Cell layer-specific cell wall modification is associated with exo-mesocarp split in pistachio (Pistacia vera L.)

Abstract

Pistachio (Pistacia vera) is a drought- and salinity-tolerant perennial whose fruit features a fleshy exo-mesocarp, or ‘hull’, that protects the kernel. Hull development and degradation are key to kernel quality, yet the anatomy and mechanisms driving hull breakdown during late-stage development remain largely unknown. Here, we show that the hull contains anatomically distinct layers of hypodermal parenchyma and filler parenchyma. Using a combination of transcriptome analyses and immunohistochemistry, we show that changes in pectin-associated gene expression and modification of this polysaccharide are involved in hull cell size increase, loss of cell–cell adhesion, and hull softening. Anatomical analysis shows that filler parenchyma expands during late-stage hull development while hypodermal parenchyma remains constant in size. Field data suggest that irrigation and humidity affect pistachio hull split, implicating a role for water status in cell expansion. In summary, the complex interplay between molecular, cellular, and environmental changes suggests that cell layer-specific modifications of the cell wall are linked to exo-mesocarp splitting, forming a model for understanding the mechanism of fruit split during ripening in non-berry fruit crops.

Cell layer-specific modifications of the cell wall are associated with cell expansion and loss of cell–cell adhesion, leading to hull split during late-stage pistachio fruit development.

Introduction

Fruit cracking, or splitting of the pericarp during ripening, has been modeled in many species including tomato (Solanum lycopersicum), strawberry (Fragaria×ananassa), and grapes (Vitis spp.) (Domínguez et al., 2012; Correia et al., 2018; Chang et al., 2019; Michailidis et al., 2020; Chang and Keller, 2021; Hurtado and Knoche, 2023; La Spada et al., 2024). Fruit cracking during fruit growth and maturation can lead to additional issues such as insect damage, fungal infections, and cosmetic defects, which in turn reduce the marketable yield of the crop (Santos et al., 2023; La Spada et al., 2024). Fruit cracking has been linked to changes in the fruit cell wall that occur during fruit ripening, particularly to modifications in pectin, as well as changes in the fruit cuticle structure. One example of this is that suppression of polygalacturonase (PG), an enzyme involved in pectin degradation, and expansin, a class of proteins known for their role in cell wall loosening, can reduce epidermal cracking in tomatoes, resulting in firmer fruits with thicker cell walls and cuticle deposition (Cantu et al., 2008; Jiang et al., 2019). However, pericarp splitting of other non-berry fleshy fruits is less well studied.

Modification of the cell wall is critical for cell swelling, which is a source of physical tension driving fruit split (Cosgrove, 2000, 2024; Schumann and Knoche, 2020; Lahaye et al., 2021). Cell–cell separation and cell wall swelling can cause the exocarp cells to ‘unzip’, leading to cracking in the sweet cherry skin (Prunus avium), as described in the ‘zipper model’ (Brüggenwirth et al., 2014; Grimm et al., 2019; Schumann and Knoche, 2020). In addition to changes in the exocarp, swelling of the mesocarp from exposure to water after microcracks are formed on the epidermis is also implicated in fruit cracking (Grimm et al., 2019). However, the relationship between cell swelling, cell wall modification, and water status has not been consistently investigated in the context of fruit splitting in many fleshy fruit species.

The difference in stiffness between fruit skin and the fruit interior has also been implicated in fruit cracking. In grapes, epidermal and cuticle stiffness, water status, texture of the fruit, and programmed cell death, all contribute to maintenance of fruit integrity (Tilbrook and Tyerman, 2008; Chang et al., 2019; Chang and Keller, 2021). The current paradigm for fruits with thin exocarp and thick, fleshy mesocarp is that the softening of the fruit changes the physical strain placed on the skin, where the mechanical properties of the skin help determine the amount of strain the skin can withstand before splitting (Considine and Brown, 1981; Bargel et al., 2004; Knoche et al., 2004; Chang et al., 2019). In contrast, in non-berry fruits such as the pistachio, walnuts, and almonds, which have relatively thicker exocarp, thinner mesocarp, heavily lignified endocarps, and large seeds, less is known regarding the process of fruit cracking.

Pistachio (Pistacia vera) is a popular woody nut crop and emerging model species due to its nutritional benefits and its relatively high salt and drought tolerance (Ferguson et al., 2002; Bulló et al., 2015; Godfrey et al., 2019). Nonetheless, irregular weather, which results in abnormal temperature patterns and water availability, may affect the integrity of the pistachio exo-mesocarp. The fleshy exo-mesocarp, also known as the hull, is critical for protecting the edible kernel from pests and pathogens. During late-stage pistachio fruit maturation, the hull naturally detaches from the shell (the endocarp) and the shell splits along a dorsal–ventral suture while the hull remains intact (Polito and Pinney, 1999; Panahi and Khezri, 2011). Although the hull is a critical tissue for fruit integrity and consumer safety (Doster and Michailides, 1995, 1999; Doster et al., 2001), relatively little is known regarding the anatomy of its development under different environmental conditions.

Several distinct hull split phenotypes can occur that impact the economic value and safety of the pistachio. One phenotype is ‘early splitting’, in which both the hull and the shell ‘crack’ along the suture prior to harvest, leaving the kernel directly exposed to the environment (Sommer et al., 1986; Doster et al., 2001). The hull can also ‘crack’ adjacent to the suture, with the shell still providing some protection to the kernel, although the loss of hull integrity still increases kernel susceptibility to pests and pathogens (Doster et al., 2001). Another category of loss of hull integrity is ‘tattering’, where the outer layers of the hull become tattered and torn in appearance (Sommer et al., 1986) while the shell and kernel may or may not be exposed. Although the phenotype of hull split has been described in field conditions and is well known to growers, the cell types involved and mechanisms behind the different phenotypes of hull split are currently unknown. Hull softening occurs during late-stage pistachio fruit development (Adaskaveg et al., 2025), and water status has been implicated in pistachio hull cracking (Doster et al., 2001). It is unknown if these factors can induce cell swelling or if changes in hull texture are key mechanisms driving hull split.

In our study, we used transcriptome-level analysis, complemented with hull anatomy assessment and field data, to identify mechanisms that affect hull integrity in pistachio. Our study identified cell types involved in hull split and their corresponding cell wall modifications, along with candidate genes associated with hull ripening. We showed that the water status of the fruit and cell-type-specific cell expansion interact with cell wall modifications, which contribute to the different types of hull split.

Materials and methods

Plant material and sampling

In 2021, the fruits from trees of the P. vera ‘Kerman’ or ‘Golden Hills’ cultivars were sampled from a commercial orchard in Cantua Creek, CA, USA. ‘Kerman’ and ‘Golden Hills’ scions were grafted on ‘Pioneer Gold I’ (P. integerrima) rootstocks, with Kerman trees ∼20 years in age and ‘Golden Hills’ ∼15 years in age. During the flowering period, 6–12 trees per cultivar were tagged. To minimize the effect of fruit load and sun exposure on development and phenology, 4–6 branches that were spaced out throughout the canopy, each containing only one inflorescence per branch, were selected per tree and tagged for analysis. The end of the flowering period, where between 75% and 80% of the trees were post-flowering, was denoted as 0 days post-anthesis (dpa). During each time point sampled, one tagged cluster per tree from six trees was harvested and stored at 4 °C in a cooler, and transported back to the laboratory for processing within 24 h of harvest.

Similarly, in 2022, fruits of ‘Kerman’, ‘Golden Hills’, and ‘Lost Hills’ were tagged in a commercial orchard in Mendota, CA, USA. The scions were between 7 and 8 years in age and were grafted on ‘UCB-1’ (P. atlantica×P. integerrima) rootstock. In 2023, fruits of ‘Golden Hills’ were tagged in a commercial orchard located in Woodland, CA, USA. The scions were between 8 and 9 years in age and grafted on ‘UCB-1’ rootstock.

In 2024, fruits of ‘Golden Hills’ were sampled from commercial orchards located in two sites in Fresno, CA, USA for comparisons between irrigation treatments. Both were mature orchards grafted on ‘UCB-1' rootstock, using double-line drip irrigation. Trial A was managed organically, while Trial B was managed conventionally. At each location, the grower’s standard irrigation practices were used as the control (100% applied water), thus the amounts of applied water differ at each location. Irrigation treatments were implemented by altering the number of irrigation lines supplying water. For the 50% treatment, a valve was installed on one of the lines, and was shut off for most of the season. For the 150% treatment, an additional line was installed. To ensure that fertilizer applied through the irrigation system was not altered, the researchers turned off the additional line in the 150% treatment, and turned on the line in the 50% treatment during fertigation events (i.e. during fertilization events, each treatment had two lines supplying water and fertilizer). Each block was three rows wide, and only the central trees in the center row were monitored. This ensured that the trees that were being monitored were not picking up water from a neighboring row. As the treatments were implemented at the irrigation system riser, the number of trees in each row differed: at Trial A, there were 38 trees in each plot row (114 in each experimental plot). At Trial B, there were 28 trees per plot row (with a total of 84 in each experimental plot). Four blocks of each treatment were randomized at each site. Six fruit clusters of each irrigation treatment were collected at 140 dpa, near the time of commercial harvest.

The ID numbers of the trees sampled from USDA Pistachio Germplasm Collection located at Wolfskill Experimental Orchards (Winters, CA, USA) are provided in Supplementary Table S1. Fruits were collected and scored from the 2021–2024 growing seasons. Three to six fruit clusters throughout the tree were bagged and manually pollinated, and 2–6 clusters were collected during 147–154 dpa.

Sample fixation and sectioning

Whole fruits were imaged and excised into either 5 mm×5 mm segments or 5 mm rings in the transverse plane, imaged again in the transverse plane, and chemically fixed. The samples were vacuum infiltrated (Labcanco, model 5530000, USA) at a pressure of −90 kPa to −100 kPa for 1.5 h at room temperature in fixative solution containing 4% (w/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS) and 0.1% Tween-20. Fixed samples were washed for 10 min in double-deionized water prior to embedding. A vibratome (Leica Biosystems Nussloch GmbH, VT1000S, Germany) was used to section tissue to 100 μm as previously described (Zhang et al., 2021). For sectioning of both fresh and fixed samples, the samples were trimmed down to 3–5 mm around the region of interest and embedded in 5% (w/v) agarose in distilled and deionized water.

Hull firmness measurement

Hull firmness was measured on a texture analyzer (TA.XT Plus, Texture Technologies Corp. Hamilton, MA, USA) with a TA-52 probe (2 mm diameter) using the software Exponent (Version 6.1.16.0; Texture Technologies Corp.). Each fruit was split into two halves along the suture line so the samples would lie flat, and were placed on the platform of the texture analyzer. The probe was set at the speed of 1 mm s⁻¹ to penetrate the hull, shell, and kernel. The first peak, indicating the force necessary to penetrate the hull, was recorded as hull firmness as described by Adaskaveg et al. (2025).

Cellular staining

Samples were stained with undiluted Calcofluor White containing Evan’s blue (Sigma-Aldrich, #18909, USA) for 5 min at room temperature and rinsed three times with distilled and deionized water before imaging. Toluidine blue (Sigma-Aldrich, #T-3260, USA) staining was performed as described by Pradhan Mitra and Loque (2014). Briefly, sections were incubated in 0.02% (w/v, in distilled and deionized water) toluidine blue solution for 2–5 min at room temperature then rinsed three times with distilled and deionized water before imaging. Sudan IV (Matheson Coleman & Bell, #B 417, USA) staining was performed by incubating sections for 5 min in 0.09% Sudan IV in 1:1 ethanol:glycerol at room temperature as described by Ruzin (1999). Fluorol Yellow 088 staining was performed as described previously (Zhang et al., 2021). Briefly, the sections are incubated in 0.1 mg ml⁻¹ Fluorol Yellow 088 (Santa Cruz Biotechnology CAS-81-37-8, USA) in lactic acid at 70 °C for 1 h in the dark, and rinsed twice with distilled and deionized water before imaging mounted in 50% glycerol (v/v, in distilled and deionized water).

Immunofluorescence

Sections of fixed tissue (100 μm) were stained with antibodies based on the protocol modified from Gerttula and Groover (2017). Briefly, the sections were rinsed in distilled and deionized water for at least 15 min before washing three times with PBS–0.2% Tween-20 and blocking in blocking buffer [2% fish gelatin (Fisher Scientific, #50-259-35, USA), 1% BSA (Chem-Impex, #00040, USA), 0.2% Tween-20 in PBS] for 1 h at room temperature. Pectin staining was performed using JIM5 antibody (PlantProbes, University of Leeds, Leeds, UK) against low methylesterified homogalacturonan at 1:10 dilution in blocking buffer, and JIM7 antibody (PlantProbes) against highly methylesterified homogalacturonan at 1:8 dilution in blocking buffer. Staining with primary antibody was performed overnight at 4 °C in a humid chamber before washing three times with PBS–0.2% Tween-20 and blocking for 1 h at room temperature. The anti-rat Alexa Fluor™ 488 (Invitrogen, #ab150157, USA) was used as a secondary antibody for both JIM5 and JIM7 at 1:50 dilution in the blocking buffer. The secondary antibody staining was performed overnight at 4 °C in the dark before washing three times in PBS–0.2% Tween-20, mounting in the same PBS–Tween mixture, and imaging.

Microscopy

All bright field micrographs (toluidine blue, Evan’s blue, and Sudan IV) were captured using an Olympus BH-2 microscope with a Jenoptik D-07739 microscope capture camera and the ProgRes CapturePro 2.10.0.1 software (Jenoptik AG, Germany), and SPlan ×4, NA 0.13; ×10, NA 0.3; and ×40, NA 0.7 dry objectives. Fluorescence microscopy images were acquired using the confocal Laser Scanning Microscope 980 system (Carl Zeiss, Germany). A 408 nm laser (2% power) was used to capture the Calcofluor White signal and a 488 nm laser (8% power) was used for goat anti-rat secondary antibody, Alexa Fluor™ 488, for JIM5 and JIM7 immunofluorescence. Emissions were collected over a wavelength range of 410–488 nm for Calcofluor White and 494–582 nm for the Alexa Fluor 488 secondary antibody of JIM5 and JIM7. All fluorescence images were collected using either the Plan-Apochromat ×10, 0.45 NA M27 dry objective or the Plan-Apochromat ×20, 0.8 NA M27 dry objective. The pinhole was set to 1.88 AU for Calcofluor White, and 11.16 AU for JIM5 and JIM7. Zen 2011 SP3 (Zeiss) software was used for imaging and Zen Lite (ZEN 3.2) for image export, with JIM5 and JIM7 exported at a display threshold of 15 000. Autofluorescence micrographs were captured on three channels. The blue channel was excited with a 408 nm laser at 7% power, 10.59 AU pinhole, and emission collected from 410 nm to 483 nm. The green channel was excited with a 488 nm laser at 10% power, 4.4 AU pinhole, and emission collected from 494 nm to 582 nm. The red channel was excited with a 561 nm laser at 10% power, 4.72 AU pinhole, and emission collected from 573 nm to 719 nm.

Quantification of anatomical traits

Hull thickness and cell size were measured manually using the ‘measure’ function in FIJI (Fiji Is Just) Image J (1.53f51) (Schneider et al., 2012). Images and micrographs of fruits sectioned in the transverse plane, from the median of the fruit between the apex and base of the fruit, were used. The dorsal site is defined as the ∼2 mm region around the suture ridge on the side of the pistachio with the longer curvature, the ventral site is the ∼2 mm region around the suture ridge on the side of the pistachio with the shorter curvature, and the non-suture region is defined as the ∼4 mm region located 90° away from both dorsal and ventral sites. The polygon tool was used to select the cells in the micrograph for cell cross-section area measurements. The line tool was used to measure the thickness of the tissue of interest, by drawing a line at the start and end of the tissue layer perpendicular to the epidermis.

RNA sequencing

At 91 dpa, all the fruits had intact hulls. At 133 dpa the fruits were separated by hull phenotype into three categories: (i) cracked; (ii) intact; and (iii) tattered when harvested. Four fruits of each phenotype from each tree were pooled for each biological replicate, and four biological replicates, representing four trees, were analyzed per category at each time point. Due to the frequency of hull split events, only ‘Golden Hills’ was used for the tattered category, and only ‘Lost Hills’ for the cracked category.

‘Golden Hills’ and ‘Lost Hills’ fruits were harvested from the trees tagged in the 2022 commercial orchard between 11.00 h and 13.00 h at 91 and 133 dpa and transported on ice back to the laboratory, where the hull was removed and immediately snap-frozen in liquid nitrogen. Total RNA was extracted from hull tissue using the RNeasy Plant Mini Kit (Qiagen, #74904, USA) and sent to Novogene Corporation Inc. (Sacramento, CA, USA) for RNA-seq library preparation, sequencing, and bioinformatics analysis. The analysis utilized the initial version of the P. vera ‘Kerman’ genome assembly and annotation, and was then updated to incorporate the most recent gene annotations following publication of the ‘Kerman’ genome and annotation (Adaskaveg et al., 2025). RNA-seq quality statistical calculations, correlation plots, differential gene expression (DGE) analyses, Gene Ontology (GO) analyses, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed by Novogene. Briefly, DGE analysis was performed using DESeq2 (Love et al., 2014) with |log2(foldchange)| ≥1 and adjusted P-value ≤0.05 set as the threshold. GO and KEGG were both performed using the clusterProfiler (Yu et al., 2012). For GO and KEGG, GO terms and KEGG pathways with adjusted P-value <0.05 were considered to be significantly enriched. The top 30 GO terms were visualized as dot blots and the top 20 KEGG pathways were visualized as bar plots for each analysis. Weighted genome correlation network analysis (WGCNA) was performed in R using the WGCNA package version 1.73 (Langfelder and Horvath, 2008). The network was constructed with power set to 5, minimum module size of 100, and cut-off height of 0.75. Read counts of candidate genes in fragments per kilobase per million mapped reads (FPKM) identified in DGE were visualized using the ggplot2 package (Wickham, 2016) in R as heatmap.

Quantitative reverse transcription–PCR

Samples for quantitative reverse transcription–PCR (qRT–PCR) analyses were harvested and processed for RNA following the same protocol as described in the RNA-seq section above. cDNA synthesis was performed using the SuperScript III First-Strand Synthesis System (ThermoFisher Scientific, #18080051, USA) following the manufacturer’s protocol, with RNase H digestion as a final step to eliminate residual RNA. Primers were designed using Primer-BLAST (National Center for Biotechnology Innovation) using published P. vera sequences in the RefSeq mRNA database (National Center for Biotechnology Innovation) for genes of interest, with one primer bridging an exon–exon junction to avoid detection of residual genomic DNA. The primers used are listed in Supplementary Table S2. qRT–PCR was performed using Power SYBR Green PCR Master Mix (ThermoFisher Scientific, #4368577) utilizing the standard two-step amplification protocol for 10 µl reactions, following the manufacturer’s instructions, on the CFX96 Real-Time System with the C1000 Touch Thermal Cycler (Bio-Rad, #1845096, USA).

To confirm whether housekeeping genes commonly used as reference genes show stable expression across samples, we analyzed all the read counts of actin (ACT7), elongation factor 1α (EF1α), and glyceraldehyde phosphate dehydrogenase (GAPDH) gene families detected in RNA-seq, and found that while actin and GAPDH genes displayed stable expression during late stage development and hull split, EF1α showed significant variation between time points and between cracked and tattered hulls (Supplementary Fig. S1). Thus, qRT–PCRs were normalized using ACT7 (PvKer.03.g088750), chosen for its high read count and stable expression.

Cell adhesion test

Cell–cell adhesion of the pistachio hull was measured using the protocol modified from Segonne et al. (2014). Briefly, 5×5 mm of hull tissue was removed from the non-suture site at the median of the fruit, halfway between the tip and the base, submerged in 2 ml of deionized water, and vortexed (Fisher Scientific, Model G550) at maximum speed for 2 min. A 30 μl aliquot of the cell suspension was then mounted on a glass slide and the number of non-adherent cells released from the tissue was counted.

Weather data collection

The daily weather conditions for the 91–133 dpa period in 2022–2024 were collected from the California Irrigation Management Information System (CIMIS) weather station #139 (California Department of Water Resources, 2023), located at the site of the USDA Germplasm Collection in Wolfskill Experimental Orchard (Winters, CA, USA). Growing degree days (GDD) data for 2021 and 2022 commercial orchards used in this study are as published by Zhang et al. (2023).

Water exposure and solar radiation assays

Hull integrity after acute water exposure was tested by spraying the fruit clusters of seven trees from the USDA Pistachio Germplasm Collection at Wolfskill Experimental Orchard with deionized water at 91, 112, and 133 dpa during the 2024 growing season, to correspond to the time points for transcriptome analysis, with an additional time point added at 112 dpa to increase the frequency of water exposure. The fruits were sprayed until thoroughly wet before air drying. One cluster was sprayed per tree, while a cluster close-by was used as control. The clusters were harvested at 142 dpa and scored for their hull phenotype.

Hull integrity after chronic water exposure was tested by covering one fruit cluster per tree of eight ‘Golden Hills’ trees during the 2024 growing season in 10 non-dripping, wet paper towels and then enclosed in glassine paper bags (ULINE, Pleasant Prairie, WI, S-12472, 4.5 l) to mimic a prolonged high humidity condition. Unbagged clusters were used as control. The towels were moistened with 50 ml of deionized water at 2–4 biweekly time points at 91, 105, 119, and 133 dpa before harvest at 141 dpa and phenotypically analyzed for hull integrity.

Statistical analysis and figure assembly

RX64 version 4.0.3 (R Core Team, 2017) in R studio (RStudio, PBC), version 1.3.1093 was used for statistical analyses. The basic ANOVA, t-test functions, and the emmeans (version 1.6.3) and multcomp (version 1.4-17) packages were used for statistical analysis. Final graphs were generated using Microsoft Excel Office 365 (Microsoft, Redmond, WA, USA) with the exception of the heatmap, which was generated using the ggplot2 package (Wickham, 2016) in R.

Photographs and micrographs were formatted in GIMP version 2.10.38 graphics editor and FIJI (Fiji Is Just) Image J (1.53f51) (Schneider et al., 2012). All figures were assembled and edited in Inkscape version 1.0.1 (3bc2e813f5, 2020-09-07, open-source software).

Results

Two main categories of hull split in pistachio cultivars

To understand late-stage pistachio hull development and the mechanisms underlying hull split, we first characterized and categorized the different hull split phenotypes that occur during late-stage fruit development during the 2022 growing season. We selected ‘Kerman’ as the industry standard nut cultivar, with a recently published high-quality genome (Adaskaveg et al., 2025), and compared it with ‘Golden Hills’ (Kallsen et al., 2009) and ‘Lost Hills’ (Kallsen et al., 2014), two upcoming cultivars with earlier harvest dates and higher shell split rates. We set our comparison at 133 dpa at the end of August, coinciding with ‘Golden Hills’ commercial harvest (∼2000 GDD). The pistachio hull includes the fleshy exocarp and mesocarp (Supplementary Fig. S2). Our analysis revealed four main categories of hull phenotypes: (i) intact; (ii) cracked, where the exo-mesocarp was completely split along the interior–exterior axis so that the endocarp (shell) was exposed; (iii) tattered, with superficial split, largely at the exocarp layer, while the shell and kernel remained protected; and (iv) a combination of both cracking and tattering (Fig. 1A). The split occurred between 91 and 133 dpa, which was consistent in appearance between genotypes (Supplementary Fig. S3A, B). The frequency of occurrence for each category appeared cultivar specific, though ‘Golden Hills’ and ‘Lost Hills’ displayed comparable frequencies of fruits with combined tattering and cracking (Fig. 1B). In contrast to these cultivars, ‘Kerman’ showed a significant phenotypic difference with 100% hull integrity at 133 dpa, without any type of hull split (Fig. 1B, P<0.01 Fisher’s exact test). A second time point was analyzed at 142 dpa (∼2100 GDD), allowing further characterization of late-stage ‘Kerman’ hull development, since it has a later harvest date. The cultivars maintained the differences in hull split categories with the same trend (Fig. 1C, P<0.01, Fisher’s exact test).

Fig. 1.

Phenotype of pistachio hull split during late-stage fruit development. (A) Representative images of pistachio fruits demonstrating the different categories of hull integrity at 133 days post-anthesis (dpa). The site of hull cracking (black arrow) indicates a deep split, while hull tattering (red arrow) is superficial. Insets show representative areas of split. ‘Golden Hills’ fruits are used for all except cracked hull, which are ‘Lost Hills’. Scale bars=5 mm. (B) Frequency of occurrence for the different hull integrity categories observed in 2022 at 133 dpa (B) and 142 dpa (C). n=24–63 fruits per genotype per time point. P<0.01 for each time point. Fisher’s exact test.

Due to the late maturation of ‘Kerman’ and to account for the difference in orchard and weather between the years, we included a further sampling date at 154 dpa (∼3000 GDD) for ‘Kerman’ from a different growing season for a saturation point of maturity. ‘Kerman’ showed a hull split of 38% of fruits, with primarily tattered hulls at 154 dpa (Supplementary Fig. S3C).

The pistachio exocarp contains multilayered hypodermal and filler parenchyma tissue

Given the importance of the pistachio hull as the tissue layers that encapsulate and protect the endocarp and the seed (Doster and Michailides, 1995, 1999), we aimed to create a clear anatomical definition of the hull and its split categories. Thus, we examined the hull cell type composition and established its anatomical structure to better understand the cellular mechanisms behind each split type. Cell staining and anatomical analysis showed that the pistachio hull is composed of the exocarp and the mesocarp with many cell types (Fig. 2). The cuticle, epidermis, and hypodermal parenchyma (HP) form the skin, or exocarp, in an organization pattern similar to other fruits (Santos et al., 2023). We defined the mesocarp in pistachio to include all the cells between the exocarp and the endocarp. Hence, the hull consists of all the cell layers encompassing both the exocarp and the mesocarp, namely the exocarp HP and the mesocarp filler parenchyma (FP) (Fig. 2A). The HP in the exocarp tends to be smaller, more flattened, and can contain 3–7 cell layers (Fig. 2A-ii, iv, blue cell outline). The FP in the mesocarp, located interior to the HP, contains 18–26 cell layers that are larger and more isodiametric in the transverse plane (Fig. 2A-ii, iv, red cell outline). In between the HP and FP is the transition zone, containing 3–5 parenchyma cell layers with a phenotype that is intermediate between HP and FP (Fig. 2A). The transition zone tends to be the site of deposition for phenolics and pigments with high autofluorescence properties (Fig. 2A-iii). It is also the site where cell senescence initiates during late-stage hull development (Fig. 2A-iv). Thus anatomically defined, hull cracking involves the splitting of the entire exo-mesocarp, exposing the endocarp (Fig. 2B), whereas hull tattering corresponds to the split at the exocarp and the initiation of the separation of the HP from the FP along the transition zone (white dashed line, Fig. 2B), resulting in the separation of the exocarp from the mesocarp.

Fig. 2.

Anatomy of pistachio hull during late-stage development. (A) Diagram and transverse sections from the non-suture region (magenta box) depicting the anatomical features of the hull at 133 dpa (i). Scale bar=5 mm. Cell anatomy is visualized by toluidine blue staining (ii), scale bar=100 µm, and cell autofluorescence (iii), scale bar=50 µm. Cell senescence is visualized with Evans blue staining (iv), scale bar=50 µm. The exocarp (skin, cyan) contains the cuticle, epidermis, and hypodermal parenchyma, while the mesocarp (red) contains resin ducts, vascular bundles and their associated accessory tissues, as well as filler and border parenchyma. The transition zone is indicated in white, while the dashed line indicates one potential line of exocarp–mesocarp separation during hull tattering. (B) Schematic representation of the cracked and tattered hull events, with the transition zone indicated with a white dashed line. Exocarp tissue often initiates detachment from the mesocarp at the split site in tattered areas, before large areas of exocarp fully separate from the fruit.

RNA-seq identifies cell wall modification genes associated with pistachio late-stage hull development

To better understand the interaction between hull integrity and late-stage development, we performed RNA-seq analysis. We used ‘Golden Hills’ and ‘Lost Hills’ harvested in 2022, which showed cultivar-specific differences in hull split categories (Fig. 1). We selected 91 dpa, when all hulls were intact, and compared it against 133 dpa, when hull split had occurred, to examine differences in development between intact and split hulls. We focused the analysis of hull cracking on ‘Lost Hills’ and the analysis of hull tattering on ‘Golden Hills’ because they displayed a high frequency of these phenotypes (Fig. 1B, C). Statistical qualities for the RNA-seq are reported in Supplementary Table S3.

When we examined our results using WGCNA we were able to group genes into six main modules, with the largest number of genes associated with the blue and turquoise module (Supplementary Fig. S4). However, since we were interested in a multidimensional analysis featuring developmental age×genotype×phenotype, with very small matrix size for each dimension, we chose to focus on DGE analysis that was able to provide a clearer resolution. Pairwise DGE analysis between genotype, dpa, and phenotype showed that the greatest number of differentially expressed genes was detected between 91 dpa intact ‛Golden Hills' and 133 dpa of tattered ‘Golden Hills’ hull samples (6128 genes, Supplementary Fig. S5A, red arrow), while the smallest number of differentially expressed genes was detected between intact ‘Golden Hills’ and ‘Lost Hills’ hull at 91 dpa (699 genes, Supplementary Fig. S5A, black arrow). Fewer differentially expressed genes were found in analyses comparing 133 dpa intact hulls of ‘Golden Hills’ and ‘Lost Hills’ hull with their corresponding intact 91 dpa samples (4658 and 3179 genes, Supplementary Fig. S5A, cyan arrows). Pearson’s correlation analysis likewise showed that the genotypes are similar at 91 dpa (Supplementary Fig. S5B).

We then further examined the top 50 genes, sorted either by adjusted P-value or by fold change, from the pairwise DGE analysis across time points, genotypes, and phenotypes. The top 10 candidates are listed in Supplementary Tables S4–S6. Several candidate genes among the top 10 by P-value were differentially expressed across multiple analyses, including two known to be involved in cell wall remodeling (Fig. 3A). Snakin-1 (SNAK1, PvKer.10.g252770) showed increased expression from 91 to 133 dpa in intact ‘Golden Hills’ (log2-fold increase 6.03 adjusted P=3.21E-139) and from 91 to 133 dpa in tattered ‘Golden Hills’ (log2-fold increase 7.28, adjusted P=1.25E-275) (Fig. 3A). Expansin 4 (EXPA4, PvKer.03.g089030) showed a log2-fold increase of 4.1 between intact and tattered 133 dpa ‘Golden Hills’ (adjusted P=3.35E-66), and log2-fold increase of 3.51 from intact and cracked (adjusted P=1.21E-19) at 133 dpa ‘Lost Hills’ (Fig. 3A). It should be noted that many of the genes are ranked among top hits by adjusted P-value and may have relatively small fold change, making them difficult to visually detect in a heatmap (Fig. 3). Nonetheless, the difference is statistically significant, and hence they are still included in the heatmap.

Fig. 3.

Expression of top candidate genes for pistachio fruit during late-stage development. (A) Heatmap of the fragments per kilobase per million mapped reads (FPKM) of the top candidate genes identified in multiple differential gene expression analyses in RNA-seq. ‘Golden Hills’=G, ‘Lost Hills’=L, intact hull=int, tattered hull=tat, cracked hull=cra. CHIA (acidic endochitinase) is re-plotted on a separate scale to avoid swamping the signal of the other candidate genes. (B) Heatmap of the FPKM of additional genes of interest that showed differential expression for cell wall modification and hormonal responses. Each square represents the average of four biological replicates. Red boxes indicate genes with the largest fold change between samples that are known to be involved in cell expansion and cell wall remodeling. The gene abbreviations are as following: Suppressor of FRI (FRIGIDA) 4 (SUF4), Peroxidase 4 (PER4), Heat shock protein 11 (HSP11), Lipid transfer protein GPI-anchored 21 (LTPG21), Expansin 4 (EXPA4), Desiccation-related protein PCC13-62 (DRPE), Snakin-1 (SNAK1), Beta-D-xylosidase 1 (BXL1), Basic blue protein (BABL), Heparanase-like protein 3 (HPSE3), Probable disease resistance protein At4g27220 (DRL28), Dirigent protein 23 (DIR23), Pectin methylesterase inhibitor 4 (PMEI4), Probable pectin methyleserase/methylesterase inhibitor 7 (PME7), Pectin acetylesterase 3 (PAE3), Cotton Golgi-related 2 (CGR2), Pectinesterase inhibitor 10 (PMI10), Pectate lyase 5 (PL5), Ethylene responsive transcription factor ERF106 (EF106), Ethylene responsive transcription factor 4 (ERF4), Ethylene receptor 2 (ETR2), Reversion to ethylene sensitive 1 (RTE1), Forever young oxidoreductase (FEY), and Neomenthol dehydrogenase (SDR1).

Since all hulls were intact at 91 dpa, we reasoned that the fruit ripening processes may be associated with hull split. Therefore, we also examined the genes known to be implicated in fruit ripening, such as genes involved in the pectin component of the cell wall, ethylene biosynthesis, and additional genes in the expansin family in our DGE results (Brummell et al., 1999; Civello et al., 1999; Cosgrove, 2000; Santiago-Doménech et al., 2008; Forlani et al., 2019; Jiang et al., 2019; Wang et al., 2018, 2019). We observed a cultivar-specific up-regulation of expression from 91 to 133 dpa for Pectate Lyase 5 (PL5, PvKer.06.g159170, log2-fold change 5.37, adjusted P-value=5.97E-106), Pectin Methylesterase Inhibitor 4 (PMEI4, PvKer.04.g094420, log2-fold change 5.04, adjusted P-value =1.1E-49), and Ethylene responsive transcription Factor 106 [EF106, also known as ethylene response factor 106 (ERF106), PvKer.04.g093440, log2-fold change 4.65, adjusted P-value=1.27E-47] (Fig. 3B). Notably, the up-regulation is most significant when comparing 91 with 133 dpa intact ‘Golden Hills’. Expansin 8 (EXPA8, PvKer.03.g092200) also showed up-regulation in this analysis (log2-fold change=2.58, adjusted P-value=1.77E-12), ranking it among the top 50 genes by P-value in our analysis of cracked versus tattered fruit (log2-fold change=1.77 higher in tattered, adjusted P-value=1.29E-38) (Fig. 3B).

Next, we validated the expression of PL5, due to the well-established role of pectate lyase (PL) in ripening, SNAK1, which is associated with both plant defense and cell wall remodeling response (Almasia et al., 2008; Nahirñak et al., 2012), and EXPA8, as candidates to test for differences between cracked and tattered fruits. We used qRT–PCR to test ‘Golden Hills’ samples. Similar to RNA-seq, there was a significant up-regulation of PL5 at 133 dpa compared with 91 dpa, when the intact hull had higher PL5 expression than the tattered hull (Supplementary Fig. S6). SNAK1 had a threshold significance of P=0.055, with higher expression for the tattered fruits, corroborating the trend we observed in RNA-seq (Supplementary Fig. S6). However, changes in EXPA8 were below our qRT–PCR detection threshold (Supplementary Fig. S6).

Next, we used GO and KEGG analysis to identify pathways associated with hull phenotypic traits. There was a greater difference in hydrolases for O-glycosyl compounds and glycosyl bonds (GO:0004553, GO:0016798) when tattered hulls were compared against intact or cracked hulls at 133 dpa in GO (Supplementary Fig. S7A, C, E). Interestingly, the phenylpropanoid and flavonoid pathways showed the most pronounced differences when the intact hull was compared with all other phenotypes (Supplementary Fig. S7B, D), potentially due to stress caused by ruptured tissue. In contrast, plant hormonal and mitogen-activated protein kinase (MAPK) signaling pathways were the most altered when cracked hull was compared with tattered hull (Supplementary Fig. S7F), suggesting that cell cycle/senescence signaling may be different between cracked and tattered hulls.

Fruit softening is associated with hull tattering but not cracking

Since we detected significant differences in cell wall remodeling genes during late-stage development of pistachio hull, we next investigated the hull texture of fruits obtained during the 2022 growing season to determine if there was a cultivar-specific difference in the texture. We observed that hulls softened significantly from 91 to 133 dpa in all genotypes, and it was most dramatic in ‘Golden Hills’, with nearly a 5-fold decrease in penetration force from 91 to 133 dpa (Fig. 4A). To distinguish the impact of genotypes on the phenotype-specific differences, we separated the 133 dpa fruits by their hull split phenotypes. Cracked hulls were more similar in texture to intact hulls, while hulls that showed both tattering and cracking displayed a texture that is more similar to tattered hulls (Fig. 4B–D).

Fig. 4.

Changes in hull texture and cell–cell adhesion during hull late-stage development. (A) Changes in hull texture during ripening measured using the hull penetration force for ‘Golden Hills’ (GH), ‘Kerman’ (K), and ‘Lost Hills’ (LH). n=10–36 fruits from six trees. P<0.01 for time points and genotype, two-way ANOVA. (B–D) Textural differences between different hull phenotypes for cultivars GH (B), K (C), and LH (D). n=44–91 fruits per cultivar. One-way ANOVA. P<0.05. (E) Number of non-adherent cells released in the cell–cell adhesion assay during hull ripening stage in 2022. n=6–9 fruits per cultivar per time point. Two-way ANOVA. P<0.01 between time points; non-significant for cultivar and interaction. (F) Number of non-adherent cells released in the cell–cell adhesion assay for 133–142 dpa ‘Golden Hills’ and ‘Lost Hills’ fruits with different hull integrity, harvested in 2022. The P-value is non-significant for all. Two-way ANOVA. n=7–11 fruits. All samples are from the 2022 growing season. Different letters indicate least square means post-hoc analysis significant with α=0.05. For (B–D), analysis of texture by split phenotype: empty no fill bar=intact fruits; diagonal fill pattern=cracked fruits; diamond fill pattern=tattered fruits; and solid fill=fruits that show both cracking and tattering.

Fruit softening during ripening and pectin modification is often associated with a decrease in cell–cell adhesion (Harker and Hallett, 1992; Segonne et al., 2014; Lahaye et al., 2021; Su et al., 2024). Therefore, we measured the number of non-adherent cells, or cells that are separated from the hull tissue, using an established cell–cell adhesion assay (Segonne et al., 2014). A nearly 600-fold increase was detected in the number of non-adherent cells for all hull phenotype categories from 91 to 133 dpa when all phenotype categories were pooled (Fig. 4E). When 133–142 dpa fruits were separated by their genotype and phenotype, tattered ‘Golden Hills’ hull had a higher number of non-adherent cells, albeit this was not statistically significant (Fig. 4F).

Late-stage development and higher split rate is associated with increased pectin detection signal

Our data suggested that significant pectin remodeling occurred during late-stage hull development. To better understand the type of pectin modification involved in pistachio hull split, we investigated the pectin changes by immunofluorescence with pectin-specific antibodies on transverse sections of ‘Golden Hills’ and ‘Lost Hills’ fruits at 91 dpa, when all hulls were intact, and at 133 dpa, when hull split had occurred. Employing the monoclonal antibody, JIM5, which detects low methylesterified pectin (Knox et al., 1990; Clausen et al., 2003), we found a fluorescence signal increase from 91 to 133 dpa, especially at the cell wall junctions and in the transition zone between the HP and FP cell layers (Fig. 5A, white arrow). Similarly, an increase in the JIM7 immunofluorescence signal, which detects high methylesterified pectin, was observed for both genotypes (Fig. 5B) (Knox et al., 1990; Clausen et al., 2003). Quantification of fluorescence normalized against no primary antibody controls, which had minimal fluorescence signal (Supplementary Fig. S8), showed that while both JIM5 and JIM7 signal intensity increased from 91 to 133 dpa, cell type responses differed. For JIM5, the FP signal increased significantly (Supplementary Fig. S9; Fig. 5C). For JIM7, the HP fluorescence increased significantly, with JIM7 also showing a cultivar-specific difference of total hull signal at 133 dpa (Fig. 5C). When hull phenotypes were compared, JIM5 showed genotype-specific differences between ‘Golden Hills’ and ‘Lost Hills’ intact and cracked fruits (Supplementary Fig. S9A). JIM7 staining showed similar levels in intact fruits of both genotypes (Supplementary Fig. S9B).

Fig. 5.

Spatiotemporal analysis of pectin modification. (A) Hull sections stained with JIM5 for low- methylesterified pectin at 91 and 133 dpa. A white arrow indicates signal at the cell-cell junction. The diagram indicates the region of the hull sampled. (B) Imaging of hull sections stained with JIM7 for high methylesterified pectin. Scale bars=50 µm. (C) Quantification of average JIM5 and JIM7 fluorescence intensity for hypodermal parenchyma (HP), filler parenchyma (FP), and whole micrograph from samples harvested in 2022. All phenotype categories are included for (B) and (C). *P<0.05, **P<0.01, Two-way ANOVA. Different letters indicate least square means post-hoc analysis significant with α=0.05; no letter indicates non-significant in post-hoc analysis. n=6–10 sections quantified from fruits of 5–6 trees.

We further examined the association of hull pectin in different growing seasons and observed that during 2021, which was the growing season with a high percentage of fruits with intact hulls (Supplementary Fig. S3C), the average fluorescence detection intensity appeared unchanged from 91 to 133 dpa (Supplementary Fig. S9C, D). Thus, environmental differences between growing seasons could contribute to differences in pectin modification during late-stage hull development and could affect hull split rate.

Effects of expansion, hull thickness, and cuticle on hull split

Given that we identified expansin and pectin remodeling genes in our RNA-seq experiment and also detected changes in the levels of pectin (Figs 3, 5), we wanted to investigate whether cell and tissue expansion contribute to hull split phenotypes. We hypothesized that cell-type-specific changes in size may contribute to the internal expansion force driving hull split. To test this, we compared the cell cross-sectioned area of FP against HP at 91 and 131 dpa for ‘Golden Hills’ and ‘Lost Hills’ sampled from the 2022 growing season. The FP cell size nearly doubled from 91 to 133 dpa, while the HP size remained constant (Fig. 6). A similar trend was observed in 2021, the growing season with high integrity. During this high integrity season, FP increased in size, albeit this increase was not statistically significant, while HP remained constant in size (Supplementary Fig. S10). In summary, we found that there were cell-type-specific differences in cell size expansion during late-stage fruit development, where FP cell size in the mesocarp increased while HP in the exocarp did not.

Fig. 6.

Cell expansion during late-stage hull development. (A) Anatomy of hull parenchyma during late-stage hull development. Calcofluor-stained hypodermal parenchyma (HP) and filler parenchyma (FP) within the skin tissue of ‘Golden Hills’ and ‘Lost Hills’. Samples were collected at 91 and 133 dpa during the 2022 growing season. Scale bars=50 µm. Boxes indicate the region of the hull sectioned and analyzed. (B) Quantification of HP and FP cross-sectional area. n=30 cells/time point from a minimum of three fruits from three trees. Difference across time points is significant for FP; non-significant for genotype, P<0.01, two-way ANOVA. Letters indicate difference by least square means with α=0.05.

The current model of fruit split assumes that an internal expanding force drives the splitting of the fruit skin, led by the turgor pressure from water absorption by the fruit or changes in the water potential (Chang and Keller, 2021; Santos et al., 2023; La Spada et al., 2024). However, questions remain on the nature and contribution of the different interior expansion forces. To test whether pistachio shell split contributes to hull split as an expansion force, we examined the fruits of P. atlantica, which only produces fruits with indehiscent shells. The observation of P. atlantica fruits with a tattered hull suggests that cell expansion, not shell split, may play a more important role in pistachio hull tattering (Supplementary Fig. S11).

We reasoned that if FP cell expansion is the main internal expanding force driving hull split, this expansion force would create tension in the hull. In this case, a location with thinner hull tissue would lead to a mechanically weaker region that is more prone to splitting. To test this hypothesis, we quantified the hull thickness. We identified an interior notch located at the dorsal suture of the hull, resulting in a region that is nearly half as thick as the non-suture site hull (Fig. 7A–C). Next, we artificially induced freeze-swelling of cells to cause tension in the hull tissue by using liquid nitrogen to snap-freeze intact pistachio fruits. When intact 133 dpa fruit were snap-frozen, the majority of the fruit hull cracked at the dorsal suture, a phenotype similar to the cracked hull observed in the field (Fig. 7D). Hulls of younger, intact 105 dpa fruits did not crack as frequently when snap-frozen (Fig. 7D–E). Quantification of the location of hull split for tattered and cracked fruits showed that indeed the location of hull cracking was at the dorsal suture site (Fig. 7B, blue line, and 7F), the location with the thinnest hull (Fig. 7A–C), while hull tattering tend to occur at the non-suture site (Fig. 7B, white line, and 7F). In summary, the data suggest that a cell expansion force may be the main driver behind hull cracking.

Fig. 7.

Effect of location-specific hull thickness and simulated cell expansion on hull cracking. (A) Anatomy of the dorsal suture interior notch at the site of cracking initiation. Sudan IV-stained hull at the dorsal suture shows initiation of tissue split from the interior of the fruit. Scale bars=50 µm. (B) Representative image of 133 dpa ‘Golden Hills’ fruit cut at the median transverse section for measurement of hull thickness at the dorsal suture (blue), non-suture (white), and ventral suture site (green), showing locations in the hull where a split may occur. The lines indicate the thickness measured at each site. Scale bar=5mm. (C) Quantification of hull thickness at each site indicated in (B). n=5–18 fruits from 5–6 trees. Asterisk indicates P<0.05 for site, non-significant for genotype and interaction, two-way ANOVA, non-significant for least square means post-hoc analysis. (D) Representative images of hull cracking at the dorsal site in field samples and in 133 dpa ‘Golden Hills’ fruit with freezing-induced cell swelling. Similar freezing does not induce cracking at 105 dpa. Scale bars=5 mm. (E) Percentage of ‘Golden Hills’ fruits that show dorsal hull cracking after snap-freezing. n=21 fruits from 7–8 trees at each time point. (F) Location of the hull split in cracked and tattered fruits from field samples in 2022. n=37–44 fruits from six trees. P<0.01, Fisher’s exact test.

Finally, cuticle properties such as thickness and integrity have been previously associated with fruit split (Knoche et al., 2004; Domínguez et al., 2012; Jiang et al., 2019; Dimopoulos et al., 2020; Chang and Keller, 2021; Santos et al., 2023). We examined the cuticles of the different genotypes to test the importance of the cuticle in pistachio hull integrity. We did not detect any discernible difference between the cuticles of the cultivars in either thickness or physical appearances over the process of late-stage hull development, between genotypes, in growing seasons in 2021 and 2022 (Supplementary Fig. S12).

Water status has a significant effect on hull split rate

Environmental factors such as humidity affect fruit split (Correia et al., 2018; La Spada et al., 2024), and likely cell expansion. The difference in ‘Golden Hills’ hull phenotypes between 2021 and 2022 at 133 dpa, changing from 95% intact with 5% tattering in 2021 to 79% intact with 16% tattering in 2022 (Fig. 1; Supplementary Fig. S3C), suggests that environmental factors may contribute to hull split in pistachio.

To examine whether humidity or irrigation was an important driver behind differences in hull phenotype, we collected fruits from two commercial orchards in 2024, from trees under 50%, 100%, and 150% irrigation treatments. We found that indeed there was a correlation of hull split occurrence with irrigation, where increasing the irrigation from 100% to 150% doubled the frequency of both tattered and cracked hulls (Fig. 8A). In addition, under the same irrigation treatment, prolonged exposure of fruits to humidity significantly increased the hull split percentage, especially of cracked hulls (Fig. 8B). Since fruit cracking is often tested by directly soaking the fruit in water (Michailidis et al., 2020; Chang and Keller, 2021), we also tested the response of pistachio to acute (short-term) overexposure to moisture by spraying P. vera fruits from trees in the USDA Germplasm Collection at Wolfskill Experimental Orchard (Supplementary Table S1) with water. However, acute exposure to humidity had no obvious effect on the hull split rate (Fig. 8C).

Fig. 8.

Effect of irrigation and humidity exposure on hull split. (A) Hull phenotype of ‘Golden Hills’ fruits collected from 50, 100, and 150% irrigation treatments during the 2024 growing season. n=331–362 fruits from 12 trees per treatment. P<0.01, χ² test. (B) Hull phenotype of ‘Golden Hills’ fruits under no treatment (control) conditions and after prolonged exposure to moisture. P<0.01, n=128 for control, 172 for treated, eight trees per treatment. χ² test. (C) Hull phenotype of fruits from the USDA Germplasm Collection at Wolfskill Experimental Orchard under no treatment (control) conditions and after acute exposure to water. The P-value is non-significant, n=104 for control and 159 for seven treated trees. χ² test.

To separate out the effect of agricultural management conditions, tree age, rootstock type, and potential genetic variations on hull split, we assayed the hulls of a single tree from the USDA Pistachio Germplasm Collection located at Wolfskill Experimental Orchard from 2022 to 2024, during which period it was grown under the same orchard management regimen. We observed that the fruits of the D 7 24 tree showed 100% intact hulls when harvested between 147 and 154 dpa in 2022, but showed <50% intact hulls in 2023, and 92% intact hulls in 2024 when harvested at the same time point (Fig. 9A). This correlated with unseasonal rain during the summer of 2023, which altered the average precipitation, humidity, temperature, and solar radiation during the late fruit developmental stage (Fig. 9B–J). Radiation and relative humidity are known to affect the cuticle and cracking (Domínguez et al., 2012; La Spada et al., 2024), and we found that there was indeed a significant increase in dew point and vapor pressure in 2023, with dew point temperature increasing ∼2 °C and vapor pressure increasing by 0.3 kPa (Fig. 9B–D). Interestingly, there was a change in relative humidity, and the average relative humidity correlated the most strongly with the total hull split rate across the years (Fig. 9A, H–J).

Fig. 9.

Effect of environmental factors on hull split. (A) Hull phenotype of fruits harvested 147–154 dpa during the 2022–2024 growing seasons from a single P. vera tree, D 7 24, in the USDA Germplasm Collection at Wolfskill Experimental Orchard. P<0.01, χ² test. n=43–219 fruits. (B–J) Average of daily climate data from an onsite weather station at the Wolfskill collection site for 91–133 dpa in 2022–2024. (B) Precipitation, P-value is non-significant. (C) Dew point temperature, P<0.01. (D) Vapor pressure, P<0.01. (E) Average air temperature, P-value is non-significant. (F) Average soil temperature P<0.01. (G) Solar radiation, P<0.01. (H) Maximum relative humidity, P<0.01. (I) Minimum relative humidity, P<0.01. (J) Average relative humidity, P<0.01. n=41. One-way ANOVA. Letters indicate difference by least square means with α=0.05.

We next tested the effect of the environment on pectin modification and hull integrity by comparing the fluorescence intensity of high and low methylesterified pectin in cultivars we sampled from the 2021–2023 growing seasons (Fig. 10A). We observed that there was a strong correlation between the fluorescence detection intensity of low methylesterified pectin in the HP and the frequency of cracked fruits (columns marked in cyan, Fig. 10A, B). There was also a negative correlation between the ratio of HP to FP low methylesterified pectin signal and the percentage of intact fruit (Fig. 10C). The data suggest that the environment and the relative amount of substituted pectin in the cell wall, particularly in the HP, are associated with the form of hull split.

Fig. 10.

Pectin modification correlates with hull phenotype across growing seasons. (A) Summary of hull integrity, and high (JIM7) and low (JIM5) methylesterified pectin signal across different growing seasons in hypodermal parenchyma (HP) and filler parenchyma (FP). Data sets with strong correlation are in columns indicated/shaded in cyan. N=minimum of six fruits sampled from six trees for each immunofluorescence experiment each year. (B) Correlation of JIM5 signal in HP with percentage of cracked hull and (C) correlation of the HP:FP ratio of JIM5 fluorescence signal with the percentage of intact hull for data reported in (A).

Discussion

Although recent studies provide a developmental profile of gene expression for the pistachio kernel, shell, and hull during ripening (Adaskaveg et al., 2025), the mechanisms behind the exo-mesocarp split in pistachio remain largely unknown. This study provides an anatomical framework of pistachio hull split phenotypes and correlates both transcriptional and in muro changes of cell wall polysaccharides across genotypic, developmental, phenotypic, and environmental levels (Fig. 11A). Integration of the data shows the contrast between the expanding FP in the mesocarp and the non-expanding HP, and that environmental factors may affect cell wall modification and phenotype in a cell-type-specific manner (Fig. 11A).

Fig. 11.

Proposed model of the interacting factors associated with the phenotype and rate of pistachio hull split (A), demonstrating the factors affecting split at the morphological (B), cellular (C), and molecular level (D). (A) Factors (boxes) and interactions (lines) experimentally tested in this study are colored green, established interactions associated with fruit cracking and splitting in other species are colored yellow, while factors and interactions not directly tested are colored gray. Asterisks indicate assay results that show cell type specificity. (B) Proposed differences in hull mechanical properties in intact (i), cracked (ii), and tattered hull (iii) as affected by hull thickness and cell–cell adhesion. (C) Proposed model of the distribution of tension in intact (i), cracked (ii), and tattered (iii) hull as affected by cell-type-specific cell expansion and humidity. (D) Difference in pectin substitution pattern in intact versus split hull.

Pectin remodification occurs during pistachio hull ripening and is associated with cell expansion and hull softening

We have identified PL5 as one of the up-regulated genes during late-stage hull development (Fig. 3B). A decrease in cell–cell adhesion, increase in cell wall loosening, and increase in cell expansion has been previously associated with the PL protein family (Santiago-Doménech et al., 2008; Wang et al., 2019; Huang et al., 2023; Su et al., 2024), which supports the role of PL in hull breakdown. We observed a similar increase in PL expression at our last late-stage hull development time point (Fig. 3B; Supplementary Fig. S6), corroborating earlier studies (Blanco-Ulate et al., 2022), and a correlating decrease in cell–cell adhesion (Fig. 4E). PL5 expression may be higher in intact than in tattered ‘Golden Hills’ at 133 dpa due to the amount of damaged and senescing cells in the latter (Fig. 1B; Supplementary Fig. S3B), resulting in tissues that were past the peak of PL activity.

Our immunofluorescence results of low and high methylesterified pectin also indicate intense reorganization of the cell wall during late-stage pistachio hull development (Figs 5, 10). In addition, there was an association between the increase in low methylesterified pectin in HP and hull split, particularly of cracked hull (Fig. 10A). Multilevel cell wall remodeling is a well-established ripening process in many fruits such as sweet cherry, strawberry, and tomato, and is associated with fruit softening and fruit split (Cantu et al., 2008; Santiago-Doménech et al., 2008; García-Gago et al., 2009; Basanta et al., 2014; Uluisik et al., 2016; Yang et al., 2017; Jiang et al., 2019; Wang et al., 2019; Santos et al., 2023; Ortega-Salazar et al., 2024). Indeed, we observed increased expression of SNAK1, CHIA, PG, xyloglucan endotransglucosylase/hydrolase, beta-glucosidase, as well as PMEI4 in our DGE (Fig. 3B; Supplementary Tables S4–D6), indicating multilevel wall remodeling and signaling responses that may also overlap with pathogen/stress response expected during hull split. At the same time, PL and PG are known to generate oligogalacturonides, which can function in plant defense signaling (Decreux and Messiaen, 2005; Liu et al., 2024; Anderson and Pelloux, 2025). Hence, the complexity of cell wall modification that is occurring and the increased expression of both pectin degradation and biosynthetic genes may explain why we observed an increase in both low and high methylesterified pectin from 91 to 133 dpa (Fig. 5).

A notable finding of this study is that the ratio of low methylesterified pectin in HP versus FP is negatively correlated with hull split (Fig. 10). This can generate a change in cell wall mechanical properties between FP and HP that can lead to either tattering or cracking depending on the location and the number of regions of mechanical weakness (Fig. 11B, C). The relative levels of substituted pectin in different cell layers, in combination with loss of cell–cell adhesion, may create regions with different mechanical properties in the hull, leading to different categories of hull split (Fig. 11B–D).

Filler parenchyma cell expansion may contribute to pistachio hull split during late-stage development

Modification in cell walls is necessary for cell expansion (Cosgrove, 2000; Caffall and Mohnen, 2009; Braybrook and Jönsson, 2016; Bou Daher et al., 2018). The fact that snap-freezing only induced cracking at 133 dpa but not at 105 dpa suggests that both the changes in cell wall modification as a result of remodeling genes and the expansion of FP during late-stage fruit development are critical for induced hull cracking (Figs 3, 7; Supplementary Tables S4–S6).

There are two potential sources of internal expansion force in the hull during development: (i) the internal force from the shell towards the hull as the shell splits; and (ii) the FP cell expansion. The two may act independently or together to increase the tension on the skin. However, the presence of tattered hulls in P. atlantica, where the shell is indehiscent, led us to conclude that pressure from the shell is not driving hull tattering (Supplementary Fig. S11). Thus, we can form a working model of pistachio hull split based on the expansion of FP cells generating the expansion force in the hull (Fig. 6). The distribution of the tension from expanding cells in the FP layers may contribute to the specific phenotype of hull split. We hypothesize that in hulls that have high cell–cell adhesion, the expansion force from the FP is more evenly distributed throughout the hull (Fig. 11B-i). In this scenario, when the hull is experiencing the same amount of tension at all locations, it will tend to break at the region of mechanical weakness, or the thinnest point of the hull, at the dorsal suture (Figs 7C, 11C-ii). In contrast, if FP cell swelling is occurring in a hull that has low cell–cell adhesion, there will be multiple mechanical weak points throughout the hull (Fig. 11B-iii). These areas and the transition zone where cell senescence is taking place (Fig. 2A-iv) may have lower cell–cell adhesion. In this scenario, if the exocarp is stiff, non-expanding, and mechanically stronger relative to the mesocarp, the expansion force from the FP in the mesocarp may result in the separation of the exocarp from the mesocarp, resulting in the tattering phenotype (Fig. 11B-iii, C-iii).

The cuticle can contribute to the tension difference between the exocarp and mesocarp. The properties of the cuticle contribute to cracking in grape, tomato, strawberry, and sweet cherry (Domínguez et al., 2012; Nadakuduti et al., 2012; Dimopoulos et al., 2020; Hurtado and Knoche, 2023; Santos et al., 2023). However, no obvious difference in cuticle between time points and genotypes was observed in our study (Supplementary Fig. S12). This suggests that in fruits with thicker skin (Fig. 2A) such as pistachio, the difference in the relative amount of low methylesterified pectin in the HP and FP layers of split fruits (Fig. 10A–C) may provide the difference in elasticity between the skin and interior of the fruit, leading to hull split.

Increased irrigation and prolonged humidity promote hull split

Our data indicate that increased irrigation for the entire growing season and prolonged exposure of fruit to humidity can promote hull split (Fig. 8). The effect of humidity on fruit split is consistent with studies in tomato and other berries, for which rainfall and humidity are involved in fruit splitting and cracking (Cline et al., 1995; Domínguez et al., 2012; Correia et al., 2018; Haokip et al., 2020; Santos et al., 2023; La Spada et al., 2024). Environmental effects can also help explain why ‘Golden Hills’ showed high rates of hull integrity in 2021 (Supplementary Fig. S3), but a high rate of tattering in 2022 and 2023 growing seasons under normal irrigation (Figs 1B, 8B; Supplementary Fig. S3). This is further supported by the variation in phenotype of the tree D 7 24 in Wolfskill Experimental Orchard between growing seasons (Fig. 9A). However, maximum acute exposure to moisture, an assay commonly performed in grapes and sweet cherries (Michailidis et al., 2020; Chang and Keller, 2021), did not increase the split rate (Fig. 8C), while prolonged exposure did (Fig. 8B), suggesting that the length of time that the tension in the hull is maintained is more important than the intensity. Future work is needed to more directly assess the hull skin tension and osmotic potential of the hull tissue under different humidity and irrigation conditions.

In conclusion, in our study we have identified two main parenchyma cell types key to hull split. Cell wall modification, especially of pectin, occurs during late-stage pistachio hull development, and this modification is associated with hull softening and loss of cell–cell adhesion. We established the complex interactions between phenotype and environment that occur at the molecular and cellular level and identified candidate genes for cell wall modification and cell expansion. Furthermore, we showed that increased irrigation and prolonged exposure to humidity promote hull split, suggesting a possible link between water status and changes at the cellular level. Together, we have built a model where cell-type-specific expansion constitutes the main driving force behind the hull split, while the changes in cell wall modification and cell adhesion may determine the amount of cell expansion and the category of hull split.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. Heatmap of the RNA-seq read count of actin, EF1a, and GAPDH gene families commonly used as reference genes in qRT–PCR of plant tissues for all samples analyzed in RNA-seq.

Fig. S2. Organization of the pistachio fruit pericarp.

Fig. S3. Additional characterization of hull split phenotypes.

Fig. S4. Weighted gene co-expression network analysis (WGCNA).

Fig. S5. Differential gene expression analysis summarized using a bar graph and Pearson’s correlation plot.

Fig. S6. qRT–PCR of genes of interest identified in RNA-seq. qRT–PCR verification of EXPA8, PL5 (PL5), and SNAK1.

Fig. S7. RNA-seq enrichment analyses for intact versus tattered, intact versus cracked, and cracked versus tattered fruit samples.

Fig. S8. Negative control for antibodies and stains used.

Fig. S9. Immunofluorescence analyses for low methylesterified pectin (JIM5) and high methylesterified pectin (JIM7).

Fig. S10. Estimated cell size of hull hypodermal and filler parenchyma for fruits collected in 2021.

Fig. S11. Shell split rate in Pistacia atlantica.

Fig. S12. Cuticle thickness of pistachio cultivars during late-stage fruit development.

Table S1. USDA Germplasm ID from Wolfskill Experimental Orchard of trees used in this study in various assays.

Table S2. Primers used in qRT–PCR.

Table S3. Statistical analysis of the RNA-seq read quality.

Table S4. Top 10 genotype-specific candidate genes identified in DGE analyses comparing ‘Golden Hills’ and ‘Lost Hills’ at 91 and 133 dpa.

Table S5. Top 10 phenotype-specific candidate genes identified in DGE analyses comparing intact, cracked, and tattered hull at 133 dpa.

Table S6. Top 10 candidate genes for late-stage hull development identified in DGE analyses comparing 91 dpa hull against intact, cracked, and tattered hull at 133 dpa.

Author contributions

SZ, GD, and PG: conceived the research and designed the experiments; SZ, SE, AC, MW, AB, PG, KMB, PT, and RL: performed the wet lab experiments; CL, JAA, YW, and GM: performed the assembly and annotation of the reference genome; SZ and GDL; wrote the manuscript; SZ, JAA, BB-U, and GD: contributed to the review and editing. All authors discussed the results and have read and agreed to the published version of the manuscript.

Funding

The work is supported by the California Pistachio Research Board and the United States Department of Agriculture Hatch (CA-D-PLS-2132-H) to GD. SZ, JAA, and YW were partially supported by the Department of Plant Sciences Graduate Student Research fellowship funded by the McDonald Endowments and facilitated by Agriculture and Natural Resources from University of California, Davis. SZ is partially supported by the Henry Jastro Graduate Research Award, the Katherine Esau Summer Graduate Fellowship, and the Agriculture and Food Research Initiative Competitive Grants Program pre-doctoral fellowship project award no. 2023-67011-40395, from the U.S. Department of Agriculture’s National Institute of Food and Agriculture. This study was made possible in part through NIH grant S10OD026702 that allowed funding of the Zeiss 980 LSM and training and support by Thomas Wilkop.

Data availability

The RNA-seq data underlying this article are available in the Gene Expression Omnibus Database at https://www.ncbi.nlm.nih.gov/geo/and can be accessed with the accession number GSE300184.