# Source paper archive

- Source title: High-throughput microarray analysis provides new insight into optimising Rubisco yield from diverse plants
- Author(s): Pattinson, J. C. D., Fletcher, C. S., Lelenaite, I., Lietz, G., Willats, W. G. T., Kapralov, M. V.
- Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766128v1
- PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766128v1.full.pdf
- PDF SHA-256: 80d5f02e3ca451caa34af18d5ae084004f2ea30af4dc3dc2ca3d1939ff86e0ca
- 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/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)

# 1  High-throughput microarray analysis provides new insight into optimising 2 Rubisco yield from diverse plants

3

4
5 Jack C. D. Pattinson [a*†], Charlotte Stenkjær Fletcher [a*†], Ieva Lel ė nait ė [a], Georg Lietz [b], William G. T.
6 Willats [a], Maxim V. Kapralov [a*],
7
8 a School of Natural and Environmental Sciences, Newcastle University, Newcastle upon Tyne, NE1
9 7RU, UK.
10 b Human Nutrition and Exercise Research Centre, Newcastle University, Newcastle upon Tyne, NE1
11 2HH, UK
12
13  - These authors contributed equally to this work
14  - Joint corresponding authors
15

16

17 **Abstract**


18 Interest in plant-based proteins is increasing due to environmental, ethical, and health considerations.


19 Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), the most abundant protein on Earth, is


20 present in all green leafy biomass and possesses excellent nutritional and functional properties.


21 Despite its potential, Rubisco remains underutilised because extraction methods are often costly,


22 inefficient, and influenced by species-specific differences. Moreover, the impact on Rubisco yield of


23 cell wall-protein interactions remains poorly understood. We applied Microarray Polymer Profiling


24 (MAPP), a high-throughput platform for simultaneous polysaccharide and protein mapping, to


25 characterise diverse leafy biomass sources. This approach revealed interspecies variation and glycan

26 protein interactions that influence protein extractability and these insights inform biomass selection


27 and optimisation of extraction protocols to improve Rubisco yield. MAPP was also used to evaluate


28 the effects of carbohydrate-active enzymes on Rubisco recovery through targeted cell wall


29 modification.


30 **Graphical abstract**


31 **Keywords:** Rubisco; plant protein, microarrays; molecular probes; glycan-protein interactions;


32 carbohydrate active enzymes


33

34

35

36

37


1


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


38 **1.** **Introduction**


39


40 Global agricultural production is facing unprecedented pressure to produce enough nutritious food


41 while simultaneously reducing its environmental impact. With the global population projected to


42 reach approximately 9.7 billion by 2050, rising incomes, urbanisation, and changing dietary


43 preferences are expected to drive greater demand for protein-rich foods (United Nations, Department


44 of Economic and Social Affairs, Population Division, 2022). However, conventional livestock


45 production contributes substantially to greenhouse gas emissions, freshwater use, land degradation,


46 and biodiversity loss (Poore & Nemecek, 2018; Tilman & Clark, 2014; Willett et al., 2019).


47 Consequently, the identification of sustainable protein sources capable of meeting future nutritional


48 requirements without placing additional pressure on natural resources has become a research priority


49 (Aiking & de Boer, 2020; van der Goot et al., 2016).


50


51 Plant-derived proteins have emerged as attractive alternatives to animal proteins due to their generally


52 lower environmental footprint and increasing consumer acceptance. Current markets are dominated


53 by soy, pea, and wheat proteins because of their high protein content and established processing


54 technologies (Day, 2013; Kyriakopoulou et al., 2019). However, dependence on a limited number of


55 crop species raises concerns regarding land use, monoculture production, and supply chain resilience.


56 Many conventional plant proteins also require extensive processing to remove antinutritional


57 compounds and undesirable sensory characteristics, increasing production costs and limiting


58 consumer appeal (McClements & Grossmann, 2021). These challenges have stimulated interest in


59 alternative protein sources that can be recovered from underutilised biomass streams while


60 maintaining favourable nutritional and functional properties.


61


62 Green leafy biomass represents one of the largest yet least exploited sources of plant protein. Leaves


63 are produced in large quantities as agricultural residues from vegetable, forage, and horticultural


64 crops, while dedicated biomass crops can generate substantial protein yields per hectare (Pérez-Vila et


65 al., 2022; Tamayo Tenorio et al., 2016). Despite this abundance, leafy biomass is commonly used as


2


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


66 animal feed, compost, or returned to soil, despite containing large quantities of soluble proteins.


67 Recovering proteins from these biomass streams offers an opportunity to increase the value of existing


68 agricultural systems and supports circular bioeconomy principles aimed at maximising resource


69 efficiency and reducing waste generation (van der Goot et al., 2016).


70


71 Among the proteins present in green leaves, ribulose-1,5-bisphosphate carboxylase/oxygenase


72 (Rubisco) has attracted considerable attention because of its abundance, nutritional quality, and


73 functional properties. Rubisco is the key enzyme responsible for carbon fixation during


74 photosynthesis and is widely recognised as the most abundant protein on earth (Andersson &


75 Backlund, 2008; Bar-On & Milo, 2019; Ellis, 1979; Spreitzer & Salvucci, 2002). In many C � species


76 it accounts for 30-50% of soluble leaf protein and up to 30% of total leaf nitrogen (Bar-On & Milo,


77 2019; Ellis, 1979; Spreitzer & Salvucci, 2002). This exceptional abundance, combined with the


78 renewability of leaf biomass, makes Rubisco an attractive candidate for sustainable protein


79 production.


80


81 Interest in Rubisco has increased as advances in green biorefinery technologies have demonstrated the


82 feasibility of producing food-grade proteins from fresh leaf biomass (Krishnan et al., 2026; Pérez-Vila


83 et al., 2022; Tamayo Tenorio et al., 2016). Unlike seed storage proteins, Rubisco is recovered from


84 vegetative tissues that are continuously produced throughout plant growth and are often underutilised.


85 Exploiting Rubisco therefore offers the potential to increase protein production without requiring


86 additional agricultural land or major changes to existing cropping systems. Moreover, Rubisco


87 possesses several characteristics that make it attractive as a food ingredient, and it is considered one of


88 the highest-quality plant proteins available. Unlike many seed storage proteins, which evolved


89 primarily as nutrient reserves, Rubisco functions as a metabolic enzyme and exhibits a balanced


90 amino acid composition that closely aligns with human nutritional requirements (Shewry, 2007). It


91 contains high concentrations of lysine, threonine, valine, leucine, and isoleucine, amino acids that are


92 often limiting in cereal proteins (Rhodes et al., 1980; Tamayo Tenorio et al., 2016). Furthermore,


93 purified Rubisco preparations are pale in colour, possess a relatively neutral flavour and contain


3


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


94 relatively low levels of antinutritional compounds such as trypsin inhibitors and lectins (Food and


95 Agriculture Organization of the United Nations, 2013; Martin et al., 2019; McClements &


96 Grossmann, 2021). Previous studies have demonstrated excellent solubility across a wide pH range,


97 together with strong emulsifying, foaming, and gel-forming capacities (Krishnan et al., 2026; Ma et


98 al., 2022; Pérez-Vila et al., 2022; Tamayo Tenorio et al., 2016). These characteristics make Rubisco a


99 promising ingredient for a variety of food applications, including beverages, dairy alternatives,


100 aerated products, and meat analogues. Compared with conventional plant proteins, Rubisco offers a


101 unique combination of nutritional quality and functional performance, suggesting it could


102 complement existing protein ingredients in future food systems. Despite these advantages,


103 commercial utilisation of Rubisco remains limited by challenges associated with efficient extraction


104 and purification.


105


106 Rubisco is located within the chloroplast stroma and is enclosed by multiple structural barriers,


107 including chloroplast membranes, the plasma membrane, and the surrounding plant cell wall (Pérez

108 Vila et al., 2022; Tamayo Tenorio et al., 2016). Conventional extraction methods generally involve


109 mechanical disruption of fresh biomass followed by pressing and separation of protein-rich green


110 juice from fibrous residues. However, these approaches often recover only a fraction of the soluble


111 protein present within leaf tissues, with substantial quantities remaining associated with the fibre


112 fraction (Pérez-Vila et al., 2022; Tamayo Tenorio et al., 2016). In addition, extraction efficiency varies


113 considerably among plant species, suggesting that intrinsic biological characteristics strongly


114 influence protein accessibility and release.


115


116 The plant cell wall is increasingly recognised as a major barrier limiting recovery of intracellular


117 proteins (Kleuter et al., 2024; Sharma et al., 2025; Wang et al., 2025). Cell walls are complex


118 composite structures composed primarily of cellulose, hemicelluloses, pectin, structural proteins, and


119 phenolic polymers (Carpita & Gibeaut, 1993; Cosgrove, 2005; Cosgrove, 2018; Harholt et al., 2010;


120 Ridley et al., 2001; Scheller & Ulvskov, 2010; Showalter, 1993; Somerville et al., 2004). Their


121 composition and organisation vary according to species, tissue type, and developmental stage,


4


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


122 influencing wall rigidity, porosity, and susceptibility to disruption (Carpita, 2011; Fangel et al., 2012).


123 Such variation is likely to affect the release of Rubisco during processing and hence the effectiveness


124 of extraction technologies. However, while considerable effort has been directed towards optimising


125 extraction parameters such as pH, temperature, and mechanical pressing, relatively little attention has


126 been given to understanding how differences in cell wall composition influence protein recovery


127 (Pérez-Vila et al., 2022; Tamayo Tenorio et al., 2016).


128


129 One promising approach to overcoming these structural barriers is enzyme-assisted extraction.


130 Cellulases, hemicellulases, and pectinases selectively degrade major components of the cell wall,


131 increase tissue permeability and enhance access to intracellular proteins (Harholt et al., 2010; Scheller


132 & Ulvskov, 2010). Previous studies have shown that enzyme-assisted processing can improve protein


133 recovery and reduce the energy required for biomass disruption (Krishnan et al., 2026; Ma et al.,


134 2022; Pérez-Vila et al., 2022; Tamayo Tenorio et al., 2016). However, responses vary substantially


135 among species, suggesting that differences in cell wall architecture influence the effectiveness of


136 enzyme treatments and the specific relationships between cell wall polysaccharide composition and


137 Rubisco extractability remain poorly understood.


138


139 In this study, we used an analytical platform (Microarray Polymer Profiling or ‘MAPP’) that


140 combines the specificity of molecular probes (monoclonal antibodies, mAbs, and carbohydrate


141 binding modules, CBMs) with the high-throughput capacity of microarrays (Knox, 2008; Møller et


142 al., 2007; Pattathil et al., 2010; Pattathil et al., 2015; Pedersen et al., 2012; Verhertbruggen et al.,


143 2009). We assessed Rubisco abundance in 128 phylogenetically diverse plant species and investigated


144 plant cell wall polysaccharides that may influence biomass processing and Rubisco extraction. By


145 linking variations in cell wall architecture with protein recovery, we aimed to identify the key cell


146 wall structural determinants limiting Rubisco extraction and to establish new principles for feedstock


147 selection and process optimisation. These insights address a critical barrier to the commercial


148 utilisation of leaf proteins and provide a framework for developing more efficient, species-specific


149 extraction technologies.


5


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


150 **2.** **Materials and methods**


151 All chemicals and reagents were of analytical grade, unless otherwise stated.


152 **2.1 Collection of plant material**


153 Leaves from a representative sample set of 128 plants were used to understand the impact of cell wall


154 polysaccharides on soluble leaf protein Rubisco extractability. The plants chosen were a


155 representation of common crops, house and commonly encountered garden plants, where plant


156 material was obtained from living plants. The species in the collection are available at Newcastle


157 University. Details of the plant material used are included in **Table S1.**


158


159 **2.2 Sample preparation**


160 For each species, leaves were collected in abundance from multiple plants where possible. The leaves


161 were combined into a Falcon tube for each species to normalise sampling per species and immediately


162 flash-frozen in liquid nitrogen. The samples were then stored in the -80 °C before being freeze-dried


163 (LaboGene). The freeze-dried leaves were homogenised by shaking using one 12 mm steel bead per


164 sample. Once homogenised, 3 samples of approximately 30 mg, weighed with 1 decimal, was taken


165 from each Falcon tube to yield 3 replicate samples per species. This procedure was repeated to yield a


166 total of 6 replicate samples per species to account for the two-part extraction series.


167


168 **2.3 Protein and polysaccharide extractions**


169 A modified two-part extraction series was used to extract constituent glycans and proteins. The


170 extractant volumes were adjusted to the exact weight of each replicate sample (30 µL mg [-1] ). The


171 extraction series is based on the procedure used by Bakshani et al. (2023) and Møller et al. (2007)


172 with the addition of a cellulase extraction and a separate protein extraction.


173


6


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


174 **2.4 Protein extraction**


175 Plant leaf material has a large amount of protein present, hence the first extraction targeted these. 3


176 replicate samples per species were reconstituted in a 50 mM HEPES, pH 7.7, 150 mM NaCl, 1 mM


177 EDTA, 2 mM MgCl 2, 10 % (v/v) Glycerol, 0.1 % Triton X-100, 1 % (w/v) PVPP and 1 µL mL [-1] of


178 protease inhibitor cocktail (added just before use). A steel bead was added to each sample tube and


179 then shaken at 27 s [-1] for 2 minutes in a Tissuelyser II (Qiagen, UK). The speed was then reduced to 10


180 s [-1] for 2 hours. After extraction, the samples were centrifuged at 13000 RPM for 10 minutes (min).


181 The supernatant was carefully removed from each of the samples and transferred into an Eppendorf.


182 Extracts were stored at 4 °C.


183


184 **2.5 Polysaccharide extraction**


185 A new set of 3 replicate samples per species were subjected to a 3-step sequential extraction. Firstly,


186 the solvent 50 mM CDTA (pH 7.5) was used, a steel bead was added, and the extraction process was


187 carried out as described above. The second extraction step used a 4 M NaOH + 0.1 % (w/v) NaBH 4


188 buffer and the extraction step was similarly carried out as described above. The third and final


189 extraction of the precipitate pellet from second extraction step used 20 mM Tris, pH 8.8, with the


190 addition of 2 µL mL [-1] of _Bacillus sp._ Cellulase 5A (NZYTech, Portugal). This extraction buffer


191 requires a 16-hour incubation at 45 °C, on a dry heat block followed by centrifugation at 13000 RPM


192 for 10 min, and the supernatant being carefully removed. All extracts were stored at 4 °C.


193


194 **2.6 Microarray printing**


195 A serial dilution of each sample was made in a 384-well microtiter plate. Each extract was diluted 4

196 fold in ArrayJet buffer (47 % glycerol, 52.9 % deionised water, 0.06 % Triton X-100 and 0.04 %


197 ProClin™ 200) to a final volume of 40 μL in each well. Plates with diluted extracts were centrifuged


198 at 3700 RPM for 10 min before they were printed on a nitrocellulose membrane using an Arrayjet


199 Marathon Argus Microarray Printer (Arrayjet, Roslin, UK) as described previously (Bakshani et al.,


7


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


200 2023; Fangel et al., 2024). For each sample, the dilution series was printed in duplicate on each


201 microarray.


202


203 **2.7 Microarray probing**


204 The printed arrays were probed and developed following procedures described by Bakshani et al.


205 (2023) and Fangel et al. (2024). A total of 53 arrays were used for this analysis, one for each antibody


206 and 5 for negative controls of the secondary antibodies. They were cut and placed into 4-well plate


207 and a 5 % (w/v) semi-skimmed milk powder solution (MPBS) was made in phosphate-buffered saline


208 (PBS; 140 mM NaCL, 2.7 mM KCl, 10 mM Na 2 HPO 4, 1.7 mM KH 2 PO 4, pH 7.5). The arrays were


209 left agitating in MPBS to block any non-specific binding to the nitrocellulose for 1 hour. The arrays


210 were then placed face down onto an MPBS solution containing one of the antibodies; one array is


211 used per primary antibody. The control arrays were left in the pure MPBS solution because they are


212 needed for negative controls of the secondary antibodies. The primary antibody solutions were made


213 by adding antibodies to the MPBS solution and aspirating thoroughly. The optimal primary antibody


214 concentrations were given by the manufacturers ( **Table S2** ). The microarrays were left in the antibody


215 and MPBS solution for 2 hours and then washed in PBS 4 times and left agitating for 5 minutes after


216 every 2 washes. The microarrays were subsequently probed with the secondary antibody, either anti

217 rat, anti-mouse, anti-rabbit or anti-polyhistidine, depending on the origin of the primary antibody used


218 on the specific array ( **Table S2** ). The secondary antibodies were conjugated to alkaline-phosphatase


219 (Jackson ImmunoResearch and Sigma, Merck Life Science) and diluted 1:1000 in MPBS. The arrays


220 were left agitating in the secondary antibody and MPBS solution for 2 hours before being washed in


221 PBS as described above. Finally, the arrays were submerged in a developing solution containing 5

222 bromo,4-chloro,3-indolylphosphate (BCIP) and 4-Nitro blue tetrazolium chloride (NBT). BCIP/NBT


223 was added to an alkaline phosphatase buffer (100 mM Tris, 5 mM MgCl 2  - 6H 2 O, pH 9.5). The


224 developing arrays were rinsed in deionised water before they were mounted on a thin sheet of


225 commercially available graph paper and prepared for scanning.


226


8


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


227 **2.8 Microarray quantification**


228 The developed arrays were scanned with a Canon 9000F Mark II desktop scanner (Canon, Newcastle,


229 England) using IJ Scan Utility at a resolution of 2400 dpi. The arrays were scanned individually using


230 grayscale colour settings, and the images were saved as JPG files. The antibody binding signals are


231 visualised on the arrays as lighter or darker spots, thus meaning darker spots present higher


232 concentrations of a specific epitope. If a specific epitope is not present in a sample, there will be no


233 binding of the antibody, hence no spot is shown on the grayscale array image.


234


235 The array images were uploaded and quantified one at a time using ArrayPro Analyzer (v.6.3.1, Media


236 Cybernetics, Rockville, USA). Spots were quantified by overlaying a unique .GAL file, and spots are


237 defined as the mean pixel value within each spot minus the median pixel value of the surrounding


238 local background. The data were then exported and saved as a .txt file and analysed using Microsoft


239 Office Excel (Microsoft, Redmond, WA). Averages were calculated for each set of replicates across


240 each of the dilutions and the highest value of the data set was assigned the value of 100. The rest of


241 the data set was adjusted accordingly as described by Bakshani et al. (2023) and Fangel et al. (2024).


242


243 **2.9 Total soluble protein (TSP) quantification**


244 Total soluble protein (TSP) concentrations (mg mL � ¹) were determined using the Pierce Bradford


245 Protein Assay Kit (Thermo Fisher Scientific; catalogue no. 23200), following the manufacturer’s


246 microplate protocol. The assay is based on the Bradford method (Bradford, 1976), and absorbance


247 was measured using a Varioskan LUX multimode microplate reader (Thermo Fisher Scientific,


248 Waltham, MA, USA).


249


250 **2.10 Enzyme Treatment**


251 Enzyme treatments were conducted using an adaptation of Vidal-Melgosa et al. (2015). Leaf material


252 from cucumber ( _Cucumis sativus_ ), tomato ( _Solanum lycopersicum_ ) and perennial ryegrass ( _Lolium_


9


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


253 _perenne_ ) was obtained from the same biomass batches used for the screening panel. Pulp was


254 prepared by thawing 200 g of frozen leaves at 4 °C and processing them using a twin-screw press


255 (Angel Juicer 5500, Angel Co., Ltd., Busan, Republic of Korea); the resulting solid fraction was


256 retained. Freeze-dried and frozen-thawed leaf and pulp samples were prepared in triplicate. Freeze

257 dried material was weighed directly to approximately 30 mg dry weight (DW), whereas frozen

258 thawed material was adjusted using its measured dry matter content to provide an equivalent dry mass


259 ( **Table S3** ). Seven enzyme treatments were evaluated: Celluzyme 0.7T, Pectinex Ultra SP-L,


260 Macerozyme R-10, Driselase, Xylio Pre, Pectinex Ultra SP-L combined with Macerozyme R-10, and


261 Pectinex Ultra SP-L combined with Celluzyme 0.7T. Samples were suspended in 50 mM sodium


262 acetate buffer (pH 5.5) containing the relevant enzyme preparation at 0.1–0.5 % (w/v), or buffer alone


263 as the no enzyme control. Treatment specific enzyme concentrations are provided in **Table S3** .


264 Samples were incubated at 30 °C for 1 hour with gentle agitation. A steel bead was then added, and


265 the samples were disrupted at 10 s � ¹ for 1 hour using a TissueLyser II (Qiagen, UK), followed by


266 centrifugation at 13000 RPM for 10 min. Supernatants were stored at 4 °C until MAPP ( **section 2.6 -**


267 **2.8** ) and TSP ( **section 2.9** ) analyses, was carried out. Please note, only RbcL (2) was used for probing.


268


269 **2.11 Data preparation and visualisation**


270 Unless otherwise stated, measurements were performed using three replicates and are presented as


271 mean ± SD. Replicates were averaged for the biomass-screening analyses, whereas enzyme-treatment


272 analyses used replicate-level data. Enzyme treatment TSP yield was expressed as mg g � ¹ DW and


273 calculated by multiplying TSP concentration by extraction volume and dividing by sample dry mass.


274 The weighed mass was used for freeze-dried material, while frozen-thawed dry mass was calculated


275 from the wet mass and measured dry matter fraction ( **Table S3** ). Protein and glycan antibody


276 heatmaps were generated in GraphPad Prism (v11.0.2; GraphPad Software, San Diego, CA, USA); all


277 other figures were generated in R using RStudio.


278


279


10


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


280 **2.12 Statistical analysis**


281 Statistical analyses were performed using R v4.5.2 (R Core Team, 2025) within RStudio


282 v2026.07.1+147. Associations between TSP and Rubisco antibody signals were assessed using two

283 sided Pearson correlations and ordinary least-squares regression with 95% confidence intervals.


284 Protein-glycan associations were evaluated using pairwise-complete Pearson correlations between


285 signals from 13 protein and 15 glycan antibodies across 128 biomass samples (195 comparisons).


286 Associated p values were adjusted across the 195 comparisons using the Benjamini-Hochberg false


287 discovery rate. Correlation profiles were hierarchically clustered using Euclidean distance and


288 complete linkage. TSP was included separately in a sensitivity analysis ( **Fig. S1** ).


289


290 Principal component analyses used complete, mean-centred and unit-variance-scaled data. PC1 scores


291 and loadings were jointly reversed where necessary so that the summed loadings of the three Rubisco


292 antibodies were positive. Ellipses drawn at the 68% level summarised within-group dispersion. Group


293 effects were assessed on the complete standardised matrices using permutational multivariate analysis


294 of variance (PERMANOVA) with Euclidean distances and 9,999 permutations. Differences in


295 dispersion were assessed using permutational analysis of multivariate dispersion (PERMDISP). The


296 vine/climber category (n = 2) was excluded from the primary growth-form tests because the group


297 size was insufficient for stable inference; including all groups in a sensitivity analysis did not alter the


298 overall interpretation. Associations between the six targeted glycan antibodies and the multivariate


299 response across the three Rubisco antibodies were assessed by redundancy analysis (RDA) using


300 permutation testing and adjusted R². Conditional associations with Rubisco-only PC1 and TSP were


301 examined using standardised multiple linear regression with HC3 robust standard errors and 95 %


302 confidence intervals. Pearson and Spearman correlations were used as sensitivity analyses, with


303 Benjamini-Hochberg adjustment applied within each analysis family.


304


305 Enzyme-treatment effects on TSP yield and RbcL (2) signal in frozen-thawed material were analysed


306 separately within each species × matrix combination by one-way analysis of variance (ANOVA),


11


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


307 followed by Dunnett-adjusted comparisons of each treatment with the matched no enzyme control.


308 Model assumptions were assessed using residual diagnostic plots, Shapiro-Wilk tests for residual


309 normality and Levene tests for homogeneity of variance. Unless otherwise stated, statistical


310 significance was accepted at _p_ < 0.05.


311 **3.** **Results and discussion**


312 **3.1 Development of a high-throughput analytical platform**


313 Using MAPP technology as a basis, we developed a specialised platform for the high-throughput


314 parallel analysis of proteins involved in photosynthesis (including Rubisco) and polysaccharides from


315 128 samples of plant biomass. The workflow used is shown **Fig. 1** and involved extraction of


316 polysaccharides and proteins which were printed as microarrays and subsequentially interrogated


317 using panels of antibodies ( **Table S2** ). Microarray data is presented as heatmaps (full heatmap can be


318 seen in **Data S1** ). TSP was also determined using Bradford assays from the same sample set. It is


319 important to recognise that MAPP data does not provide information about absolute amounts of


320 glycans or proteins but rather reports differences in relative abundance across samples.


321


322 **Figure 1.**


323


324 **3.2 MAPP analysis of Rubisco and other photosynthesis related proteins**


325 MAPP was used to assess the relative abundance of Rubisco and associated photosynthetic proteins


326 across the plant sample set. Microarrays of extracted proteins were probed with monoclonal


327 antibodies specific for Rubisco subunits and a variety of other protein with data presented as a


328 heatmap in **Fig. 2** .


329


330 **Figure 2.**


331


12


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


332 The relative abundance of extracted proteins varied widely across the biomass sample set, as reported


333 by different binding signals. Regarding the three Rubisco specific antibodies (anti-RBcL (1), anti

334 RbcL (2) and RbcS), _Brassica rapa_ ‘Joi Choi’ produced the highest RbcL (2) and RbcS signals, while


335 _Lolium perenne_ produced the highest RbcL (1) signal. _Tarenaya spinosa_, _Beta vulgaris_ subsp. _cicla_


336 and _Spinacia oleracea_ ‘Semi-Savoy’ also showed consistently high Rubisco related profiles.


337 Conversely, _Fuchsia magellanica_ and _Araucaria heterophylla_ showed lower signals for all three


338 probes. RbcL (1/2) and RbcS were the most strongly associated antibodies in terms of binding


339 profiles, consistent with their stoichiometric assembly in the Rubisco holoenzyme and coordinated


340 expression across the chloroplast and nucleus (Suzuki & Makino, 2012; Vitlin Gruber & Feiz, 2018).


341 However, there were also notable variations in the binding of individual Rubisco antibodies within


342 some biomass samples. For example, _Anemanthele lessoniana_ produced a high RbcL (2) signal but a


343 substantially lower RbcS signal, whereas _Pelvetia canaliculata_ had low RbcL (1) and higher RbcL


344 (2)/RbcS signals. This supports the notion that RbcL and RbcS protein concentration is not necessarily


345 present as a 1:1 ratio, and as such must be measured separately to understand the abundance of each


346 (Recuenco-Muñoz et al., 2015; Wostrikoff & Stern, 2007). These findings also highlight the


347 importance of using a panel of three antibodies with specificity for different subunits to obtain a more


348 complete picture of overall Rubisco abundance. Signals from Calvin–Benson cycle/regulatory


349 proteins, photosystem II targets and associated carbon-metabolism enzymes were also heterogeneous


350 and did not completely parallel the Rubisco profiles, indicating that the array resolved complementary


351 components of the photosynthetic-protein phenotype.


352


353 Antibody binding profiles were not generally correlated to taxonomic groups, with Joi Choi,


354 Tarenaya, sugar beet, spinach and Lolium yielding high signals despite their phylogenetic divergence.


355 This is consistent with evidence that nitrogen allocation to Rubisco varies with leaf traits and


356 environment rather than following a single plant-functional-type rule (Luo et al., 2021).


357 Developmental state also affects both the soluble-protein pool and extractability, and a recent tomato


358 leaf proteomics study linked declining extractability during maturation to a shift from anabolic to


359 catabolic/stress functions and to incomplete recovery of soluble proteins (Kleuter et al., 2025). These


13


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


360 data also highlight the importance of empirical phenotyping of candidate feedstocks instead of


361 assuming that crop identity or taxonomy predicts Rubisco potential.


362


363 Plants are classified according to their photosynthetic carbon fixation pathway into C 3, C 4 and


364 crassulacean acid metabolism (CAM) plants ( **Data S2)** . C 3 plants, which include most temperate


365 crops and dicotyledonous species such as tomato ( _Solanum lycopersicum_ ) and cucumber ( _Cucumis_


366 _sativus_ ), fix atmospheric CO � directly through the Calvin–Benson cycle via Rubisco, producing the


367 three-carbon compound 3-phosphoglycerate as the first stable product (Taiz et al., 2018). In contrast,


368 C 4 plants, including grasses such as maize and many forage species, initially fix CO � into four

369 carbon compounds using phosphoenolpyruvate carboxylase (PEPC) before concentrating CO �


370 around Rubisco within specialised bundle sheath cells. This spatial separation reduces


371 photorespiration and improves photosynthetic efficiency under conditions of high temperature, intense


372 light and limited water availability (Hatch, 1987; Sage & Zhu, 2011). CAM plants, which include


373 many succulents and cacti, separate carbon fixation temporally by opening their stomata at night to


374 minimise water loss while storing CO � as organic acids that are subsequently decarboxylated during


375 daylight for photosynthesis. This adaptation provides exceptional water-use efficiency in arid


376 environments (Winter & Smith, 2021). The differing photosynthetic pathways are associated with


377 distinct leaf anatomies, physiological characteristics and carbon allocation patterns, all of which may


378 influence biomass composition and the efficiency of protein extraction from plant tissues. In this


379 study, as previously described, C 3 plants demonstrate higher Rubisco content whilst C 4 demonstrate


380 lower Rubisco content (Galmés et al., 2014; Sugawara et al., 2026).


381


382 MAPP highlighted key differences in plant protein abundance across the sample set, demonstrating


383 the effectiveness of this technique as a way to screen a large sample set in a high-throughput manner.


384 Next, we assessed the extent to which MAPP data agreed with those generated using the established


385 Bradford assays to measure TSP.


386


14


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


387 **3.3 TSP – Variation in extractable soluble protein across the biomass panel**


388 The Bradford assay has been used for decades to quantify TSP due to its simplicity, rapidity, cost

389 effectiveness, and relative sensitivity. In the context of measuring Rubisco across the sample set,


390 Bradford measurements are informative because Rubisco often constitutes a substantial proportion of


391 soluble leaf protein.


392


393 **Figure 3.**


394


395 Extractable soluble protein was quantified across the biomass panel using a standardised protein


396 extraction workflow ( **section 2.4** ). TSP concentrations varied markedly among the 128 samples ( **Fig.**


397 **3** ), ranging from 0 to 2.906 mg mL � ¹, with a median of 0.423 mg mL � ¹ and an interquartile range of


398 0.083–1.060 mg mL � ¹. _Spinacia oleracea_ ‘Semi-Savoy’ produced the highest concentration (2.906


399 ± 0.032 mg mL � ¹), followed by _Brassica rapa_ ‘Joi Choi’ (2.662 ± 0.130 mg mL � ¹), _Tarenaya_


400 _spinosa_ (2.310 ± 0.253 mg mL � ¹), _Lemna minor_ (2.211 ± 0.161 mg mL � ¹) and _Chenopodium_


401 _giganteum_ (2.147 ± 0.032 mg mL � ¹). High-TSP samples occurred across several taxonomic groups,


402 whereas near-baseline recovery occurred among several fern, conifer and woody ornamental samples.


403 Caryophyllales and rosids contained several highly ranked samples, however, unequal taxonomic


404 representation and substantial within-group variation preclude conclusions relating TSP to plant


405 lineage. This broad variation is consistent with reported crop- and tissue-dependent differences in


406 TSP. Song et al. (2015) identified comparatively high TSP in broccoli and Chinese cabbage, broadly


407 agreeing with the presence of six _Brassica_ accessions among the 20 highest-ranked samples in the


408 present screen. However, their use of different plant organs, extraction conditions and a fresh-biomass


409 reporting basis preclude direct value comparison. Among the horticultural crops central to the wider


410 study, spinach produced 2.906 mg mL � ¹ TSP, compared with 0.974 mg mL � ¹ for cucumber and


411 0.560 mg mL � ¹ for tomato. Spinach therefore showed approximately 3-fold more TSP than cucumber


412 and 5.2-fold more than tomato, consistent with previous studies highlighting spinach as a promising


413 source material for Rubisco extraction and leaf protein concentrate production (Valente et al., 2022).


15


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


414 The high ranking of duckweed ( _L. minor)_ likewise agrees with its recognised potential as a source of


415 Rubisco-rich protein concentrates (Nieuwland et al., 2021). Comparable food- or feed-directed


416 evidence for protein recovery from cucumber leaves remains limited. The intermediate cucumber TSP


417 concentration should therefore be interpreted as a screening-specific indication of extractable-protein


418 potential rather than evidence of established suitability as a protein feedstock.


419


420 A notable contrast was observed within _Solanum_ : _S. tuberosum_ ‘Carolus’ produced 1.973 mg mL � ¹


421 TSP, approximately 3.5-fold more than _S. lycopersicum_ . This difference shows that close taxonomic


422 relatedness did not translate into similar protein recovery under the common workflow. Nevertheless,


423 the lower tomato value does not necessarily indicate lower total leaf-protein content. During tomato

424 leaf processing in other studies, substantial losses have been attributed to retention of protein

425 containing liquid within fibrous pulp and the presence of insoluble protein, while younger upper


426 leaves produced higher protein yields than older leaves (Yu et al., 2024). Reduced tomato leaf


427 extractability during development has also been associated with decreased protein synthesis, increased


428 protein degradation and incomplete extraction of the soluble-protein fraction (Kleuter et al., 2025).


429 Thus, the relatively low TSP values observed for tomato and other biomass sources should not


430 necessarily be interpreted as indicative of low protein abundance but may instead result from


431 developmental changes and matrix-related limitations that reduce protein extractability. The constant


432 biomass-to-extractant ratio (30 µL mg � ¹ dry biomass) excluded biomass loading as an explanation for


433 the observed variation. Nevertheless, a common protocol is unlikely to recover protein equally from


434 structurally and chemically contrasting biomass matrices. Measured TSP may reflect the intrinsic


435 soluble protein pool, developmental stage, protein degradation or aggregation, tissue accessibility, and


436 interactions with phenolic compounds, pigments and cell wall polysaccharides. In plant-based protein,


437 increased alkaline protein recovery was associated principally with disruption of the tissue and cell


438 wall matrix rather than increased protein solubility or hydrolysis (Zhang et al., 2015). Different


439 physical treatments similarly produced different soluble protein and Rubisco recoveries from sugar

440 beet leaves (Duki ć et al., 2023), confirming that recovery is process dependent.


441


16


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


442 TSP should therefore be interpreted as an operational measure of soluble protein recovered under the


443 applied conditions, rather than a direct measure of total tissue protein, extraction yield, protein purity


444 or Rubisco abundance. Rubisco may account for up to approximately 50 % of soluble leaf protein in


445 some plants (Parry et al., 2012), this proportion varies substantially with species, tissue type and


446 developmental stage, as demonstrated by changes in the association between soluble protein and


447 Rubisco abundance during rice development (Ishimaru et al., 2001).


448


449 **Figure 4.**


450


451 Despite the reliance on Bradfords to measure TSP it fails to distinguish individual proteins, and as


452 such caution is required when interpreting TSP data. With MAPP, the individual protein is assessed,


453 however it is not quantifiable, so efforts were made to understand the relationship between MAPP and


454 TSP data. **Fig. 3** establishes contrasting extractable protein phenotypes, whereas **Fig. 4** tests whether


455 this independent biochemical measurement is associated with the Rubisco directed MAPP signals. All


456 three Rubisco directed MAPP signals were strongly and positively correlated with TSP ( _r_ = 0.69–


457 0.78, _p_ < 0.001) ( **Fig. 4** ), indicating that biomasses with greater Bradford reactive soluble protein


458 recovery generally produced stronger Rubisco directed MAPP signals. This agreement is biologically


459 consistent with Rubisco comprising a substantial fraction of the soluble leaf protein pool, while the


460 consistent relationships across two RbcL antibodies and RbcS suggest that the pattern was not specific


461 to a single antibody. However, because TSP includes non-Rubisco proteins and MAPP signals are


462 relative, the two measurements should be considered complementary rather than quantitatively


463 interchangeable.


464


465 **3.4 MAPP analysis of cell wall polysaccharides**


466 Cell walls are comprised of complex three-dimensional polysaccharide matrixes that collectively


467 constitute most of the plant body. In some cases, cell wall polymers are known to impact protein


468 accessibility and yield from biomass, and our goal was to investigate these effects for Rubisco. We


17


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


469 therefore used MAPP to assess the relative abundance of cell wall structures across the plant sample


470 set ( **Fig. 5** ) and integrated these data with protein (including Rubisco) yield data ( **Data S1** ).


471


472 **Figure 5.**


473


474 Summed signals for polysaccharide-specific probes across the CDTA, NaOH and cellulase extraction


475 fractions revealed extensive variation in cell wall epitopes. As expected, (1 → 3,1 → 4)- β -glucan signals


476 (detected by mAb BS-400-3) was highest in Poales with the nine highest signals originating from this


477 group, led by _Anemanthele lessoniana_ and _Miscanthus sinensis_ . These data agree with the well

478 established occurrence of (1 → 3,1 → 4)- β -glucan in the cell walls of grasses and cereals. However,


479 immunochemical and phylogenetic work has demonstrated mixed linkage glucan beyond Poaceae and


480 even beyond Poales, so the non-Poales signals should be treated as biologically plausible rather than


481 anomalous (Little et al., 2018; Sørensen et al., 2008).


482


483 Poales also showed high signals for the xylan directed mAb, LM11 and LM28, signals, with


484 _Fascicularia bicolor_ producing the maximum for both probes. The LM11/LM28 xylan signals are


485 consistent with xylan being the dominant non-cellulosic polysaccharide in many grass secondary


486 walls, and LM11 recognises both unsubstituted xylan and wheat arabinoxylan epitopes (McCartney et


487 al., 2005; Tryfona et al., 2023). The strong Fascicularia, Koeleria, Fargesia, Anemanthele and bamboo


488 signals support a recurring grass wall phenotype, while high signals in selected asterids, rosids and


489 other monocots highlight the cross-lineage occurrence of xylan.


490


491 Non-Poales monocots showed a distinct profile for mannan containing polysaccharides (as detected


492 by CBM27, and mAbs BS-400-4 and LM21). _Monstera deliciosa_ produced the highest signal for all


493 three probes, and _Musa basjoo_, _Agapanthus_ sp. and _Aloiampelos striatula_ also ranked highly. The


494 mannan directed profile in non-Poales monocots is corroborated by previous MAPP data showing


495 prominent soluble mannans in several non-commelinid monocot leaves and succulent tissues


496 (Fradera-Soler et al., 2022). The agreement of CBM27, BS-400-4 and LM21, together with the


18


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


497 repeated prominence of Monstera, Musa, Agapanthus and Aloiampelos, makes this more persuasive


498 than a single probe maximum. Strong mannan signals in selected Brassica samples remain an


499 important exception.


500


501 The glycan screen provides insight into the three-dimensional cell wall architectures from which


502 soluble, and organellar proteins must be released, but it should not be presented as a direct assay of


503 recalcitrance. Cell walls are dynamic composite materials whose porosity, hydration, charge and


504 mechanics emerge from interactions among cellulose, hemicelluloses, pectin, proteins and phenolics


505 (Anderson & Kieber, 2020; Thompson & Islam, 2021). A highly accessible epitope signal can


506 therefore indicate composition, extractability or exposure; it does not by itself show that the epitope


507 physically traps Rubisco. In green-tea leaf residue, alkaline protein recovery tracked tissue disruption


508 and co-extraction of rhamnogalacturonan-I pectin and lignin rather than simple protein solubility,


509 although that mechanism was specific to the material and treatment studied (Zhang et al., 2015). Work


510 on tomato leaf likewise suggests that developmental changes in the soluble proteome and inhibitory


511 compounds affect recovery (Kleuter et al., 2025), while recent studies on sugar-beet show that protein


512 content and extractability can be screened at scale and have exploitable genetic variation (Rijken et


513 al., 2025; Rijken et al., 2026).


514


515 Taken together, these results shed new light on the impact of cell wall architectures and glycan


516 diversity in protein, including Rubisco, extraction from plant biomass, and highlight the need for


517 detailed understanding of these effects to optimise extraction protocols.


518


519 **3.5 Deciphering relationships between protein and glycan profiles and implications for Rubisco**


520 **yield.**


521 To better understand correlative, and possibly causal, relationships between profiles obtained for TSP,


522 cell wall glycans and Rubisco, we undertook Pearson correlation coefficient and Principal component


523 analysis (PCA).


19


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


524


525 Pearson correlations were calculated across 128 plant biomass samples for all combinations of 13


526 protein antibodies and 15 glycan antibodies, yielding 195 pairwise correlations. Rows and columns


527 were clustered using Euclidean distance and complete linkage applied to their full correlation profiles


528 ( **Fig. 6** ). Proteins clustering together therefore showed similar patterns of correlation across the 15


529 glycans, while glycans clustering together showed similar relationships across the 13 protein


530 antibodies. Of the 195 correlations, 146 were negative and 49 were positive, with a mean absolute


531 correlation coefficient (| _r_ |) of 0.139.


532


533 **Figure 6.**


534


535 The glycan dendrogram separated a region of mixed or positive correlations, containing LM1 and


536 several antibodies with weak or near null associations, from a predominantly negative block


537 comprising mAbs LM15, LM5, INCh1, BS-400-2, LM11 and CBM21. Many photosynthetic proteins


538 shared broadly similar negative correlation profiles, whereas Lhcb2/Lhcb4 and PEPC/ACT formed


539 more distinct branches.


540


541 Following false discovery rate (FDR) correction, 41 correlations remained significant, comprising 39


542 negative and 2 positive associations. The strongest negative correlations involved PRK–BS-400-2 ( _r_ =


543 −0.365), PsbP–BS-400-2 ( _r_ = −0.347), GLDP–LM11 ( _r_ = −0.340) and Lhcb2–LM25 ( _r_ = −0.337).


544 ACT–LM1 ( _r_ = 0.258) and NTRC–LM1 ( _r_ = 0.256) were the only significant positive correlations.


545 Nine Rubisco–glycan correlations remained significant, all negative. The strongest involved RbcL


546 (2)–BS-400-2 ( _r_ = −0.316), RbcL (1)–LM15 ( _r_ = −0.304), RbcL (1)–INCh1 ( _r_ = −0.297) and RbcS–


547 LM11 ( _r_ = −0.275). LM15 showed significant negative correlations with all three Rubisco antibodies,


548 while LM11 and INCh1 showed consistently negative associations across the same antibodies. This


549 agreement suggests that the observed patterns were not specific to a single Rubisco antibody.


550


20


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


551 Across the biomass samples, stronger Rubisco and photosynthetic-protein signals tended to coincide


552 with weaker signals for certain structural glycan epitopes. However, these associations were generally


553 weak and do not establish that individual glycans directly restrict Rubisco extraction. The


554 predominantly negative correlation block may instead reflect differences in tissue composition or


555 biomass type. LM1 showed a contrasting correlation profile, while several other glycan antibodies had


556 weak or near-null associations, indicating that the relationships between protein and glycan signals


557 were not uniformly negative. In a sensitivity analysis including TSP ( **Fig. S1),** five TSP–glycan


558 correlations remained significant after FDR correction. The strongest negative associations involved


559 BS-400-2 ( _r_ = −0.404) and LM11 ( _r_ = −0.354), suggesting that similar relationships may extend


560 beyond Rubisco to soluble protein more generally. These findings identify glycan epitopes that


561 warrant further investigation as potential markers of biomass extractability, particularly through


562 validation against protein recovery and responses to enzyme treatment. Their associations with protein


563 signals do not, however, establish a causal role in limiting extraction.


564


565 **Figure 7.**


566


567 PCA of the targeted protein-glycan dataset identified two complementary axes of variation across the


568 biomass panel ( **Fig. 7** ). In the targeted PCA ( **Fig. 7A** ), PC1 explained 29.7% of the variance and was


569 dominated by RbcL (1), RbcL (2) and RbcS, which together contributed 82.5% of the axis, whereas


570 the glycan antibodies BS-400-4, LM1, LM11, LM15, LM19 and JIM13 contributed 98.2% of PC2,


571 which explained 19.8% of the variance. Together, PC1 and PC2 accounted for 49.4% of the total


572 variation. All three Rubisco antibodies loaded in the same direction on PC1, whereas LM11 and


573 LM15 loaded oppositely, consistent with the inverse Rubisco-glycan relationships identified in **Fig. 6** .


574 LM1, LM19, JIM13 and BS-400-4 together contributed 96.4% of PC2, indicating a contrast among


575 glycan epitopes rather than a general increase in glycan antibody signal. The six targeted glycans


576 collectively explained 15.9% of the multivariate Rubisco antibody variation in redundancy analysis


577 (adjusted R² = 0.117, BH-adjusted P < 0.001), with LM15 retaining the strongest marginal association


578 (BH-adjusted P = 0.0066). In the HC3 robust multivariable model, LM15 was negatively associated


21


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


579 with Rubisco PC1 (standardised β = -0.300, 95% CI -0.432 to -0.168, BH-adjusted P < 0.001). This


580 relationship was also supported by Pearson and Spearman analyses, identifying LM15 as the most


581 consistent candidate marker of lower Rubisco associated signal across the different analytical


582 approaches.


583


584 When grouped by growth form ( **Data S2** ), forbs (non-graminoid herbaceous flowering plants) were


585 shifted towards the Rubisco associated zone of PC1, woody plants towards the opposing end, and


586 graminoids towards negative PC2 values. A separate Rubisco-only PCA explained 78.7% and 15.2%


587 of the variance on PC1 and PC2, respectively; together, these axes accounted for 93.9% of the


588 variance ( **Fig. 7B-D** ). In this space, forbs and crop plants were descriptively shifted towards the


589 Rubisco-associated end of PC1, whereas woody plants were shifted towards the opposing end; broad


590 taxonomic groups overlapped extensively. Growth form explained 14.6% of the targeted profile


591 variation and 16.1% of the Rubisco only variation (PERMANOVA, both BH-adjusted P < 0.001).


592 However, multivariate dispersion also differed among growth forms (both PERMDISP P = 0.001),


593 indicating that these effects reflected differences in both group position and within group


594 heterogeneity. Crop status produced the strongest group effect on the Rubisco only profile (R² =


595 0.249, BH-adjusted P < 0.001), without a supported difference in dispersion (P = 0.064). Substantial


596 within group variation remained, particularly among the forbs.


597


598 These patterns suggest that variation in Rubisco associated signal is not organised by broad taxonomy


599 alone. Crop status and growth form may provide useful, although imperfect, guidance for candidate


600 selection, while the targeted PCA shows that Rubisco associated signal and cell-wall epitope profiles


601 represent complementary dimensions of biomass phenotype. Complementary HC3 modelling showed


602 that LM11 and LM15 were negatively associated with TSP, whereas LM1 was positively associated;


603 all three directions were also supported by Pearson and Spearman analyses. Together with **Fig. 6**,


604 these findings identify LM15 as the most consistent Rubisco associated marker, LM11 as a candidate


605 marker of the broader soluble protein phenotype, and LM1 as a marker associated with higher TSP


606 rather than Rubisco specifically. The separate protein and glycan heatmaps ( **Fig. 2 and Fig. 5** ) showed


22


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


607 substantial sample level variation, while **Fig. 4, Fig. 6 and Fig. 7** show that TSP, Rubisco antibody


608 and glycan antibody signals covary through complementary pairwise and multivariate analyses. This


609 interpretation is consistent with reports that biomass structure is associated with the success of leaf


610 protein fractionation (Nynäs et al., 2021) and that pectin composition is associated with protein


611 extractability in tomato leaves (Kleuter et al., 2024). However, these associations do not establish that


612 cell-wall composition directly restricts Rubisco extraction. Crop derived biomass may provide a


613 practical starting point for screening, but individual feedstocks must still be evaluated directly and


614 may require biomass-specific processing strategies. These findings provide a basis for selecting


615 contrasting feedstocks and evaluating whether cell-wall degrading enzyme treatments can improve


616 protein recovery biomasses with contrasting cell-wall epitope profiles.


617


618 **3.6 Impact of enzyme treatment on soluble protein yield and Rubisco microarray signals**


619 It has previously been highlighted that cell wall degrading enzymes can enhance protein yield by


620 disassembling the polysaccharide matrix that is the dominant component of biomass, and from which


621 proteins must be released to be extracted effectively (Harholt et al., 2010; Krishnan et al., 2026;


622 Pérez-Vila et al., 2022; Scheller & Ulvskov, 2010; Tamayo Tenorio et al., 2016). To assess these


623 effects in relation to Rubisco yield, we used 7 different enzyme treatments tested against 3 different


624 biomass types ( **Table S3** ). The three biomass types tested were grass, cucumber and tomato due to the


625 abundance of waste generated annually from these crops. It was found that enzyme pretreatment


626 significantly increased soluble protein yield in 51 of 84 biomass matrix preparation comparisons ( **Fig.**


627 **8** **and** **Fig. S2** ).


628


629 **Figure 8.**


630


631 Grass demonstrated the strongest response to enzymatic pre-treatment of all matrices examined, with


632 Celluzyme treatment producing the greatest increase in both in TSP and RbcL (2) abundance ( **Fig. 8**


633 **and** **Fig. S2** ). The convergence of these independent measurements provides strong evidence that


23


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


634 enzymatic degradation of the grass cell wall enhanced the recovery of intracellular proteins,


635 particularly chloroplast associated Rubisco. Similar responses were also observed in grass pulp,


636 suggesting that enzyme assisted extraction is effective not only for intact leaf tissue but also for


637 residual biomass streams that currently have limited commercial value. The marked improvement in


638 protein recovery is consistent with the structural characteristics of grass cell walls. Grasses possess


639 Type II primary cell walls enriched in glucuronoarabinoxylans and mixed-linkage β -glucans, with


640 comparatively low pectin content and frequent ferulate mediated cross linking between


641 polysaccharides and lignin (Carpita & Gibeaut, 1993; Vogel, 2008). These features reduce wall


642 porosity and limit solvent penetration, making mechanical homogenisation alone insufficient to fully


643 release intracellular proteins. Cellulolytic enzyme preparations such as Celluzyme likely weaken this


644 structural network, increasing wall permeability and facilitating disruption of chloroplast containing


645 mesophyll cells. Similar mechanisms have been reported during enzyme assisted extraction of


646 proteins from plant biomass, where degradation of structural polysaccharides enhances diffusion of


647 soluble proteins into the extraction medium (Sari et al., 2013; Wang et al., 2025). Previous studies


648 have demonstrated that matching enzyme composition to cell wall architecture significantly improves


649 biomass deconstruction and intracellular product recovery (Park et al., 2016; Shawky et al., 2025).


650 Consequently, optimisation of enzyme formulations for monocotyledonous biomass may further


651 enhance protein yields while reducing enzyme loading and processing costs.


652


653 Cucumber exhibited a positive but less consistent response to enzymatic pre-treatment than grass.


654 Celluzyme produced the highest TSP recovery from frozen-thawed leaves ( **Fig. 8A** ), suggesting that


655 combining freeze-thaw disruption with enzymatic hydrolysis improved access to intracellular


656 proteins. Freeze-thaw processing is known to induce membrane rupture through ice crystal formation,


657 thereby increasing tissue permeability and enhancing subsequent enzyme penetration (Mazur, 1984).


658 The present findings therefore suggest that physical disruption and enzymatic degradation act


659 synergistically to improve protein extraction from cucumber tissues.


660


24


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


661 In contrast to TSP recovery, RbcL abundance did not consistently increase following the same


662 treatments. Macerozyme generated the highest Rubisco signal in freeze-dried leaves and pulp,


663 whereas Celluzyme produced the strongest response in frozen-thawed leaves ( **Fig. 8B and Fig. S2B** ).


664 This divergence suggests that the mechanisms governing total protein extraction differ from those


665 controlling chloroplast protein recovery. While Celluzyme may effectively disrupt structural cellulose


666 networks and release soluble cytoplasmic proteins, Macerozyme may more effectively degrade pectic


667 components surrounding chloroplast rich tissues or better preserve Rubisco during extraction. Similar


668 differences between total protein recovery and individual protein abundance have been reported in


669 enzyme assisted extraction studies, reflecting the complexity of plant tissue organisation and the


670 compartmentalisation of intracellular proteins (Komatsu & Jorrin-Novo, 2021; Wang et al., 2025).


671 Compared with grass pulp, cucumber pulp showed smaller improvements in both TSP and Rubisco


672 abundance. This may reflect differences in tissue composition, as cucumber tissues possess relatively


673 thin primary cell walls, high water content and lower structural polysaccharide concentrations than


674 grass (Cosgrove, 2005). Consequently, a larger proportion of proteins may already be accessible


675 following mechanical disruption, reducing the additional benefit provided by enzymatic hydrolysis.


676 Alternatively, endogenous protease activity released during tissue disruption may contribute to partial


677 degradation of soluble proteins before extraction. Protease mediated losses have previously been


678 identified as a significant limitation during protein recovery from fresh green biomass (Bals & Dale,


679 2012).


680


681 Tomato displayed a clear dependence on both processing method and enzyme formulation. Enzymatic


682 effects were substantially greater in frozen-thawed samples than in freeze-dried material, with


683 Macerozyme and the Pectinex + Macerozyme combination consistently producing the highest TSP


684 recoveries from both leaf and pulp tissues ( **Fig. 8A** **and** **Figure S2A** ). These findings indicate that


685 freeze-thaw processing enhanced enzyme accessibility and that degradation of pectin-rich cell wall


686 domains was particularly effective in improving protein extraction. The observed response is


687 consistent with the biochemical composition of tomato cell walls, which are characteristic of Type I


688 primary walls containing abundant pectin, xyloglucans and cellulose microfibrils (Carpita & Gibeaut,


25


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


689 1993). Pectin contributes significantly to cell adhesion through the middle lamella, and enzymatic


690 degradation of these polymers reduces tissue cohesion, facilitating cell separation and intracellular


691 protein release (Willats et al., 2001). The superior performance of Pectinex-containing treatments


692 therefore supports the hypothesis that pectin degradation is a key mechanism governing protein


693 extraction from tomato tissues. This interpretation is further supported by the strongest RbcL (2)


694 responses observed in tomato pulp following treatment with Pectinex + Macerozyme ( **Fig. 8B** **and**


695 **Fig. S2B** ). Tomato processing generates substantial quantities of pomace and pulp that retain


696 appreciable amounts of protein despite being predominantly utilised for low value applications such


697 as animal feed or compost (Reguengo et al., 2022). Improving protein recovery from these residues


698 could therefore contribute to increased resource efficiency within tomato processing industries and


699 support circular bioeconomy strategies. Similar approaches have recently been proposed for the


700 valorisation of vegetable processing side streams through enzyme assisted extraction technologies


701 (Shawky et al., 2025; Wang et al., 2025). Collectively, the tomato results demonstrate that enzyme


702 specificity should be matched to tissue composition, with pectin degrading enzymes offering


703 particular advantages for dicotyledonous tissues rich in pectic polysaccharides. The strong responses


704 observed in tomato pulp further suggest that vegetable processing residues represent promising


705 substrates for sustainable protein recovery.


706


707 Overall, these findings demonstrate the potential of enzyme assisted extraction to improve protein


708 recovery across different plant biomass types, although extraction efficiency and Rubisco recovery


709 depend on substrate composition, enzyme specificity and the intracellular localisation of target


710 proteins. Grass exhibited the strongest response, particularly following Celluzyme treatment,


711 suggesting that cellulose rich structural components represent an important barrier to protein release.


712 Cucumber also benefited from enzymatic treatment, although the optimal enzyme preparation may


713 differ depending on whether maximising total protein recovery or preserving Rubisco is the primary


714 objective. In tomato, increased TSP recovery was not always accompanied by corresponding increases


715 in RbcL abundance, highlighting the importance of distinguishing total protein recovery from target


26


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


716 protein abundance. Such discrepancies may reflect biological variation or matrix-dependent effects on


717 immunodetection rather than changes in Rubisco recovery alone (Komatsu & Jorrin-Novo, 2021).


718 Consequently, further validation using complementary analytical techniques, such as SDS-PAGE or


719 targeted LC-MS/MS, would help clarify treatment-specific changes in Rubisco abundance. Future


720 optimisation should therefore consider both enzyme specificity and substrate composition, alongside


721 the preservation and reliable quantification of target proteins, to maximise the potential of enzyme


722 assisted extraction for sustainable protein recovery from agricultural biomass.


723 **4.** **Conclusion**


724 There is broad agreement that Rubisco is a promising source of plant protein that could make a


725 valuable contribution to balancing future global nutritional needs with environmental considerations.


726 However, it is also clear that significant optimisation of extraction protocols is required if this


727 potential is to be realised in a cost-feasible way. In this work we demonstrate how a high-throughput


728 technology platform (MAPP), can provide important new insights into three parameters that play key


729 roles in determining Rubisco yield: biomass source; cell wall matrix composition and enzymatic


730 disruption of cell wall polymers.


731


732 Our results show that Rubisco abundance varies considerably across plant biomass types, and that


733 abundance is not directly linked to taxonomic groupings. To our knowledge, this is the first parallel


734 analysis of cell wall matrix polysaccharides and soluble proteins across a diverse plant biomass panel,


735 allowing relationships between these glycans and Rubisco yield to be investigated. We explored these


736 effects further by demonstrating how the targeted use of cell wall degrading enzymes can enhance


737 Rubisco yield. Additionally, crop plants generally produced stronger Rubisco-associated signals, with


738 potato haulm and Brassica leaf residues emerging as promising feedstocks for further extraction and


739 valorisation.


740


741 This work highlights the importance of understanding biomass properties when developing Rubisco


742 extraction methods and demonstrates that MAPP can aid in addressing this knowledge gap. We


27


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


743 anticipate that MAPP could also be used to examine other sources of variation in Rubisco yield and


744 quality, including seasonal and developmental variation, batch-to-batch variation, and Rubisco purity


745 and composition.


746 **5.** **CRediT authorship contribution statement**


747 **John C. D. Pattinson:** Conceptualisation, data curation, formal analysis, investigation, methodology,


748 project administration, resources, software, validation, visualization, writing – original draft, writing –


749 review and editing. **Charlotte S. Fletcher:** Conceptualisation, data curation, formal analysis,


750 investigation, methodology, project administration, resources, software, validation, visualization,


751 writing – original draft, writing – review and editing. **Ieva Lelenaite:** Project administration, writing


752 – review and editing. **Georg Lietz:** Supervision **William G. T. Willats:** Funding acquisition,


753 Supervision, writing – review and editing. **Maxim. V. Kapralov:** Funding acquisition, Supervision,


754 writing – review and editing.


755 **6.** **Declaration of competing interest**


756 The authors declare that they have no known competing financial interests or personal relationships


757 that could have appears to influence the work reported in this paper.


758 **7.** **Acknowledgements**


759 The authors would like to acknowledge Ewelina Bien and Max Kelly for their support and guidance.


760 This research was supported by a PhD studentship funded through the Institute for Agri-Food


761 Research and Innovation (IAFRI) Doctoral Training Programme at Newcastle University (JCDP).


762 This work was supported by a PhD studentship funded by the BBSRC in relation to the BiSCoP


763 (Bioscience for Sustainable Consumer Products) Collaborative Training Partnership.


764 **8.** **Appendix A. Supplementary data**


765 Data S1


28


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


766 Data S2


767


768 **9.** **Appendix B. Supplementary figures and tables**


769 Table S1


770 Table S2


771 Table S3


772 Fig. S1


773 Fig. S2


774 **10.** **Data availability**


775 Data will be made available upon request.


776 **11.** **References**


777 Aiking, H., & de Boer, J. (2020). The next protein transition. Trends in Food Science & Technology,
778 105, 515–522. https://doi.org/10.1016/j.tifs.2018.07.008


779 Anderson, C. T., & Kieber, J. J. (2020). Dynamic construction, perception, and remodeling of plant
780 cell walls. Annual Review of Plant Biology, 71(1), 39–69. https://doi.org/10.1146/annurev-arplant781 081519-035846


782 Andersson, I., & Backlund, A. (2008). Structure and function of Rubisco. Plant Physiology and
783 Biochemistry, 46(3), 275–291. https://doi.org/10.1016/j.plaphy.2008.01.001


784 Aragón, C., Pascual, P., González, J., Escalona, M., Carvalho, L., & Amancio, S. (2013). The
785 physiology of ex vitro pineapple (Ananas comosus L. Merr. var. MD-2) as CAM or C3 is regulated by
786 the environmental conditions: Proteomic and transcriptomic profiles. Plant Cell Reports, 32(11),
787 1807–1818. https://doi.org/10.1007/s00299-013-1493-3


788 Bakshani, C. R., Sangta, J., Sommano, S., & Willats, W. G. T. (2023). Microarray polymer profiling
789 (MAPP) for high-throughput glycan analysis. Journal of Visualized Experiments, (199), Article
790 65443. https://doi.org/10.3791/65443


791 Bals, B. D., & Dale, B. E. (2012). Developing a model for assessing biomass processing technologies
792 within a local biomass processing depot. Bioresource Technology, 106, 161–169.
793 https://doi.org/10.1016/j.biortech.2011.12.024


794 Bar-On, Y. M., & Milo, R. (2019). The global mass and average rate of Rubisco. Proceedings of the
795 National Academy of Sciences, 116(10), 4738–4743. https://doi.org/10.1073/pnas.1816654116


29


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


796 Boraston, A. B., Revett, T. J., Boraston, C. M., Nurizzo, D., & Davies, G. J. (2003). Structural and the
797 rmodynamic dissection of specific mannan recognition by a carbohydrate binding module, TmCBM2
798 7. Structure, 11(6), 665–675. https://doi.org/10.1016/S0969-2126(03)00100-X


799 Bradford, M. M. (1976). A rapid and sensitive method for the quantitation of microgram quantities of
800 protein utilizing the principle of protein-dye binding. Analytical Biochemistry, 72(1–2), 248–254.
801 https://doi.org/10.1016/0003-2697(76)90527-3


802 Burne, R. A., Schilling, K., Bowen, W. H., & Yasbin, R. E. (1987). Expression, purification, and
803 characterization of an exo- β -D-fructosidase of Streptococcus mutans. Journal of Bacteriology,
804 169(10), 4507–4517. https://doi.org/10.1128/jb.169.10.4507-4517.1987


805 Carpita, N. C. (2011). Update on mechanisms of plant cell wall biosynthesis: How plants make
806 cellulose and other (1 → 4)- β -D-glycans. Plant Physiology, 155(1), 171–184.
807 https://doi.org/10.1104/pp.110.163360


808 Carpita, N. C., & Gibeaut, D. M. (1993). Structural models of primary cell walls in flowering plants:
809 Consistency of molecular structure with the physical properties of the walls during growth. The Plant
810 Journal, 3(1), 1–30. https://doi.org/10.1111/j.1365-313x.1993.tb00007.x


811 Chen, P., Yan, K., Shao, H., & Zhao, S. (2013). Physiological mechanisms for high salt tolerance in w
812 ild soybean (Glycine soja) from Yellow River Delta, China: Photosynthesis, osmotic regulation, ion fl
813 ux and antioxidant capacity. PLOS ONE, 8(12), Article e83227. https://doi.org/10.1371/journal.pone.
814 0083227


815 Chen, Y.-E., Zhang, C.-M., Su, Y.-Q., Ma, J., Zhang, Z.-W., Yuan, M., Zhang, H.-Y., & Yuan, S.
816 (2017). Responses of photosystem II and antioxidative systems to high light and high temperature co817 stress in wheat. Environmental and Experimental Botany, 135, 45–55.
818 https://doi.org/10.1016/j.envexpbot.2016.12.001


819 Clausen, M. H., Ralet, M.-C., Willats, W. G. T., McCartney, L., Marcus, S. E., Thibault, J.820 F., & Knox, J. P. (2004). A monoclonal antibody to feruloylated-(1 → 4)- β -D821 galactan. Planta, 219(6), 1036–1041. https://doi.org/10.1007/s00425-004-1309-3


822 Cornuault, V., Buffetto, F., Marcus, S. E., Crépeau, M., Guillon, F., Ralet, M.-C., & Knox, J. P.
823 (2017). LM6-M: A high avidity rat monoclonal antibody to pectic α -1,5-L-arabinan. BioRxiv. 161604
824 https://doi.org/10.1101/161604


825 Cornuault, V., Buffetto, F., Rydahl, M. G., Marcus, S. E., Torode, T. A., Xue, J., Crépeau, M.-J.,
826 Faria-Blanc, N., Willats, W. G. T., Dupree, P., Ralet, M.-C., & Knox, J. P. (2015). Monoclonal
827 antibodies indicate low-abundance links between heteroxylan and other glycans of plant cell walls.
828 Planta, 242(6), 1321–1334. https://doi.org/10.1007/s00425-015-2375-4


829 Cosgrove, D. J. (2018). Diffuse growth of plant cell walls. Plant Physiology, 176(1), 16–27.
830 https://doi.org/10.1104/pp.17.01541


831 Cosgrove, D. J. (2005). Growth of the plant cell wall. Nature Reviews Molecular Cell Biology, 6(11),
832 850–861. https://doi.org/10.1038/nrm1746


833 Day, L. (2013). Proteins from land plants—Potential resources for human nutrition and food security.
834 Trends in Food Science & Technology, 32(1), 25–42. https://doi.org/10.1016/j.tifs.2013.05.005


835 Duki ć, J., Košpi ć, K., Kelava, V., Mavri ć, R., Nutrizio, M., Balen, B., Butorac, A., Öztop, M. H., &
836 Režek Jambrak, A. (2023). Alternative methods for Rubisco extraction from sugar beet waste: A


30


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


837 comparative approach of ultrasound and high voltage electrical discharge. Ultrasonics Sonochemistry,
838 99, Article 106535. https://doi.org/10.1016/j.ultsonch.2023.106535


839 Ellis, R. J. (1979). The most abundant protein in the world. Trends in Biochemical Sciences, 4(11),
840 241–244. https://doi.org/10.1016/0968-0004(79)90212-3


841 Fangel, J. U., Sørensen, K. M., Jacobsen, N., Mravec, J., Ahl, L. I., Bakshani, C., Mikkelsen, M. D.,
842 Engelsen, S. B., Willats, W., & Ulvskov, P. (2024). The legacy of terrestrial plant evolution on cell
843 wall fine structure. Plant, Cell & Environment, 47(4), 1238–1254. https://doi.org/10.1111/pce.14785


844 Fangel, J. U., Ulvskov, P., Knox, J. P., Mikkelsen, M. D., Harholt, J., Popper, Z. A., & Willats, W. G.
845 T. (2012). Cell wall evolution and diversity. Frontiers in Plant Science, 3, Article 152.
846 https://doi.org/10.3389/fpls.2012.00152


847 Food and Agriculture Organization of the United Nations. (2013). Dietary protein quality evaluation
848 in human nutrition (FAO Food and Nutrition Paper No. 92).
849 https://www.fao.org/ag/humannutrition/35978-02317b979a686a57aa4593304ffc17f06.pdf


850 Fradera-Soler, M., Leverett, A., Mravec, J., Jørgensen, B., Borland, A. M., & Grace, O. M. (2022).
851 Are cell wall traits a component of the succulent syndrome? Frontiers in Plant Science, 13, Article
852 1043429. https://doi.org/10.3389/fpls.2022.1043429


853 Galmés, J., Kapralov, M. V., Andralojc, P. J., Conesa, M. À., Keys, A. J., Parry, M. A. J., & Flexas, J.
854 (2014). Expanding knowledge of the Rubisco kinetics variability in plant species: Environmental and
855 evolutionary trends. Plant, Cell & Environment, 37(9), 1989–2001. https://doi.org/10.1111/pce.12335


856 Grimmer, J., Rödiger, A., Hoehenwarter, W., Helm, S., & Baginsky, S. (2014). The RNA857 binding protein RNP29 is an unusual Toc159 transport substrate. Frontiers in Plant Science, 5, 258. htt
858 ps://doi.org/10.3389/fpls.2014.00258


859 Harholt, J., Suttangkakul, A., & Vibe Scheller, H. (2010). Biosynthesis of pectin. Plant Physiology,
860 153(2), 384–395. https://doi.org/10.1104/pp.110.156588


861 Hatch, M. D. (1987). C4 photosynthesis: A unique blend of modified biochemistry, anatomy and
862 ultrastructure. Biochimica et Biophysica Acta (BBA)—Reviews on Bioenergetics, 895(2), 81–106.
863 https://doi.org/10.1016/S0304-4173(87)80009-5


864 Ishimaru, K., Kobayashi, N., Ono, K., Yano, M., & Ohsugi, R. (2001). Are contents of Rubisco,
865 soluble protein and nitrogen in flag leaves of rice controlled by the same genetics? Journal of
866 Experimental Botany, 52(362), 1827–1833. https://doi.org/10.1093/jexbot/52.362.1827


867 Jones, L., Seymour, G. B., & Knox, J. P. (1997). Localization of pectic galactan in tomato cell walls
868 using a monoclonal antibody specific to (1 → 4)- β -D-galactan. Plant Physiology, 113(4), 1405–1412.
869 https://doi.org/10.1104/pp.113.4.1405


870 Khoshravesh, R., Stata, M., Busch, F. A., Saladié, M., Castelli, J. M., Dakin, N., Hattersley, P. W.,
871 Macfarlane, T. D., Sage, R. F., Ludwig, M., & Sage, T. L. (2020). The evolutionary origin of C �
872 photosynthesis in the grass subtribe Neurachninae. Plant Physiology, 182(1), 566–583.
873 https://doi.org/10.1104/pp.19.00925


874 Kleuter, M., Yu, Y., Pancaldi, F., Nagtzaam, M., van der Goot, A. J., & Trindade, L. M. (2024). Cell
875 wall as a barrier for protein extraction from tomato leaves: A biochemical study. Plant Physiology and
876 Biochemistry, 208, 108495. https://doi.org/10.1016/j.plaphy.2024.108495


877 Kleuter, M., Yu, Y., Verdegaal, L., Pancaldi, F., America, A. H. P., van der Goot, A. J., & Trindade, L.
878 M. (2025). Characterizing the extractable proteins from tomato leaves—A proteomics study. Food
879 Chemistry: X, 25, Article 102114. https://doi.org/10.1016/j.fochx.2024.102114


880 Knox, J. (2008). Revealing the structural and functional diversity of plant cell walls. Current Opinion
881 in Plant Biology, 11(3), 308–313. https://doi.org/10.1016/j.pbi.2008.03.001


31


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


882 Komatsu, S., & Jorrin-Novo, J. V. (2021). Plant proteomic research 3.0: Challenges and perspectives.
883 International Journal of Molecular Sciences, 22(2), Article 766. https://doi.org/10.3390/ijms22020766


884 Krishnan, H., Lakshmipathy, B., Ramaswamy, J., Nallamuthu, I., Dubey, M., Xu, B., Reddy, C. K., &
885 Kumar, M. (2026). Green extraction of Rubisco-rich leaf protein for sustainable food systems: Process
886 efficiency, functional quality, and sustainability assessment. Food Chemistry, 521, Article 149914.
887 https://doi.org/10.1016/j.foodchem.2026.149914


888 Kyriakopoulou, K., Dekkers, B., & van der Goot, A. J. (2019). Plant-based meat analogues. In
889 Sustainable meat production and processing (pp. 103–126). Elsevier. https://doi.org/10.1016/B978-0890 12-814874-7.00006-7


891 Lang, E. G. E., Mueller, S. J., Hoernstein, S. N. W., Porankiewicz-Asplund, J., Vervliet-Scheebaum,
892 M., & Reski, R. (2011). Simultaneous isolation of pure and intact chloroplasts and mitochondria from
893 moss as the basis for sub-cellular proteomics. Plant Cell Reports, 30(2), 205–215.
894 https://doi.org/10.1007/s00299-010-0935-4


895 Little, A., Schwerdt, J. G., Shirley, N. J., Khor, S. F., Neumann, K., O’Donovan, L. A., Lahnstein, J.,
896 Collins, H. M., Henderson, M., Fincher, G. B., & Burton, R. A. (2018). Revised phylogeny of the
897 Cellulose Synthase gene superfamily: Insights into cell wall evolution. Plant Physiology, 177(3),
898 1124–1141. https://doi.org/10.1104/pp.17.01718


899 Luo, X., Keenan, T. F., Chen, J. M., Croft, H., Prentice, I. C., Smith, N. G., Walker, A. P., Wang, H.,
900 Wang, R., Xu, C., & Zhang, Y. (2021). Global variation in the fraction of leaf nitrogen allocated to
901 photosynthesis. Nature Communications, 12(1), Article 4866. https://doi.org/10.1038/s41467-021902 25163-9


903 Ma, K. K., Greis, M., Lu, J., Nolden, A. A., McClements, D. J., & Kinchla, A. J. (2022). Functional
904 performance of plant proteins. Foods, 11(4), Article 594. https://doi.org/10.3390/foods11040594


905 Machovi č, M., Svensson, B., MacGregor, E. A., & Jane č ek, Š. (2005). A new clan of CBM families
906 based on bioinformatics of starch-binding domains from families CBM20 and CBM21. The FEBS
907 Journal, 272(21), 5497–5513. https://doi.org/10.1111/j.1742-4658.2005.04942.x


908 Marcus, S. E., Blake, A. W., Benians, T. A. S., Lee, K. J. D., Poyser, C., Donaldson, L., Leroux, O., R
909 ogowski, A., Petersen, H. L., Boraston, A., Gilbert, H. J., Willats, W. G. T., & Knox, J. P. (2010). Res
910 tricted access of proteins to mannan polysaccharides in intact plant cell walls. The Plant Journal, 64(2)
911, 191–203. https://doi.org/10.1111/j.1365-313X.2010.04319.x


912 Marcus, S. E., Verhertbruggen, Y., Hervé, C., Ordaz-Ortiz, J. J., Farkas, V., Pedersen, H. L., Willats,
913 W. G. T., & Knox, J. P. (2008). Pectic homogalacturonan masks abundant sets of xyloglucan epitopes
914 in plant cell walls. BMC Plant Biology, 8, Article 60. https://doi.org/10.1186/1471-2229-8-60


915 Martin, A. H., Castellani, O., de Jong, G. A., Bovetto, L., & Schmitt, C. (2019). Comparison of the
916 functional properties of Rubisco protein isolate extracted from sugar beet leaves with commercial
917 whey protein and soy protein isolates. Journal of the Science of Food and Agriculture, 99(4), 1568–
918 1576. https://doi.org/10.1002/jsfa.9335


919 Mayfield, A. B., Hsiao, Y.-Y., Chen, H.-K., et al. (2014). Rubisco expression in the dinoflagellate
920 Symbiodinium sp. is influenced by both photoperiod and endosymbiotic lifestyle. Marine
921 Biotechnology, 16(4), 371–384. https://doi.org/10.1007/s10126-014-9558-z


922 Mazur, P. (1984). Freezing of living cells: Mechanisms and implications. American Journal of
923 Physiology—Cell Physiology, 247(3), C125–C142. https://doi.org/10.1152/ajpcell.1984.247.3.C125


924 McCartney, L., Marcus, S. E., & Knox, J. P. (2005). Monoclonal antibodies to plant cell wall xylans
925 and arabinoxylans. Journal of Histochemistry & Cytochemistry, 53(4), 543–546.
926 https://doi.org/10.1369/jhc.4B6578.2005


32


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


927 McClements, D. J., & Grossmann, L. (2021). The science of plant-based foods: Constructing next928 generation meat, fish, milk, and egg analogs. Comprehensive Reviews in Food Science and Food
929 Safety, 20(4), 4049–4100. https://doi.org/10.1111/1541-4337.12771


930 Meikle, P. J., Bönig, I., Hoogenraad, N. J., Clarke, A. E., & Stone, B. A. (1991). The location of
931 (1 → 3)- β -glucans in the walls of pollen tubes of Nicotiana alata using a (1 → 3)- β -glucan-specific
932 monoclonal antibody. Planta, 185(1), 1–8. https://doi.org/10.1007/BF00194507


933 Meikle, P. J., Hoogenraad, N. J., Bönig, I., Clarke, A. E., & Stone, B. A. (1994). A (1 → 3,1 → 4)- β  934 glucan-specific monoclonal antibody and its use in the quantitation and immunocytochemical location
935 of (1 → 3,1 → 4)- β -glucans. The Plant Journal, 5(1), 1–9. https://doi.org/10.1046/j.1365936 313X.1994.5010001.x


937 Møller, I., Sørensen, I., Bernal, A. J., Blaukopf, C., Lee, K., Øbro, J., Pettolino, F., Roberts, A.,
938 Mikkelsen, J. D., Knox, J. P., Bacic, A., & Willats, W. G. T. (2007). High-throughput mapping of cell939 wall polymers within and between plants using novel microarrays. The Plant Journal, 50(6), 1118–
940 1128. https://doi.org/10.1111/j.1365-313x.2007.03114.x


941 Najmudin, S., Guerreiro, C. I. P. D., Carvalho, A. L., Prates, J. A. M., Correia, M. A. S., Alves, V. D.,
942 Ferreira, L. M. A., Romão, M. J., Gilbert, H. J., Bolam, D. N., & Fontes, C. M. G. A. (2006).
943 Xyloglucan is recognized by carbohydrate-binding modules that interact with β -glucan chains. Journal
944 of Biological Chemistry, 281(13), 8815–8828. https://doi.org/10.1074/jbc.M510559200


945 Nieuwland, M., Geerdink, P., Engelen-Smit, N. P. E., van der Meer, I. M., America, A. H. P., Mes, J.
946 J., Kootstra, A. M. J., Henket, J. T. M. M., & Mulder, W. J. (2021). Isolation and gelling properties of
947 duckweed protein concentrate. ACS Food Science & Technology, 1(5), 908–916.
948 https://doi.org/10.1021/acsfoodscitech.1c00009


949 Nikkanen, L., Toivola, J., & Rintamäki, E. (2016). Crosstalk between chloroplast thioredoxin systems
950 in regulation of photosynthesis. Plant, Cell & Environment, 39(8), 1691–
951 1705. https://doi.org/10.1111/pce.12718


952 Nynäs, A.-L., Newson, W. R., & Johansson, E. (2021). Protein Fractionation of Green Leaves as an
953 Underutilized Food Source—Protein Yield and the Effect of Process Parameters. Foods, 10(11), 2533.
954 https://doi.org/10.3390/foods10112533


955 Park, S., Ong, R. G., & Sticklen, M. (2016). Strategies for the production of cell wall-deconstructing
956 enzymes in lignocellulosic biomass and their utilization for biofuel production. Plant Biotechnology
957 Journal, 14(6), 1329–1344. https://doi.org/10.1111/pbi.12505


958 Parry, M. A. J., Andralojc, P. J., Scales, J. C., Salvucci, M. E., Carmo-Silva, A. E., Alonso, H., &
959 Whitney, S. M. (2013). Rubisco activity and regulation as targets for crop improvement. Journal of
960 Experimental Botany, 64(3), 717–730. https://doi.org/10.1093/jxb/ers336


961 Pattathil, S., Avci, U., Baldwin, D., Swennes, A. G., McGill, J. A., Popper, Z., Bootten, T., Albert, A.,
962 Davis, R. H., Chennareddy, C., Dong, R., O’Shea, B., Rossi, R., Leoff, C., Freshour, G., Narra, R.,
963 O’Neil, M., York, W. S., & Hahn, M. G. (2010). A comprehensive toolkit of plant cell wall glycan964 directed monoclonal antibodies. Plant Physiology, 153(2), 514–525.
965 https://doi.org/10.1104/pp.109.151985


966 Pattathil, S., Hahn, M. G., Dale, B. E., & Chundawat, S. P. S. (2015). Insights into plant cell wall
967 structure, architecture, and integrity using glycome profiling of native and AFEX™-pre-treated
968 biomass. Journal of Experimental Botany, 66(14), 4279–4294. https://doi.org/10.1093/jxb/erv107


969 Pedersen, H. L., Fangel, J. U., McCleary, B., Ruzanski, C., Rydahl, M. G., Ralet, M.-C., Farkas, V.,
970 von Schantz, L., Marcus, S. E., Andersen, M. C. F., Field, R., Ohlin, M., Knox, J. P., Clausen, M. H.,
971 & Willats, W. G. T. (2012). Versatile high-resolution oligosaccharide microarrays for plant


33


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


972 glycobiology and cell wall research. Journal of Biological Chemistry, 287(47), 39429–39438.
973 https://doi.org/10.1074/jbc.M112.396598


974 Pennell, R. I., Knox, J. P., Scofield, G. N., Selvendran, R. R., & Roberts, K. (1989). A family of
975 abundant plasma membrane-associated glycoproteins related to the arabinogalactan proteins is unique
976 to flowering plants. The Journal of Cell Biology, 108(5), 1967–1977.
977 https://doi.org/10.1083/jcb.108.5.1967


978 Pérez-Vila, S., Fenelon, M. A., O’Mahony, J. A., & Gómez-Mascaraque, L. G. (2022). Extraction of
979 plant protein from green leaves: Biomass composition and processing considerations. Food
980 Hydrocolloids, 133, Article 107902. https://doi.org/10.1016/j.foodhyd.2022.107902


981 Pettolino, F. A., Hoogenraad, N. J., Ferguson, C., Bacic, A., Johnson, E., & Stone, B. A. (2001). A
982 (1 → 4)- β -mannan-specific monoclonal antibody and its use in the immunocytochemical location of
983 galactomannans. Planta, 214(2), 235–242. https://doi.org/10.1007/s004250100606


984 Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and
985 consumers. Science, 360(6392), 987–992. https://doi.org/10.1126/science.aaq0216


986 Recuenco-Muñoz, L., Offre, P., Valledor, L., Lyon, D., Weckwerth, W., & Wienkoop, S. (2015).
987 Targeted quantitative analysis of a diurnal Rubisco subunit expression and translation profile in
988 Chlamydomonas reinhardtii introducing a novel Mass Western approach. Journal of Proteomics, 113,
989 143–153. https://doi.org/10.1016/j.jprot.2014.09.026


990 Reguengo, L. M., Salgaço, M. K., Sivieri, K., & Maróstica Júnior, M. R. (2022). Agro-industrial by991 products: Valuable sources of bioactive compounds. Food Research International, 152, Article
992 110871. https://doi.org/10.1016/j.foodres.2021.110871


993 Rhodes, P. R., Kung, S. D., & Marsho, T. V. (1980). Relationship of ribulose-1,5-bisphosphate
994 carboxylase-oxygenase specific activity to subunit composition. Plant Physiology, 65(1), 69–73.
995 https://doi.org/10.1104/pp.65.1.69


996 Ridley, B. L., O’Neill, M. A., & Mohnen, D. (2001). Pectins: Structure, biosynthesis, and
997 oligogalacturonide-related signaling. Phytochemistry, 57(6), 929–967. https://doi.org/10.1016/S0031998 9422(01)00113-3


999 Rijken, H., Oertel, H. F., Hirschmann, C. B., Tossens, A., Hackenberg, D., Vegas, J., Bruins, M. E., &
1000 Trindade, L. M. (2025). High-throughput methods for measuring protein extractability in sugar beet
1001 (Beta vulgaris L.) leaves. Journal of the Science of Food and Agriculture, 105(15), 8412–8421.
1002 https://doi.org/10.1002/jsfa.70074


1003 Rijken, H., Saleem, A., Hirschmann, C. B., Hackenberg, D., Bruins, M. E., & Trindade, L. M. (2026).
1004 Leaf protein extractability in sugar beet: Genetic control and breeding potential for dual-purpose
1005 cultivars. Field Crops Research, 336, Article 110215. https://doi.org/10.1016/j.fcr.2025.110215


1006 Ruprecht, C., & Pfrengle, F. (2022). Synthetic plant glycan microarrays as tools for plant biology. In
1007 M. Kilcoyne & J. Q. Gerlach (Eds.), Glycan microarrays: Methods and protocols (Methods in Molecu
1008 lar Biology, Vol. 2460). Humana. https://doi.org/10.1007/978-1-0716-2148-6_7


1009 Ruprecht, C., Bartetzko, M. P., Senf, D., Dallabernardina, P., Boos, I., Andersen, M. C. F., Kotake,
1010 T., Knox, J. P., Hahn, M. G., Clausen, M. H., & Pfrengle, F. (2017). A synthetic glycan microarray
1011 enables epitope mapping of plant cell wall glycan-directed antibodies. Plant Physiology, 175(3),
1012 1094-1104. https://doi.org/10.1104/pp.17.00737


1013 Rydahl, M. G., Kra č un, S. K., Fangel, J. U., et al. (2017). Development of novel monoclonal
1014 antibodies against starch and ulvan—implications for antibody production against polysaccharides
1015 with limited immunogenicity. Scientific Reports, 7, Article 9326. https://doi.org/10.1038/s41598-0171016 04307-2


34


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1017 Sage, R. F., & Zhu, X.-G. (2011). Exploiting the engine of C � photosynthesis. Journal of
1018 Experimental Botany, 62(9), 2989–3000. https://doi.org/10.1093/jxb/err179


1019 Sari, Y. W., Bruins, M. E., & Sanders, J. P. M. (2013). Enzyme assisted protein extraction from
1020 rapeseed, soybean, and microalgae meals. Industrial Crops and Products, 43, 78–83.
1021 https://doi.org/10.1016/j.indcrop.2012.07.014


1022 Scheller, H. V., & Ulvskov, P. (2010). Hemicelluloses. Annual Review of Plant Biology, 61(1), 263–
1023 289. https://doi.org/10.1146/annurev-arplant-042809-112315


1024 Shawky, E., Zhu, W., & Tian, J. (2025). A review of innovative extraction technologies for protein
1025 recovery from plant-based by-products: A step toward zero-waste processing. International Journal of
1026 Biological Macromolecules, 315, Article 144301. https://doi.org/10.1016/j.ijbiomac.2025.144301


1027 Sharma, A., Sharma, S., Ramaraju, G., Rasane, P., Ercisli, S., & Singh, J. (2025). Enzyme-assisted
1028 extraction of leaf proteins: Efficiency, functionality, and structural insights. Food Chemistry: X, 31,
1029 Article 103181. https://doi.org/10.1016/j.fochx.2025.103181


1030 Shewry, P. R. (2007). Improving the protein content and composition of cereal grain. Journal of
1031 Cereal Science, 46(3), 239–250. https://doi.org/10.1016/j.jcs.2007.06.006


1032 Showalter, A. M. (1993). Structure and function of plant cell wall proteins. The Plant Cell, 5(1), 9–23.
1033 https://doi.org/10.1105/tpc.5.1.9


1034 Smallwood, M., Martin, H., & Knox, J. P. (1995). An epitope of rice threonine- and hydroxyproline1035 rich glycoprotein is common to cell wall and hydrophobic plasma-membrane glycoproteins. Planta,
1036 196, 510–522. https://doi.org/10.1007/BF00203651


1037 Somerville, C., Bauer, S., Brininstool, G., Facette, M., Hamann, T., Milne, J., Osborne, E., Paredez,
1038 A., Persson, S., Raab, T., Vorwerk, S., & Youngs, H. (2004). Toward a systems approach to
1039 understanding plant cell walls. Science, 306(5705), 2206–2211.
1040 https://doi.org/10.1126/science.1102765


1041 Song, I., Kim, D. S., Kim, M. K., Jamal, A., Hwang, K. A., & Ko, K. (2015). Comparison of total
1042 soluble protein in various horticultural crops and evaluation of its quantification methods.
1043 Horticulture, Environment, and Biotechnology, 56(1), 123–129. https://doi.org/10.1007/s13580-0151044 0097-y


1045 Sørensen, I., Pettolino, F. A., Wilson, S. M., Doblin, M. S., Johansen, B., Bacic, A., et al. (2008).
1046 Mixed-linkage (1 → 3),(1 → 4)- β -D-glucan is not unique to the Poales and is an abundant component of
1047 Equisetum arvense cell walls. The Plant Journal, 54(3), 510–521. https://doi.org/10.1111/j.13651048 313x.2008.03453.x


1049 Spreitzer, R. J., & Salvucci, M. E. (2002). RUBISCO: Structure, regulatory interactions, and
1050 possibilities for a better enzyme. Annual Review of Plant Biology, 53(1), 449–475.
1051 https://doi.org/10.1146/annurev.arplant.53.100301.135233


1052 Sugawara, S., Ito, K., Miyazawa, S. I., Makino, A., & Suzuki, Y. (2026). Enzymatic and quantitative
1053 properties of Rubisco in C � herbaceous plants with early-spring persistent leaves and some alpine
1054 plants. Journal of Plant Research, 139(3), 511–518. https://doi.org/10.1007/s10265-026-01701-8


1055 Sun, J., Sui, X., Wang, S., et al. (2014). The response of rbcL, rbcS and rca genes in cucumber
1056 (Cucumis sativus L.) to growth and induction light intensity. Acta Physiologiae Plantarum, 36, 2779–
1057 2791. https://doi.org/10.1007/s11738-014-1648-z


35


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1058 Suzuki, Y., & Makino, A. (2012). Availability of Rubisco small subunit up-regulates the transcript
1059 levels of large subunit for stoichiometric assembly of its holoenzyme in rice. Plant Physiology,
1060 160(1), 533–540. https://doi.org/10.1104/pp.112.201459


1061 Taiz, L., Zeiger, E., Møller, I. M., & Murphy, A. (2018). Plant physiology and development. Sinauer
1062 Associates; Oxford University Press.


1063 Tamayo Tenorio, A., Gieteling, J., de Jong, G. A. H., Boom, R. M., & van der Goot, A. J. (2016).
1064 Recovery of protein from green leaves: Overview of crucial steps for utilisation. Food Chemistry, 203,
1065 402–408. https://doi.org/10.1016/j.foodchem.2016.02.092


1066 Thompson, D. S., & Islam, A. (2021). Plant cell wall hydration and plant physiology: An exploration
1067 of the consequences of direct effects of water deficit on the plant cell wall. Plants, 10(7), Article 1263.
1068 https://doi.org/10.3390/plants10071263


1069 Tilman, D., & Clark, M. (2014). Global diets link environmental sustainability and human health.
1070 Nature, 515(7528), 518–522. https://doi.org/10.1038/nature13959


1071 Torode, T. A., O’Neill, R., Marcus, S. E., Cornuault, V., Pose, S., Lauder, R. P., Kra č un, S. K.,
1072 Rydahl, M. G., Andersen, M. C. F., Willats, W. G. T., Braybrook, S. A., Townsend, B. J., Clausen, M.
1073 H., & Knox, J. P. (2018). Branched pectic galactan in phloem-sieve-element cell walls: Implications
1074 for cell mechanics. Plant Physiology, 176(2), 1547–1558. https://doi.org/10.1104/pp.17.01568


1075 Tryfona, T., Bourdon, M., Delgado Marques, R., Busse-Wicher, M., Vilaplana, F., Stott, K., & Dupree,
1076 P. (2023). Grass xylan structural variation suggests functional specialization and distinctive interaction
1077 with cellulose and lignin. The Plant Journal, 113(5), 1004–1020. https://doi.org/10.1111/tpj.16096


1078 Tsukimoto, K., Takada, R., Araki, Y., Suzuki, K., Karita, S., Wakagi, T., Shoun, H., Watanabe, T., &
1079 Fushinobu, S. (2010). Recognition of cellooligosaccharides by a family 28 carbohydrate1080 binding module. FEBS Letters, 584(6), 1205–1211. https://doi.org/10.1016/j.febslet.2010.02.027


1081 United Nations Department of Economic and Social Affairs, Population Division. (2022). World
1082 population prospects 2022: Summary of results. United Nations.
1083 https://www.un.org/development/desa/pd/sites/www.un.org.development.desa.pd/files/wpp2022_sum
1084 mary_of_results.pdf


1085 Valente, A. I., Ferreira, A. M., Almeida, M. R., Mohamadou, A., Freire, M. G., & Tavares, A. P. M.
1086 (2021). Efficient extraction of the Rubisco enzyme from spinach leaves using aqueous solutions of
1087 biocompatible ionic liquids. Sustainable Chemistry, 3(1), 1–18.
1088 https://doi.org/10.3390/suschem3010001


1089 van der Goot, A. J., Pelgrom, P. J. M., Berghout, J. A. M., Geerts, M. E. J., Jankowiak, L., Hardt, N.
1090 A., Keijer, J., Schutyser, M. A. I., Nikiforidis, C. V., & Boom, R. M. (2016). Concepts for further
1091 sustainable production of foods. Journal of Food Engineering, 168, 42–51.
1092 https://doi.org/10.1016/j.jfoodeng.2015.07.010


1093 Venditto, I., Luis, A. S., Rydahl, M., Schückel, J., Fernandes, V. O., Vidal-Melgosa, S., Bule, P.,
1094 Goyal, A., Pires, V. M. R., Dourado, C. G., Ferreira, L. M. A., Coutinho, P. M., Henrissat, B., Knox,
1095 J. P., Baslé, A., Najmudin, S., Gilbert, H. J., Willats, W. G. T., & Fontes, C. M. G. A. (2016).
1096 Complexity of the Ruminococcus flavefaciens cellulosome reflects an expansion in glycan
1097 recognition. Proceedings of the National Academy of Sciences, 113(26), 7136–7141.
1098 https://doi.org/10.1073/pnas.1601558113


36


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1099 Verhertbruggen, Y., Marcus, S. E., Haeger, A., Ordaz-Ortiz, J. J., & Knox, J. P. (2009). An extended
1100 set of monoclonal antibodies to pectic homogalacturonan. Carbohydrate Research, 344(14), 1858–
1101 1862. https://doi.org/10.1016/j.carres.2008.11.010


1102 Verhertbruggen, Y., Marcus, S. E., Haeger, A., Verhoef, R., Schols, H. A., McCleary, B. V., McKee,
1103 L., Gilbert, H. J., & Knox, J. P. (2009). Developmental complexity of arabinan polysaccharides and
1104 their processing in plant cell walls. The Plant Journal, 59(3), 413–425. https://doi.org/10.1111/j.13651105 313x.2009.03876.x


1106 Vidal-Melgosa, S., Pedersen, H. L., Schückel, J., Arnal, G., Dumon, C., Amby, D. B., Monrad, R. N.,
1107 Westereng, B., & Willats, W. G. T. (2015). A new versatile microarray-based method for high1108 throughput screening of carbohydrate-active enzymes. Journal of Biological Chemistry, 290(14),
1109 9020–9036. https://doi.org/10.1074/jbc.m114.630673


1110 Vitlin Gruber, A., & Feiz, L. (2018). Rubisco assembly in the chloroplast. Frontiers in Molecular
1111 Biosciences, 5, Article 24. https://doi.org/10.3389/fmolb.2018.00024


1112 Vogel, J. (2008). Unique aspects of the grass cell wall. Current Opinion in Plant Biology, 11(3), 301–
1113 307. https://doi.org/10.1016/j.pbi.2008.03.002


1114 Wang, H., Meng, F., Zhao, Y., Liu, Y., Chen, J., & Liu, X. (2025). Enzyme-assisted extraction of
1115 proteins from radish leaf and their structure-function relationship in stabilizing high internal phase
1116 emulsions compared to soy protein and whey protein. Food Hydrocolloids, 168, Article 111550.
1117 https://doi.org/10.1016/j.foodhyd.2025.111550


1118 Willats, W. G., McCartney, L., Mackie, W., & Knox, J. P. (2001). Pectin: Cell biology and prospects
1119 for functional analysis. Plant Molecular Biology, 47(1–2), 9–27.


1120 Willats, W. G. T., McCartney, L., Steele-King, C. G., Marcus, S. E., Mort, A. J., Huisman, M. M. H.,
1121 van Alebeek, G. J. W. M., Schols, H. A., Voragen, A. G. J., Le Goff, A., Bonnin, E., Thibault, J.-F., &
1122 Knox, J. P. (2004). A xylogalacturonan epitope is specifically associated with plant cell detachment.
1123 Planta, 218(4), 673–681. https://doi.org/10.1007/s00425-003-1147-8


1124 Willett, W., Rockström, J., Loken, B., Springmann, M., Lang, T., Vermeulen, S., Garnett, T., Tilman,
1125 D., DeClerck, F., Wood, A., Jonell, M., Clark, M., Gordon, L. J., Fanzo, J., Hawkes, C., Zurayk, R.,
1126 Rivera, J. A., De Vries, W., Sibanda, L. M., Afshin, A., . . . Murray, C. J. L. (2019). Food in the
1127 Anthropocene: The EAT–Lancet Commission on healthy diets from sustainable food systems. The
1128 Lancet, 393(10170), 447–492. https://doi.org/10.1016/S0140-6736(18)31788-4


1129 Winter, K., & Smith, J. A. C. (2022). CAM photosynthesis: The acid test. New Phytologist, 233(2),
1130 599–609. https://doi.org/10.1111/nph.17790


1131 Wostrikoff, K., & Stern, D. (2007). Rubisco large-subunit translation is autoregulated in response to
1132 its assembly state in tobacco chloroplasts. Proceedings of the National Academy of Sciences, 104(15),
1133 6466–6471. https://doi.org/10.1073/pnas.0610586104


1134 Yates, E. A., Valdor, J.-F., Haslam, S. M., Morris, H. R., Dell, A., Mackie, W., & Knox, J. P. (1996).
1135 Characterization of carbohydrate structural features recognized by anti-arabinogalactan-protein
1136 monoclonal antibodies. Glycobiology, 6(2), 131–139. https://doi.org/10.1093/glycob/6.2.131


1137 Yu, Y., Kleuter, M., Zhang, R., Trindade, L. M., Boom, R. M., & van der Goot, A. J. (2024).
1138 Understanding the protein extraction potential of tomato leaves (Solanum lycopersicum) through mass
1139 balance modelling. Innovative Food Science & Emerging Technologies, 96, Article 103790.
1140 https://doi.org/10.1016/j.ifset.2024.103790


1141 Zhang, C., Sanders, J. P. M., Xiao, T. T., & Bruins, M. E. (2015). How does alkali aid protein
1142 extraction in green tea leaf residue: A basis for integrated biorefinery of leaves. PLOS ONE, 10(7),
1143 Article e0133046. https://doi.org/10.1371/journal.pone.0133046


37


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1144 Zheng, B. B., Fang, Y. N., Pan, Z. Y., Sun, L., Deng, X. X., Grosser, J. W., & Guo, W. W. (2014).
1145 iTRAQ-based quantitative proteomics analysis revealed alterations of carbohydrate metabolism
1146 pathways and mitochondrial proteins in a male sterile cybrid pummelo. Journal of proteome research,
1147 13(6), 2998–3015. https://doi.org/10.1021/pr500126g


1148


1149


1150 **Figure 1. Overview of the experimental workflow used for parallel glycan and protein profiling**


1151 **of leaf biomass.**


1152 Leaf material was (1) collected and (2) freeze-dried and homogenised. (3) Glycans and proteins were


1153 extracted in parallel using the indicated extraction procedures. (4) The extracts were transferred to


1154 microplates for microarray printing, and (5) the printed arrays were probed with selected glycan- and


1155 protein-binding reagents. (6) Probe-binding signals were quantified and visualised for subsequent


1156 analysis. In parallel, (7) protein extracts were analysed for total soluble protein (TSP). This workflow


1157 enabled integrated comparison of cell-wall glycan epitope profiles, protein-probe signals and TSP


1158 concentrations across the biomass panel.


1159 **Figure 2. Protein antibody binding profiles across the biomass panel.**


1160 Rows represent biomass samples and columns represent functionally grouped protein antibodies;


1161 relative signal intensity (0–0.3) increases from blue through green to yellow. The antibodies targeted


1162 the Rubisco large subunit [RbcL (1), Abcam antibody; RbcL (2), Agrisera antibody], Rubisco small


1163 subunit (RbcS), Rubisco activase (RA), phosphoribulokinase (PRK), NADPH-dependent thioredoxin


1164 reductase C (NTRC), ribulose-phosphate 3-epimerase (RPE), LHCII type II chlorophyll a/b-binding


1165 protein (Lhcb2), CP29 chlorophyll a/b-binding protein of photosystem II (Lhcb4), oxygen-evolving


1166 enhancer protein 2 of photosystem II (PsbP), glycine decarboxylase P-protein (GLDP),


1167 phosphoenolpyruvate carboxylase (PEPC) and actin (ACT). Controls comprised a general negative


1168 control (Neg.) and a rabbit secondary-antibody control (Neg. Rabbit). Antibody specificities,


1169 dilutions, suppliers and references are provided in Supplementary table 2.


1170 **Figure 3. Total soluble protein concentrations across the plant biomass panel.**


38


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1171 Bars show mean total soluble protein concentrations from three replicates (n = 3), with error bars


1172 representing standard deviation. Samples are grouped by broad taxonomic group, indicated by colour


1173 and separated by dashed lines. TSP was quantified using the Bradford assay and is expressed as mg


1174 mL � ¹.


1175 **Figure 4. Relationships between Rubisco antibody signal intensity and total soluble protein**


1176 **across the biomass panel.**


1177 Scatterplots show mean total soluble protein (TSP) concentration against mean signal intensity for (A)


1178 RbcL (1), (B) RbcL (2) and (C) RbcS. Each point represents the mean of three biological replicates


1179 for one biomass sample. Solid lines show fitted linear regressions and shaded bands represent 95%


1180 confidence intervals. Pearson correlation coefficients ( _r_ ) and corresponding _p_ values are shown.


1181 Antibody designations: RbcL (1), Abcam anti-Rubisco large subunit; RbcL (2), Agrisera anti-Rubisco


1182 large subunit; RbcS, Agrisera anti-Rubisco small subunit. Further antibody details are provided in


1183 Supplementary table 2.


1184 **Figure 5. Glycan antibody binding profiles across the biomass panel.**


1185 Rows represent biomass samples and columns represent glycan-binding antibodies grouped by their


1186 principal targets; relative signal intensity (0–0.8) increases from black through pink to yellow.


1187 Controls comprised a general negative control (Neg.) and detection-specific negative controls for rat,


1188 mouse and polyhistidine antibodies (Neg. Rat, Neg. Mouse and Neg. poly-His, respectively).


1189 Antibody identities, epitope specificities, dilutions, suppliers and references are provided in


1190 Supplementary table 2.


1191 **Figure 6. Pearson correlation heatmap of protein and glycan antibody signals across the plant**


1192 **biomass panel.**


1193 Each cell represents the Pearson correlation coefficient ( _r_ ) calculated between mean protein and


1194 glycan antibody signal intensities across 128 plant samples. Red indicates positive correlations, with


1195 protein and glycan antibody signals tending to increase together; blue indicates negative correlations,


1196 with higher protein antibody signals associated with lower glycan antibody signals. Near-white cells


39


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1197 indicate little or no linear association. Protein and glycan antibodies were hierarchically clustered


1198 using Euclidean distance and complete linkage according to the similarity of their complete


1199 correlation profiles. RbcL (1) and RbcL (2) are Abcam and Agrisera antibodies against the Rubisco


1200 large subunit, respectively; RbcS is an Agrisera antibody against the Rubisco small subunit. Further


1201 antibody identities and specificities are provided in Supplementary table 2.


1202 **Figure 7. Principal component analysis of targeted protein-glycan and Rubisco antibody profiles**


1203 **across the plant biomass panel.**


1204 (A) Targeted PCA based on RbcL (1), RbcL (2), RbcS and six selected glycan antibodies (BS-400-4,


1205 LM1, LM11, LM15, LM19 and JIM13), points are coloured by growth form. (B) Rubisco-only PCA


1206 based on the same three Rubisco antibodies, points are coloured by growth form. The same Rubisco

1207 only scores are displayed by (C) crop status and (D) broad taxonomic group. Each point represents


1208 one plant sample (n = 128), based on the mean of three biological replicates; ellipses represent 68%


1209 within-group dispersion. In PCA analyses, positive PC1 scores indicate increasing combined Rubisco


1210 antibody signal, whereas PC2 represents the largest remaining source of variation independent of


1211 PC1. Percentages in the axis labels indicate the variance explained by each principal component.


1212 Samples positioned closer together have more similar antibody profiles, whereas greater separation


1213 indicates greater dissimilarity. Biomass identities and classifications are provided in Table 1.


1214 **Figure 8. Effects of enzyme treatment on total soluble protein yield and RbcL (2) signal in**


1215 **frozen-thawed plant biomass.**


1216 (A) Total soluble protein yield and (B) MAPP-derived RbcL (2) signal intensity in leaf and pulp


1217 fractions of cucumber ( _Cucumis sativus_ ), tomato ( _Solanum lycopersicum_ ) and perennial ryegrass


1218 ( _Lolium perenne_ ). Bars show means ± SD from three biological replicates (n = 3). Green and brown


1219 bars represent leaf and pulp fractions, respectively; dotted horizontal lines indicate the corresponding


1220 mean no-enzyme controls. Asterisks indicate Dunnett-adjusted differences from the matched species

1221 and matrix-specific no-enzyme control: _P_ < 0.05, **P** < 0.01 and _**P**_ < 0.001. RbcL (2) denotes the


40


[bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766128; this version posted November 13, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.11.12.766128)

(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-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/)


1222 Agrisera antibody targeting the Rubisco large subunit. Biomass weights and enzyme concentrations


1223 are provided in Supplementary table 3. DW, dry weight.


1224


1225

1226

1227


41