How to build a fruit: Transcriptomics of a novel fruit type in the Brassiceae

Comparative gene expression studies are invaluable for predicting how existing genetic pathways may be modified or redeployed to produce novel and variable phenotypes. Fruits are ecologically important organs because of their impact on plant fitness and seed dispersal, modifications in which results in morphological variation across species. A novel fruit type in the Brassicaceae known as heteroarthrocarpy enables distinct dispersal methods in a single fruit through segmentation via a lateral joint and variable dehiscence at maturity. Given the close relationship to Arabidopsis, species that exhibit heteroarthrocarpy are powerful models to elucidate how differences in gene expression of a fruit patterning pathway may result in novel fruit types. Transcriptomes of distal, joint, and proximal regions from Erucaria erucarioides and Cakile lanceolata were analyzed to elucidate within fruit and between species differences in whole transcriptome, gene ontology, and fruit patterning expression profiles. Whole transcriptome expression profiles vary between fruit regions in patterns that are consistent with fruit anatomy. These transcriptomic variances do not correlate with changes in gene ontology, as they remain generally stable within and between both species. Upstream regulators in the fruit patterning pathway, FILAMENTOUS FLOWER and YABBY3, are expressed in the distal and proximal regions of E. erucarioides, but not in the joint, implicating alterations in the pathway in heteroarthrocarpic fruits. Downstream gene, INDEHISCENT, is significantly upregulated in the abscissing joint region of C. lanceolata, which suggests repurposing of valve margin genes for novel joint disarticulation in an otherwise indehiscent fruit. In summary, these data are consistent with modifications in fruit patterning genes producing heteroarthrocarpic fruits through different components of the pathway relative to other indehiscent, non-heteroarthrocarpic, species within the family. Our understanding of fruit development in Arabidopsis is now extended to atypical siliques within the Brassicaceae, facilitating future studies on seed shattering in important Brassicaceous crops and pernicious weeds.


Introduction
Studying gene expression patterns across plant structures and species can elucidate how their modification may produce morphological variation [1,2]. Fruits are diverse and ecologically a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 indehiscent L. apellianum [28,29]. Upregulation in upstream regulator AP2 has been suggested as a factor in this origin of indehiscence [29].
A notable morphological adaptation is the evolution of a complex fruit type known as heteroarthrocarpy, which is only found in some members of the tribe Brassiceae [21,30,31]. This modified silique is defined by the presence of a variably abscising central joint, an indehiscent distal region, and a variably dehiscent proximal region (Fig 2). As such, this novel morphology offers an opportunity to investigate fruit variation beyond shifts from dehiscent to indehiscent. Ancestral state reconstructions reveal that typical siliques are ancestral in the tribe with multiple origins to heteroarthrocarpy. However, reconstructions vary in the precise number of times this trait has evolved. In contrast, evolutionary patterns of dehiscence and joint articulation are less clear with closely related taxa exhibiting variation in these features [32]. Anatomically, heteroarthrocarpic fruits appear most like Arabidopsis siliques in their proximal regions, varying by a lack of a valve margin cell layer in indehiscent variants [32][33][34]. There are three described variations of heteroarthrocarpy: a non-abscising joint with a dehiscent proximal region, an abscising joint with an indehiscent proximal region, and an abscising joint with a dehiscent proximal region [34]. These subtypes have evolved multiple times, perhaps as a bet hedging strategy in response to selective pressure from hostile desert environments [6,32]. The evolution of the joint and subsequent heteroarthrocarpic subtypes may be developmental enablers that have facilitated changes in fruit morphology across the tribe, which would explain heteroarthrocarpy's evolutionary lability [34]. Regardless of lability, all types are linked by the mechanism in which seeds from the same fruit are released by different means. In other words, the joint is the novel and unifying feature of heteroarthrocarpy [34].
A comparison of expression patterns between heteroarthrocarpic subtypes is potentially informative for formulating hypotheses about its evolutionary origins. Erucaria erucarioides and Cakile lanceolata, hereafter referred to as Erucaria and Cakile, are two well-studied representatives for heteroarthrocarpy because of their close relation and divergent subtypes [6,7,32,34]. In all heteroarthrocarpic subtypes the distal region remains indehiscent. Erucaria represents the subtype where the proximal region dehisces at maturity, releasing seeds, and the distal region remains attached to the mother plant via the persistent replum (Fig 2). Cakile exhibits the subtype where the proximal region also remains indehiscent with joint abscission such that all seeds are dispersed as protected propagules (Fig 2). In previous studies it was hypothesized that the formation of the joint is the result of repositioning of the valve margin, such that the valve is only present in the proximal region of the fruit, unlike in Arabidopsis where it is found in the entire ovary [34]. In other words, the joint is the distal portion of the valve margin. This hypothesis was partially supported by comparative gene expression data of some, but not all, genes in the valve margin pathway using a candidate gene approach [7]. However, that study did not definitively determine how the pathway has been repositioned because it did not investigate upstream genes. Candidate gene approaches will, by design, overlook non-targeted genes, and a lack of in situ hybridization does not necessarily indicate a lack of expression. Further, the basis of the joint remains unknown. No study to date has investigated transcriptional variation of heteroarthrocarpic fruits sectioned transversely into distal, joint and proximal regions. This approach is complementary to prior research because it quantifies expression of all transcripts in discrete regions of a whole system. Expression profiles from these regions will elucidate broad patterns and potentially identify key players involved in the formation of heteroarthrocarpy. They will clarify unique and shared gene expression patterns between and within Erucaria and Cakile, and will set the groundwork for future research regarding the evolution of the joint. Herein, the objective is to uncover transcript patterns, unique or shared, between and within, two variant heteroarthrocarpic species. We expect gene expression to be consistent with anatomical features within fruits, and that expression of fruit patterning transcripts will be consistent with repositioning of the valve margin in heteroarthrocarpy.

Plant material
Seeds from Erucaria erucarioides (Coss. and Durieu) Müll. Berol and Cakile lanceolata (Willd.) O. E. Schulz were obtained from the late César Gómez-Campo's and KEW royal botanical garden's seed collections, respectively. Vouchers for Cakile and Erucaria have been deposited in the Vascular Plant Herbarium at the University of Alberta, and the Harvard University Herbaria, respectively. Seeds were germinated in 1% agar and transferred to clay pots containing a 2:1 soil (Sungro sunshine mix #4, Agawam, MA, USA) to perlite mixture. Plants were grown under a 16/8-hour light/dark schedule at 24˚C with scheduled watering in the University of Alberta, Department of Biological Sciences, growth chambers.
Distal, joint, and proximal regions from 10mm fruits (~10 days post fertilization) were collected and flash frozen in liquid nitrogen prior to storage at -80˚C. Distal and proximal regions were classified as all tissue~1mm above or below the joint, and the joint is remaining tissue between distal and proximal regions (Fig 2). The 10mm fruit size is roughly equivalent to Arabidopsis stage 17A fruits [35], which go through elongation and cell expansion before maturity. This size was chosen to capture late stage valve margin gene expression because the valve margin is easily distinguished at this stage, and an increase in lignification is observed in key layers, e.g., enb. [36].

RNA isolation and cDNA library preparation
RNA was extracted from frozen tissue using manual grinding and a Qiagen RNeasy micro kit (Hilden, Germany) with the following amendments to protocol: RNA was incubated in nuclease free water for five minutes prior to elution, and this eluate was spun through the same extraction column to maximize RNA yield. RNA concentration was verified using a Nanodrop ND-1000 spectrophotometer (Software version 3.1.2), and quality was confirmed using the Agilent 2100 bioanalyzer (Software version B.02.09.SI720). All cDNA samples were set at the same concentration of the most dilute RNA extraction. Samples were processed using the Illumina TruSeq stranded mRNA LT sample prep kit RS-122-2101 (California, U.S.), and the procedure was followed as described in the low sample protocol. The mRNA from each sample was isolated and purified using AMPure XP magnetic beads (Agencourt; Beverly, Massachusetts) before primary and secondary strand cDNA synthesis. Unique Illumina adapters were ligated, and each sample was PCR amplified before validation. PCR was run for 15 cycles of: 98˚C for 10 seconds, 60˚C for 30 seconds and 72˚C for 30 seconds followed by 5 minutes at 72˚C and a final hold at 4˚C. Samples were normalized, pooled, and sequenced by the center for applied genetics (TCAG) facilities of the Toronto Sick Kids hospital, Ontario, Canada.

De novo transcript assembly, differential expression, and annotation
Raw reads were trimmed and quality checked using Trim Galore! (Version 0. 4 . The raw reads are available at the Sequence Read Archive (SRA) database under the BioProject ID PRJNA545186. The Transcriptome Shotgun Assembly projects have been deposited at DDBJ/ EMBL/GenBank under the accessions GHNY00000000 and GHOR00000000 for Erucaria and Cakile, respectively. The versions described in this paper are the first versions, GHNY01000000 and GHOR01000000. Corset (Version 1.0.6) [39] was used to estimate contig abundance by grouping contigs into representative gene clusters as the first step of the differential expression analysis. Contigs are defined as continuous overlapping paired-end reads. Next, edgeR (Version 3.6.2) [40,41] was used to perform pairwise differential expression analysis of Trinity gene, Trinity contig, and Corset clusters between proximal, joint, and distal regions of fruits from the same species. Genes, contigs, and clusters were classified as significantly differentially expressed if log2(fold-change) was greater than 2 and the False Discovery Rate (FDR)-corrected p-value (α) was less than 0.05. The analyze_diff_expr.pl script, provided with Trinity, was used to generate z-score heatmaps of all significantly differentially expressed contig clustered transcripts (α < 0.05). A z-score is used to indicate how many standard deviations a value is above the mean. The transcriptomes were annotated using the Basic Local Alignment Search Tool (BLAST) [42] algorithm on a local copy of both the National Center for Biotechnology Information (NCBI) non-redundant protein (nr) database and The Arabidopsis Information Resource (TAIR) 10 database [43]. BLASTx (E-value<10 −10 ) was used to identify highly similar sequences, and transcripts with the highest bit-score from the TAIR database were used as representative transcripts for heatmap generation. Whole transcriptome and fruit patterning heatmaps were generated using ggplot2 [44] and ggplot in R, respectively (Version 3.4.2) [45]. These global gene expression patterns were compared to previously published in situ hybridization and semi-quantitative PCR of select fruit genes [7]. Assembly completeness was determined using Benchmarking Universal Single Copy Orthologs (BUSCO) (Version 2.0) [46] (S1 Table).

Orthologous clustering
Orthofinder (Version 1.1.8) [47] was used to group orthologous transcripts from unfiltered Erucaria and Cakile transcriptomes. These groupings (orthogroups) with transcripts from both species as well as top BLAST matches for fruit patterning genes of interest were used to generate heatmaps. For Venn diagram generation, high-throughput sequencing (HTS) [48] filtered transcripts, sorted by regions, were translated to longest open reading frame (ORF) protein fasta files using TransDecoder (Version 5.0.0) [49]. These files were uploaded for comparison using the Orthovenn webserver [50]. HTS filtering was used to reduce file size due to the web server upload limit, and to reduce the number of insubstantial transcripts.

Gene ontology
Transcriptome fasta files from Erucaria and Cakile were imported to BLAST2GO (Version 2.8) [51]. Annotation files were exported and filtered, using a merged profile of all three biological replicates, to generate gene ontology (GO) terms for each region and species. These GO terms were used to produce graphs containing transcriptome hits for chosen terms. Terms were chosen based on searches for lignin, abscission, dehiscence, specific hormone keywords, and top hits. For comparison between transcriptomes, the log2 of selected GO term counts were divided over the log2 of all GO term counts (log2(n)/log2(N)).

De novo assembly of Erucaria and Cakile transcriptome data
RNA-seq libraries were constructed from 9 total replicates of triplicate distal, proximal, and joint regions. RNA samples from segmented fruits of two distinct plants were combined before sequencing to achieve optimal yield for library preparation. Sequencing from both libraries averaged 27.41 and 29.41 million paired-reads for Erucaria and Cakile, respectively. After quality trimming read counts were reduced to 27.36 million and 28.36 million high quality reads, respectively. Inter-quartile ranges per base were minimally 33 for Erucaria for the first 5 base pairs, and minimally 32 in the 90 th percentile; Cakile's inter-quartile ranges were minimally 33 for the first 5 base pairs, and minimally 29 in the 90 th percentile.
The transcriptome from Erucaria had an average contig length of 942.83, and Cakile's had an average length of 877.15. The total transcript count for Erucaria and Cakile was 227,530 and 314,194 reads, respectively (Table 1). Corset cluster counts averaged 365,257 (Erucaria) and 436,177 (Cakile). Notably, the first replicate for Cakile had a read count of 269,732, which is minimally 130,000 fewer than replicate 2 and 3. This inconsistency may have caused some issues in downstream analyses, but overall, both transcriptomes were of adequate quality and read-depth. This is supported by a BUSCO analysis, as Erucaria and Cakile's assemblies had overall completeness of 96.4% and 94.8%, respectively (S1 Table).

Annotation of assembled transcripts
Both transcriptomes were compared to the nr and TAIR peptide database using a BLASTx algorithm, and all downstream analyses used the TAIR10 annotation for facilitated comparison to Arabidopsis. A total of 254,592 (Cakile) and 213,757 (Erucaria) transcripts with e- The majority of selected orthogroups were similar between and within species (lignin, abscission, and dehiscence processes, and hormone response) (Fig 3). Exceptions include: cell wall modification related to abscission, general abscission, and catabolic lignification. Cakile has a greater ratio of cell wall modification processes and a lower ratio of general abscission processes relative to Erucaria. Erucaria has a higher ratio of catabolic lignification processes in the joint region despite having similar ratios relative to Cakile in the distal and proximal regions (Fig 3). Overall, the GO analysis results are consistent between and within species.
Additional results from OrthoVenn showed minimal difference in orthologous clustering within species, but some differences between species (Fig 4). There are a greater number of shared clusters between the proximal and distal regions in Erucaria (2548) than Cakile (2306) despite Cakile having substantially more overall clusters than Erucaria (50,003 vs 32,757). Additionally, there are fewer clusters unique to the joint for Cakile (21) than Erucaria (112). In sum, there are fewer orthologous clusters in common within regions of Cakile fruits than within regions of Erucaria fruits.

Identification of differentially expressed transcripts in 10mm fruit
For whole transcriptome comparison, two heatmaps of significant pairwise differentially expressed transcripts (α = 0.01) were generated (Fig 5). Contig clustering was chosen for this analysis because it is a more conservative estimation of significant differential expression at the transcript level, i.e., there are a greater number of transcripts being compared with more stringent FDR correction relative to corset clustering. Values were then converted to z-score to facilitate interspecies comparison, and for visual clarity. Dendrograms highlight the differences in number of differentially expressed transcripts between both species, and show that all replicates clustered together, respectively. The joint and proximal regions of Erucaria are most alike in expression and are both dissimilar to the distal region ( Fig 5). All three regions in Cakile have different expression patterns, and the distal region has a relatively large inter-replicate variance (Fig 5). There are 15,345 (Erucaria) and 74 (Cakile) significantly differentially  difference in significant differential expression between the distal region relative to the joint and proximal region in Erucaria, and little significant variation between all three Cakile regions.
We compared expression profiles of 21 genes important for valve margin formation and positioning in Arabidopsis [2,11,16,[52][53][54][55][56][57][58][59][60][61][62][63] (Fig 6). Contig clustered transcripts were also chosen for this analysis based on matches against the TAIR database. Row dendrograms highlight the different clustering of expression patterns of VM genes between both species, e.g., ETT, RPL and BP cluster together in Erucaria but not in Cakile (Fig 6). Most fruit patterning genes for both species have no significant differences in expression across all regions, except for FIL and YAB3 which were significantly upregulated in the distal region relative to the joint in Erucaria, and IND which was significantly upregulated in the joint relative to both the distal and proximal regions in Cakile. Upstream regulators FIL and YAB3 are not expressed in late stage Cakile fruits, despite global expression in Erucaria fruits. Downstream regulator IND is expressed in the whole fruit in Erucaria, but only in the joint region of Cakile (Fig 6).

Gene ontology of heteroarthrocarpic fruits
Overall, GO terms within fruits and between species are similar (Fig 3 and S1 Fig), as expected, because all sections and replicates are from developing fruit with shared components (e.g., ovary wall, septum). Additionally, GO analyses of top terms do not usually vary between closely related species [64,65]. However, despite similarities in gene ontology, the origin of heteroarthrocarpy may still be explained by deviation in expression patterns of one or more of the valve margin pathway genes [18,[22][23][24][25][26]. Similarities in gene ontology do not imply similarity between all expressed transcripts, so variation of just a few transcripts may be the driving factor behind heteroarthrocarpy.

Global transcript expression of heteroarthrocarpic fruits are consistent with anatomy
Transcript expression patterns are consistent with anatomical variances within and between fruits. The distal region of Erucaria has opposing transcript expression relative to both its joint and proximal regions (Fig 5), i.e., when transcripts are upregulated distally in Erucaria they are downregulated proximally. This pattern is consistent with heteroarthrocarpic fruit anatomy, as distal regions contain no valve or valve margin, and proximal regions have both [34]. In contrast, all regions of Cakile have variable transcript expression, with the clearest distinction between the proximal and joint regions, i.e., when genes are upregulated proximally they will be downregulated in the joint ( Fig 5). As with Erucaria, expression profiles in Cakile vary in a manner consistent with anatomy. Superficially, one might expect the Cakile silique to have similar expression between all regions because the entire fruit is indehiscent, which is consistent with the pattern of significantly fewer DE genes between regions of Cakile (74) than Erucaria (15,345) ( Fig 5). However, anatomically, the distal region of Cakile is more similar to the distal region of Erucaria than to its own proximal region [34], and its abscising joint is anatomically reminiscent to a valve margin [34]. Thus, we would expect regions of the fruit to exhibit different expression patterns, which is supported by our data (Fig 5). Abscission zones are also found between septum and seeds, and they too share similar anatomy and expression to typical silique valve margins [66]. Heteroarthrocarpic distal regions are unlike indehiscent non-heteroarthrocarpic siliques such as L. appelianum, because heteroarthrocarpic distal regions have no remnant valve margin in contrast to indehiscence observed in Lepidium and the proximal region of Cakile [29,34]. Thus, we expect different expression patterns within heteroarthrocarpic fruits, as well as between heteroarthrocarpic and non-heteroarthrocarpic fruits. In summary, there is a clear difference between distal and proximal expression profiles for both Erucaria and Cakile, which is consistent with a repositioning of the valve margin, i.e., the distal region is quite distinct from the proximal region due to the lack of valve margin, or its remnant, in the distal region. This consistency is further explored by analysis of fruit patterning transcript expression involved in valve margin formation.

Fruit patterning genes
Despite the substantial differences in anatomy, most valve margin genes reveal similar expression patterns across fruits in both Erucaria and Cakile (Fig 6). Although overall these expression patterns are consistent with a previous investigation of some, but not all, members of the VM pathway [10], they vary with regards to expression of one gene in Erucaria. EeFUL1, one of two FUL homologs found in Erucaria, was previously shown to only be expressed in the proximal region in earlier stages of carpel development [7], but all FUL transcripts are expressed across all regions in this study of later stage development (Fig 6). This discrepancy may be due to dynamic gene expression at different stages or because our methodology cannot distinguish within region differences (e.g., genes expressed in valve but not replum), so differences within regions cannot be distinguished. In contrast to EeFUL1, our data are consistent with a previous publication which demonstrated that other fruit patterning genes have broader expression domains than observed in Arabidopsis [7]. EeALC and EeIND and ClALC were expressed in the septum of Erucaria and Cakile, respectively, which is found throughout all regions sampled in this study.
It is a compelling finding that upstream regulators FIL/YAB3 and JAG have variable expression across Erucaria (Fig 6). These three genes positively regulate expression of FUL and valve margin genes in Arabidopsis such that their cooperative function has been designated together as JAG/FIL activity [19]. Our data suggest a decoupling of this cooperation in heteroarthrocarpic fruits because these three genes do not exhibit the same expression patterns across Erucaria fruits (Fig 6). That is, no expression of JAG was detected in any region of either species at this stage. FIL and YAB3 showed different expression patterns across fruits of Erucaria, but neither were detected in Cakile. It is important to note that plants of double mutants' fil/yab3 in Arabidopsis have fruits that are remarkably reminiscent of heteroarthrocarpy, e.g., they lack valve margin in the distal region of fruit while maintaining ovary wall identity [8]. In contrast to heteroarthrocarpy, these mutants have ectopic valve margin in the proximal region of their fruits [8]. As these genes exhibit different patterns across Cakile and Erucaria and are expressed in both proximal and distal regions of Erucaria, heteroarthrocarpy cannot be explained by a simple lack of expression of these key regulators. Further, FIL/YAB are absent in the joint region of Erucaria (Fig 6), which is confounding since the joint contains small portions of both proximal and distal regions, an unavoidable consequence of segmentation during tissue collection. Nonetheless, deviation in expression patterns of these upstream regulators between Arabidopsis and heteroarthrocarpic fruits implicates variation in their expression profiles in the origin of heteroarthrocarpy.
When exploring heteroarthrocarpy, we need to consider fruit patterning beyond the basalapical differences that distinguish distal, joint, and proximal regions. That is, the medial (replum) patterning (Fig 1) is maintained in heteroarthrocarpic fruits whereas the lateral is not. This pattern is due to differences in dehiscence between proximal and distal segments: undifferentiated ovary wall is present in the distal region whereas valve or remnant valve is present in proximal region. In other words, replum tissue is present in distal, joint, and proximal regions of heteroarthrocarpic fruits regardless of whether the ovary wall has differentiated into valve and it is appropriately sized [36]. FIL/YAB3 and JAG function antagonistically with replum promoting gene, WUSCHEL RELATED HOMEOBOX 13 (WOX13), which positively regulates RPL in turn. This interaction is necessary for proper medial-lateral formation of Arabidopsis fruits. Further, ASYMMETRIC LEAVES1 (AS1) and AS2 collaborate with JAG/ FIL function as promoters of lateral factors [9]. The loss of both AS1/2 and JAG/FIL in Arabidopsis results in dramatic medial-lateral differences and substantially enlarged replum, which is interestingly more pronounced in the basal portion of the fruit [9,67]. As AS1/2 and AS1 are expressed throughout Cakile and Erucaria regions, respectively, this pattern suggests that AS1 alone is sufficient for proper replum (aka medial-lateral) formation in heteroarthrocarpic fruits. In other words, the collaboration between JAG/FIL function and AS1/2 is not maintained in heteroarthrocarpic fruits. Further, JAG/FIL activity is non-detectable in the entire fruit, at least in Cakile at later stages of development. Thus, it appears that some redundancy in lateral-medial patterning of Arabidopsis fruits has been lost in heteroarthrocarpic fruits, as supported by the different clustering of VM genes for each species (Fig 6), while simultaneously gaining apical-basal differences, e.g., dehiscence and indehiscence in the proximal and distal regions of Erucaria.

Valve margin pathway recruitment for abscission of the Cakile joint
The fruit of Cakile is distinct in that the joint abscises (disarticulates) at maturity. The joint, which represents the distal portion of the valve margin, thus represents a novel abscission zone in Cakile, completely separating the distal portion of the fruit. This is an unusual feature of certain heteroarthrocarpic subtypes, as there is no equivalent abscission zone in Arabidopsis. Our data strongly implicate the recruitment of downstream valve margin genes as responsible for joint abscission, although how that zone is positioned remains elusive. IND is significantly upregulated in joint region (Fig 6) and is primarily responsible for formation of separation and lignification layers in typical siliques [22,24], a juxtaposition of cell types also observed in the abscising joint region. Its presence in the joint may be due to a co-option of downstream valve margin pathway genes to facilitate formation of the joint abscission zone. Similar cooption is observed in seed abscission zones, although these zones typically involve SEEDSTICK (STK) in lieu of SHP, and the functionally similar transcription factor HEC3 in lieu of IND [66]. SHP1/2 and ALC expression are both consistent with this co-option, as they are expressed in all three regions (Fig 6). Additionally, SPT expression is consistent with expression of IND, as expected from its downstream role in valve margin formation (Fig 6) [11]. Further, both representative transcripts are among the 21 unique orthologous clusters in the joint of Cakile (Fig 4). This pattern is consistent with in situ hybridization data that showed SHP2 expressed in septum and ovules of Cakile, and in ovules of Erucaria [7]. Thus, the likely function of SHP1/2 and ALC in the joint region would be to promote expression of IND (SHP1/2), and the formation of the separation layer (ALC). What is unusual about joint abscission is that for the joint to separate, the distal and proximal regions of the replum must also separate. This expression pattern then implies that the mechanism used to physically separate valve from replum may also be in play for replum in the joint region. Taken together with anatomical studies, our data strongly suggests that there is a repurposing of the valve margin pathway in an otherwise indehiscent Cakile fruit, and that this pathway may be capable of initializing disarticulation in multiple tissue types.

Conclusion
Transcriptomic expression from late stage Erucaria and Cakile fruits is consistent with some conservation and some deviation of the valve margin pathway, specifically in upstream regulation, e.g., FIL/YAB3 and JAG. Thus, different upstream regulators are implicated in the loss of dehiscence in Brassiceae relative to Lepidium, where AP2 is likely responsible [29]. Loss of expression of FIL/YAB3 and JAG in Arabidopsis results in differing apical and basal phenotypes, which may help to explain the apical/basal differences in heteroarthrocarpic fruits [8]. Further, heteroarthrocarpic fruits likely recruit the same mechanism used in valve and seed abscission for joint abscission (Fig 6). Functional tests are necessary to confirm whether redeployment of FIL/YAB3, IND, and possibly SPT have key roles in the origin of heteroarthrocarpy as well as joint abscission.
There have been multiple whole genome duplications in the Brassicales, which has resulted in many polyploids within the Brassicaceae family [68][69][70]. We considered the possibility of transcriptional differences between gene copies in distal, joint, and proximal regions that were undetected because we were unable to determine copy number in our transcriptome. For example, there are four copies of FUL in the Brassiceae [71], but each potential FUL copy had multiple hits from the same transcripts in both transcriptomes, so there is no definitive answer about copy number and expression (Fig 6). That is, we could not confirm or refute subfunctionalization of some fruit patterning genes as having a role in the origin of heteroarthrocarpy. An analysis of multiple transcripts for every fruit patterning gene showed generally similar expression for each, but further analyses are needed to determine if neo/subfunctionalization plays a role in heteroarthrocarpy.
Understanding the nature of heteroarthrocarpy, and how it relates to fruit development in Arabidopsis, will facilitate future studies on seed shattering in important Brassicaceous crops, and pernicious heteroarthrocarpic weeds. Further, these studies inform on the origin of important variation in seed packaging and dispersal capabilities.