ( _Diospyros kaki_ Thunb.) germplasms using an _ASTRINGENCY_ -linked marker.
_Tree Genet._ Genomes **14**, 28 (2018).
18. S. Koren, A. Rhie, B. P. Walenz, A. T. Dilthey, D. M. Bickhart, S. B. Kingan, S.
Hiendleder, J. L. Williams, T. P. L. Smith, A. M. Phillippy, De novo assembly of
haplotype-resolved genomes with trio binning. _Nat. Biotechnol._ **36**, 1174–1182 (2018).
19. Y. Suo, P. Sun, H. Cheng, W. Han, S. Diao, H. Li, Y. Mai, X. Zhao, F. Li, J. Fu, A highquality chromosomal genome assembly of _Diospyros oleifera_ Cheng. _Gigascience_ **9**, 1–
10 (2020).
20. N. Onoue, R. Matsuzaki, A. Azuma, T. Saito, T. Shimizu, A. Sato, Accurate marker
assisted selection for non-astringent persimmon using a CAPS marker to complement a
SCAR marker. _Euphytica_ **220**, 137 (2024).
21. A. Kato, M. Suzuki, A. Kuwahara, H. Ooe, K. Higano-Inaba, Y. Komeda, Isolation and
analysis of cDNA within a 300 kb Arabidopsis thaliana genomic region located around
the 100 map unit of chromosome 1. _Gene_ **239**, 309–316 (1999).
22. T. Akagi, Y. Takeda, K. Yonemori, A. Ikegami, A. Kono, M. Yamada, S. Kanzaki,
Quantitative Genotyping for the Astringency Locus in Hexaploid Persimmon Cultivars
using Quantitative Real-time PCR. _J. Am. Soc. Hortic. Sci._ **135**, 59–66 (2010).
23. D. G. MacArthur, T. A. Manolio, D. P. Dimmock, H. L. Rehm, J. Shendure, G. R. Abecasis,

D. R. Adams, R. B. Altman, S. E. Antonarakis, E. A. Ashley, J. C. Barrett, L. G. Biesecker,
D. F. Conrad, G. M. Cooper, N. J. Cox, M. J. Daly, M. B. Gerstein, D. B. Goldstein, J. N.
Hirschhorn, S. M. Leal, L. A. Pennacchio, J. A. Stamatoyannopoulos, S. R. Sunyaev, D.
Valle, B. F. Voight, W. Winckler, C. Gunter, Guidelines for investigating causality of
sequence variants in human disease. _Nature_ **508**, 469–476 (2014).
24. K. Yonemori, S. Kanzaki, C. Honsho, T. Akagi, D. E. Parfitt, Phylogeny and cultivar
development of _Diospyros kaki_ : A survey based on molecular analyses. _Adv. Hortic. Sci._
**22**, 261–268 (2008).
25. H. H. Hume, A kaki classification _J. Hered._ **5**, 400-406 (1914).
26. Y. Fujiwara, S. Nishiyama, A. Kusumi, K. Okamoto-Furuta, H. Yamane, K. Yonemori, R.
Tao, Developmental dynamics of cellular specialization during proanthocyanidin
accumulation in persimmon fruit. _Plant Physiol._ **200**, kiaf645 (2026).
27. R. Workman, R. Fedak, D. Kilburn, S. Hao, K. Liu, W. Timp, High molecular weight
DNA extraction from recalcitrant plant species for third generation sequencing. _protocol_
_exchange_, (2018). doi: 10.1038/protex.2018.059
28. Doyle, J. J., & Doyle, J. L. A rapid DNA isolation procedure for small quantities of fresh
leaf tissue. _Phytochemical Bulletin_ **19**, 11–15 (1987).
29. E. S. Rice, S. Koren, A. Rhie, M. P. Heaton, T. S. Kalbfleisch, T. Hardy, P. H. Hackett, D.
M. Bickhart, B. D. Rosen, B. V. Ley, N. W. Maurer, R. E. Green, A. M. Phillippy, J. L.
Petersen, T. P. L. Smith, Continuous chromosome-scale haplotypes assembled from a
single interspecies F1 hybrid of yak and cattle. _Gigascience_ **9** (2020).
30. H. Cheng, G. T. Concepcion, X. Feng, H. Zhang, H. Li, Haplotype-resolved de novo
assembly using phased assembly graphs with hifiasm. _Nat. Methods_ **18** (2021).


12


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


31. S. Kurtz, A. Phillippy, A. L. Delcher, M. Smoot, M. Shumway, C. Antonescu, S. L.
Salzberg, Versatile and open software for comparing large genomes. _Genome Biol._ **5**, R12
(2004).
32. H. Li, R. Durbin, Fast and accurate short read alignment with Burrows–Wheeler transform.
_Bioinformatics_ (2009).
33. H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis,
R. Durbin, The Sequence Alignment/Map format and SAMtools. _Bioinformatics_ **25**,
2078–2079 (2009).
34. C. Y. Wan, T. A. Wilkins, A modified hot borate method significantly enhances the yield
of high-quality RNA from cotton ( _Gossypium hirsutum_ L.). _Anal. Biochem._ **223**, 7–12
(1994).
35. H. Li, Minimap2: Pairwise alignment for nucleotide sequences. _Bioinformatics_ **34**, 3094–
3100 (2018).
36. J. T. Robinson, H. Thorvaldsdóttir, W. Winckler, M. Guttman, E. S. Lander, G. Getz, J. P.

Mesirov, Integrative genomics viewer. _Nat. Biotechnol._ **29**, 24–26 (2011).
37. A. Horiuchi, K. Masuda, K. Shirasawa, N. Onoue, N. Fujita, K. Ushijima, T. Akagi,
Ongoing Rapid Evolution of a Post-Y Region Revealed by Chromosome-Scale Genome
Assembly of a Hexaploid Monoecious Persimmon ( _Diospyros kaki_ ). _Mol. Biol. Evol._ **40**
(2023).
38. K. Katoh, K. Misawa, K.-I. Kuma, T. Miyata, MAFFT: a novel method for rapid multiple
sequence alignment based on fast Fourier transform. _Nucleic Acids Res._ **30**, 3059–3066
(2002).
39. T. K. F. Wong, N. Ly-Trong, H. Ren, H. Baños, A. J. Roger, E. Susko, C. Bielow, N. De
Maio, N. Goldman, M. W. Hahn, Others, IQ-TREE 3: Phylogenomic Inference Software
using Complex Evolutionary Models. _EcoEvoRxiv_ (2025). doi:10.32942/X2P62N
40. J. Rozas, A. Ferrer-Mata, J. C. Sánchez-DelBarrio, S. Guirao-Rico, P. Librado, S. E.

Ramos-Onsins, A. Sánchez-Gracia, DnaSP 6: DNA Sequence Polymorphism analysis of
large data sets. _Mol. Biol. Evol._ **34**, 3299–3302 (2017).
41. D. C. Koboldt, K. Chen, T. Wylie, D. E. Larson, M. D. McLellan, E. R. Mardis, G. M.
Weinstock, R. K. Wilson, L. Ding, VarScan: variant detection in massively parallel
sequencing of individual and pooled samples. _Bioinformatics_ **25**, 2283–2285 (2009).
42. D. C. Koboldt, Q. Zhang, D. E. Larson, D. Shen, M. D. McLellan, L. Lin, Wilson R,
VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome
sequencing. _Genome research_ **22**, 568–576 (2012).
43. P. Danecek, A. Auton, G. Abecasis, C. A. Albers, E. Banks, M. A. DePristo, R. E.

Handsaker, G. Lunter, G. T. Marth, S. T. Sherry, G. McVean, R. Durbin, 1000 Genomes
Project Analysis Group, The variant call format and VCFtools. _Bioinformatics_ **27**, 2156–
2158 (2011).

**Funding:** This work was supported by JSPS KAKENHI grants 21K14848 (S.N.), 21KK0269
(S.N.), 23K26895 (S.N. and N.O.), and 23KJ1313 (Y.F.).


13


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Competing interests:** Authors declare that they have no competing interests.


14


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Figure 1. Identification of a single candidate gene at the** _**AST**_ **locus by haplotype-resolved**
**genome assembly and polymorphism analysis**
(A) Synteny analysis of the delimited _AST_ region between the corresponding region in the
diploid _D. oleifera_ genome [9] and haplotype-resolved sequences of hexaploid _D. kaki_ ‘Taiten’.
Among six distinct haplotypes in ‘Taiten’, five haplotypes ( _A_ 224, _A_ 240, _a_ 349, _a_ 353_2a,
_a_ 353_2b) were successfully assembled, whereas _a_ 355 was not recovered. No detectable
polymorphisms were found between _a_ 353_2b and _a_ 355 (fig. S3). Polymorphisms used for the
Indel-3 marker (5R) [13] and the _A_ 224- and _A_ 240-specific markers [15] were detected in the
corresponding _A_ 224 and _A_ 240 contigs. (B) Distribution of _AST_ -linked polymorphisms detected
in the delimited _AST_ region of the _A_ 240 haplotype sequence. Arrows indicate coding regions
and their transcriptional orientation. Numbers in parenthesis indicate the number of _AST_ -linked


15


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


polymorphisms within each gene region (from 5’UTR to 3’ UTR). Gene annotations are
indicated by different colors shown in the bottom. The concentrated accumulation of _AST_  linked polymorphisms in a single gene annotated as CW7 (AT1G59520) indicates this gene as
the strongest candidate at the _AST_ locus.


16


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Figure 2. Abnormal** _**DkTAN**_ **transcription is strongly associated with PCNA and non-**
**PCNA genotypes**
(A) Distribution of read frequencies obtained from cDNA amplicon sequencing of ‘Taiten’
and ‘Kanshu’, classified by predicted peptide length. Functional full-length translations (389–
392 aa) were detected in ‘Taiten’, whereas they were detected only at trace level in ‘Kanshu’.
Black, green and red arrows indicate dominant peptide length translated from _AST_, _a_ 349 and
_a_ 353 ( _a_ 355), respectively. (B) Ratio of _DkTAN_ transcripts encoding full-length functional
proteins, estimated from amplicon-seq data using 106 non-PCNA and 25 PCNA cultivars.
cDNA sequences predicted to encode proteins longer than 380 aa were classified as functional
transcripts. A clear difference in the abundance of functional transcripts was observed between
comprehensive non-PCNA and PCNA cultivars. (C) Ratio of functional _TAN_ transcripts in


17


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


_Diospyros_ relatives. In all graphs, cDNA reads predicted to encode proteins longer than 380 aa
were classified as functional transcripts. (D) Ratio of functional _DkTAN_ transcripts in the 652population (‘Fuyu’ x ‘Yoshidagosho’). Ten sibling individuals, including both PCNA and nonPCNA progeny, together with their parental cultivars, were analyzed. YG, ‘Yoshidagosho’;
FY, ‘Fuyu’. (E) Relationship between the ratio of functional _DkTAN_ transcripts and dominant
_AST_ dosage. _AST_ dosage information of cultivars was obtained from a previous report [22] . The
non-PCNA siblings from the 652-line population analyzed in this study were genotyped as
_A_ 353 _aaaaa_ based on the genotyping of the Indel-3 region [20] .


18


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Figure 3. Complex abnormal** _**DkTAN**_ **transcription caused by two independent mutations**
**defining distinct** _**ast**_ **haplotype groups**
Schematic representation of abnormal transcripts caused by the two genomic mutations: the
_a_ 349-specific SNP near exon 3 and the _a_ 353/ _a_ 355-specific deletion spanning exon 7 and part
of exon 8. The SNP at the 5′ splice site of exon 3 causes abnormal splicing variants, including
shifts of the splice junction and exon 3 skipping. The 1.5-kb deletion including exon 7 and the
first portion of exon 8 results in skipping of the corresponding exon regions. These abnormal
transcripts generate premature stop codons and multiple truncated protein products. In addition,
a transcript variant encoding a 98-aa peptide, caused by shortening of the 3′ region of exon 3,
was detected independently of these two mutations. Deletion length and abnormal splicing
model were inferred from haplotype sequence of ‘Taiten’ and Iso-seq mapping.


19


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


**Figure 4. Sequence diversity of the** _**DkTAN**_ **region revealed by long-read target-**
**enrichment analysis in PCNA and non-PCNA cultivars**
(A) Phylogenetic analysis of assembled haplotype-resolved sequences spanning the _DkTAN_
genomic region. The tree topology suggests that _ast_ haplotypes can be classified into two
independent clades, _a_ 349 and _a_ 353/ _a_ 355. In contrast, presumed functional _TAN_ haplotype
sequences from non-PCNA cultivars formed a distinct _AST_ clade. Importantly, all sequences
obtained from PCNA cultivars were classified into either the _a_ 349 or _a_ 353/ _a_ 355 clade. (B)
Phased genotyping of the two causal mutations using the assembled haplotype-resolved
sequences. Sequence classification was based on the phylogenetic analysis shown in (A). All
sequences belonging to the _a_ 349 clade possessed only the SNP near exon 3, whereas all
sequences belonging to the _a_ 353/ _a_ 355 clade possessed only the 1.5-kb deletion encompassing
exon 7. _DkTAN_ sequences from non-PCNA deviated from the _AST_ clade ( `●` ) and those from

other _Diospyros_ species did not have either mutation. (C) Haplotype network analysis of
assembled _DkTAN_ sequences. Circle diameter represents the number of haplotypes included in
each group, and line length indicates the sequence dissimilarity between connected haplotype
groups. Yellow, orange, and green circles indicate _AST_, _ast_, and other _Diospyros_ haplotype


20


[bioRxiv preprint doi: https://doi.org/10.64898/2026.10.31.762479; this version posted November 3, 2026. The copyright holder for this](https://doi.org/10.64898/2026.10.31.762479)

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


groups, respectively. Red dots represent median vectors, which indicate intermediate
haplotypes inferred in the network analysis. (D) P opulation genetic analysis in the _AST_ locus.
Nucleotide diversity (π) and _F_ ST were calculated in 10 kb window using VCFtools based on
DNA-seq mapping data from the cultivars used for polymorphism analysis. Variant calling was
performed using VarScan, and the genotypes of the detected polymorphisms were refined using
custom Python scripts based on alternative-allele read frequency thresholds optimized for the
hexaploid genome (see Materials and Methods). Although a clear _F_ ST peak was detected around
the _DkTAN_ (Chr5: 315,216–348,804), π showed only a weak local increase, with no evidence
of reduced nucleotide diversity in the PCNA group across the _DkTAN_ region. The genomic
region containing _DkTAN_ is highlighted in blue. πP and πNP represent nucleotide diversity
within the PCNA and non-PCNA populations, respectively.


21