Genetic Architecture and Molecular Networks Underlying ...endemic to coastalregions of the Atacama...

Genetic Architecture and Molecular Networks UnderlyingLeaf Thickness in Desert-Adapted TomatoSolanum pennellii1[OPEN]

Viktoriya Coneva,a,3 Margaret H. Frank,a Maria A. de Luis Balaguer ,b Mao Li,a Rosangela Sozzani,b andDaniel H. Chitwooda,2,3

aDonald Danforth Plant Science Center, St. Louis, Missouri 63132bDepartment of Plant and Microbial Biology, North Carolina State University, Raleigh, North Carolina 27695

ORCID IDs: 0000-0002-0640-5135 (V.C.); 0000-0002-5964-1764 (R.S.); 0000-0003-4875-1447 (D.H.C.).

Thicker leaves allow plants to grow in water-limited conditions. However, our understanding of the genetic underpinnings ofthis highly functional leaf shape trait is poor. We used a custom-built confocal profilometer to directly measure leaf thickness in aset of introgression lines (ILs) derived from the desert tomato Solanum pennellii and identified quantitative trait loci. We reportevidence of a complex genetic architecture of this trait and roles for both genetic and environmental factors. Several ILs withthick leaves have dramatically elongated palisade mesophyll cells and, in some cases, increased leaf ploidy. We characterized thethick IL2-5 and IL4-3 in detail and found increased mesophyll cell size and leaf ploidy levels, suggesting that endoreduplicationunderpins leaf thickness in tomato. Next, we queried the transcriptomes and inferred dynamic Bayesian networks of geneexpression across early leaf ontogeny in these lines to compare the molecular networks that pattern leaf thickness. We show thatthick ILs share S. pennellii-like expression profiles for putative regulators of cell shape and meristem determinacy as well as ageneral signature of cell cycle-related gene expression. However, our network data suggest that leaf thickness in these two linesis patterned at least partially by distinct mechanisms. Consistent with this hypothesis, double homozygote lines combiningintrogression segments from these two ILs show additive phenotypes, including thick leaves, higher ploidy levels, and largerpalisade mesophyll cells. Collectively, these data establish a framework of genetic, anatomical, and molecular mechanisms thatpattern leaf thickness in desert-adapted tomato.

Leaves are the primary photosynthetic organs of landplants. Quantitative leaf traits have important connec-tions to their physiological functions and, ultimately, towhole-plant productivity and survival. While few as-pects of leaf morphology have been determined un-ambiguously as functional (Nicotra et al., 2011), clearassociations between leaf traits and variations in cli-mate have been drawn (Wright et al., 2004). Leafthickness, the distance between the upper (adaxial)and lower (abaxial) leaf surfaces, has been shown to

correlate with environmental variables such as wateravailability, temperature, and light quantity. Thus, ona global scale, across habitats and land plant diversity,plants adapted to arid environments tend to havethicker leaves (Wright et al., 2004; Poorter et al., 2009).

Leaf thickness is a continuous, rather than a categorical,trait. Thus, it is important to distinguish between thick-ness in the context of typical leaf morphology, generallypossessing clear dorsiventrality (adaxial/abaxial flatten-ing), in comparisonwith extremely thick leaves, describedas succulent, which are often more radial. While the def-inition of succulence is ecophysiological rather thanmorphological (Ogburn and Edwards, 2010), at the cel-lular level it is broadly associated with increased cell sizeand relative vacuole volume (Gibson, 1982; von Willert,1992). These cellular traits promote the capacity to storewater and to survive in dry environments (Becker, 2007).Allometric studies across land plant families have shownthat leaf thickness scales specifically with the size of pal-isade mesophyll cells: the adaxial layer of photosyntheticcells in leaves (Garnier and Laurent, 1994; Roderick et al.,1999; Sack and Frole, 2006; John et al., 2013). Increasedpalisade cell height leads to an increased area of contactwith the intercellular space and, thereby, to improveduptake of carbon dioxide (CO2) into mesophyll cells(Oguchi et al., 2005; Terashima et al., 2011), possiblyoffsetting the increased CO2 diffusion path in thicker

1 This work was supported by funds from the Donald DanforthPlant Science Center. R.S. is supported by an NSF CAREER grant(MCB-1453130). M.H.F. is supported by an NSF NPGI postdoctoralfellowship (IOS-1523668).

2 Current address: Independent Researcher, Santa Rosa, CA 95409.3 Address correspondence to [email protected] or

[email protected] author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policy de-scribed in the Instructions for Authors (www.plantphysiol.org) is:Daniel H. Chitwood ([email protected]).

V.C., M.H.F., and D.H.C. designed the research; V.C., M.H.F.,M.A.d.L.B., R.S., and M.L. conducted the experiments and analyzedthe data; V.C. wrote the article with contributions from the otherauthors.

[OPEN] Articles can be viewed without a subscription.www.plantphysiol.org/cgi/doi/10.1104/pp.17.00790

376 Plant Physiology�, September 2017, Vol. 175, pp. 376–391, www.plantphysiol.org � 2017 American Society of Plant Biologists. All Rights Reserved.

https://plantphysiol.orgDownloaded on December 13, 2020. - Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved.

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0003-4875-1447

http://orcid.org/0000-0002-0640-5135

http://orcid.org/0000-0002-5964-1764

http://orcid.org/0000-0003-4875-1447

http://crossmark.crossref.org/dialog/?doi=10.1104/pp.17.00790&domain=pdf&date_stamp=2017-08-24

mailto:[email protected]


http://www.plantphysiol.org


http://www.plantphysiol.org/cgi/doi/10.1104/pp.17.00790

https://plantphysiol.org

leaves. At the organismal level, thicker leaves presenta tradeoff between rapid growth versus drought andheat tolerance (Smith et al., 1997). This idea is sup-ported by global correlations between leaf mass perarea (LMA), a proxy for leaf thickness, and habits as-sociated with slower growth (Poorter et al., 2009).Although leaf thickness is a highly functional trait,

mechanistic understanding of how it is patterned dur-ing leaf ontogeny is poor. The main cellular events thatunderpin leaf development are the establishment ofadaxial/abaxial polarity, followed by cell division, di-rectional expansion, and differentiation (Efroni et al.,2008). Changes in the relative timing (heterochrony)and duration of these events can impact leaf morphol-ogy, including thickness. Several mutants have beenidentified that show clear alterations in leaf thickness.These include the Arabidopsis (Arabidopsis thaliana)angustifolia and rotundifolia3 (Tsuge et al., 1996) as wellas argonaute1, phantastica, and phabulosa (Bohmert et al.,1998), which have aberrations in the polarity of cellelongation and the establishment of adaxial/abaxialpolarity, respectively, as well as Nicotiana sylvestris fatand lam-1 (McHale, 1992, 1993), which affect the extentof periclinal cell division in leaves. However, thesedevelopmental mutants do not necessarily inform us ofthe mechanisms by which natural selection acts topattern quantitative variation in leaf thickness.Efforts to understand the genetic basis of leaf thickness in

the context of natural variation face several importantchallenges. First, direct measurement of leaf thickness at ascale that would allow the investigation of quantitativetrait loci (QTL) for the trait is not trivial. Because of thedifficulty inmeasuring leaf thickness directly, LMA is usedmost often as a proxy for this trait (Poorter et al., 2009;Muiret al., 2014). Second, in addition to genetic components, leafthickness is environmentally plastic: it is responsive to boththe quantity and quality of light (Pieruschka and Poorter,2012). Finally, because leaf thickness varies on a continuousspectrum and is not associated with any particular phylo-genetic lineage or growth habit, mechanistic questionsregarding its patterning need to be addressed in a taxon-specific manner.With these considerations in mind, we used two mem-

bers of the tomato clade (Solanum sect.Lycopersicon) that areclosely related, morphologically distinct, and occupydistinct environments (Nakazato et al., 2010) to studythe genetic basis and developmental patterning of leafthickness. The domesticated tomato species Solanumlycopersicum inhabits a relatively wide geographic rangecharacterized bywarm,wet conditionswith little seasonalvariation. By contrast, the wild species Solanum pennellii isendemic to coastal regions of the Atacama Desert of Peru,a habitat characterized by extremely dry conditions(Nakazato et al., 2010). The leaves of S. pennellii plants,therefore, exhibit morphological and anatomical featuresthat are likely adaptations to dry conditions (McDowellet al., 2011; Hali�nski et al., 2015), including thick leaves(Koenig et al., 2013). Moreover, a set of homozygousintrogression lines (ILs) harboring defined, partiallyoverlapping segments of the S. pennellii genome in an

otherwise S. lycopersicum background (Eshed and Zamir,1995) has been used to successfully map a number ofQTL, including fruit metabolite concentrations (Fridmanet al., 2004; Schauer et al., 2006), yield (Semel et al., 2006),and leaf shape (Chitwood et al., 2013). Here, we used acustom-built dual confocal profilometer to obtain precisemeasurements of leaf thickness across the IL panel andidentified QTL for this trait in tomato. Leaf thicknesscorrelates with other facets of leaf shape as well as a suiteof traits associated with desiccation tolerance and lowerproductivity. We investigated the anatomical manifesta-tions of thickness in tomato and found a prominent in-crease in palisade cell height in many thick ILs. Finally,we inferred comparative gene regulatory networks ofearly leaf development (plastochron stages P1–P4) intwo thick lines using organ-specific RNA sequencing(RNA-Seq) and identified molecular networks thatpattern S. pennellii-like desert-adapted leaves.

RESULTS

Complex Genetic Architecture of Leaf Thickness across S.pennellii ILs

To investigate the genetic architecture and patterning ofleaf thickness in the S. pennellii IL panel, we used a custom-built dual confocal profilometer device (Supplemental Fig.S1), which generates precise thickness measurementsthroughout the leaflet lamina at a range of resolutions(0.1–1 mm2) and at high throughput. The device makesuse of two confocal lasers positioned on either side ofthe sample and calculates thickness by measuring thedistance between each of the sample’s surfaces and thecorresponding laser probe. Finally, we visualize thick-ness as a heatmap of thickness values across the surfaceof the leaf lamina (Fig. 1A).

We first compared leaflet thickness in S. lycopersicumvariety M82 and its desert relative S. pennellii LA0716.Our confocal profilometer measurements showed thatS. pennellii leaflets are thicker than those of domesti-cated tomato, as reported previously (Fig. 1; Koeniget al., 2013), demonstrating the capacity of this deviceto quantitatively detect fine differences in leaf laminathickness. We compared dynamic growth patterns ofthe two species under water-limited conditions andshow that, unlike the domesticated species, S. pennelliiis unaffected by drought (Fig. 1C). This observationhighlights the importance of understanding the pat-terning of developmental traits in this species, such asleaf thickness, which may contribute to drought tol-erance. We proceeded to measure leaf thickness acrossthe S. pennellii IL panel in field conditions.

We used mixed linear regression models to compareeach of the ILs with the domesticated parent M82(Supplemental Data Set S1) and found that 31 ILs hadsignificantly thicker leaflets than the M82 parent, whilefive had transgressively thinner leaflets. The overallbroad-sense heritability for leaflet thickness is 39.1%(Fig. 2). The lines with thickest leaflets are IL5-4, IL5-3,IL8-1, IL4-3, IL8-1-1 (containedwithin IL8-1), and IL2-5,

Plant Physiol. Vol. 175, 2017 377

Leaf Thickness in Tomato


http://www.plantphysiol.org/cgi/content/full/pp.17.00790/DC1




while IL4-1-1, IL2-6-5, IL9-1-3, IL12-4-1, and IL2-1 havethinner leaves than the M82 parent.

Based on the observation that the heritability valuefor leaf thickness is 39.1%, we reasoned that environ-mental factors are likely to play a role in modulatingleaf thickness. We thus compared our field experimentwith leaf thicknessdata for vegetative leaves of greenhouse-grown plants. We selected 20 ILs, which were highlysignificant for leaf thickness differences from M82 infield conditions (P , 0.001), and observed that onlysome of these lines also are significantly thicker thanthe domesticated parent in greenhouse conditions(P , 0.05; Supplemental Fig. S2A). Finally, our ob-servations suggest that leaf thickness varies acrossthe shoot of a number of our select thick leaf ILs, withpost-flowering leaves having thicker leaves thanvegetative leaves (Supplemental Fig. S2B).

For each leaflet in our field experiment, we also quan-tified LMA,which reflects both thickness and density andis traditionally used as a proxy for leaf thickness. Al-though the heritability for LMA is similar to that for

thickness (33.2% and 39.1%, respectively), significantQTL for these two traits do not overlap consistently(Supplemental Data Set S1).

Leaf Thickness and LMA Are Correlated with DistinctSuites of Traits in Tomato

We generated pairwise correlations between leafletthickness, LMA, and a suite of other previously publishedtraits, including metabolite, morphological, enzymaticactivity in fruit pericarp, seed-related, developmental,and elemental profile-related traits (SupplementalData Sets S2–S4; Chitwood et al., 2013, and refs.therein). Spearman’s correlation coefficients with sig-nificant q values (q , 0.050) are reported in Figure 2B.Leaf thickness and LMA are correlated (rho = 0.423, q =0.003). Leaf thickness also correlates with leaf shapeparameters, such as roundness (rho = 0.328, q = 0.044),aspect ratio (rho = 20.327, q = 0.045), and the first twoprincipal components of the elliptical Fourier descriptorsof leaflet shape (EFD.PC1 rho = 0.414, q = 0.004 and EFD.

Figure 1. Desert-adapted tomato plantshave thicker leaves than domesticated to-mato and are resistant to drought. A,Thickness across leaflet blades of domes-ticated (S. lycopersicum M82) and desert-adapted (S. pennellii) tomatoes measuredwith a custom-built dual confocal profi-lometer device (Supplemental Fig. S1). Themedian thickness of the S. lycopersicumleaflet shown here is 211 mm, and that forS. pennellii is 294 mm. B, Confocal imagesof propidium iodide-stained leaflet crosssections. Bars = 200 mm. C, Total shootarea normalized by taking the square rootof pixels (px) from top-view phenotypingimages over 16 d in three water treatments(n = 8). Gray shading reflects SE.

378 Plant Physiol. Vol. 175, 2017

Coneva et al.









PC2 rho = 0.406, q = 0.005). Thickness is negatively cor-related with several reproductive traits, including yield(rho = 20.337, q = 0.037), seed weight (rho = 20.342, q =0.033), and seed number per plant (rho = 20.339, q =0.036). Moreover, leaf thickness is negatively correlatedwith leaf stomatal ratio, the relative density of stomata onthe abaxial and adaxial sides of the leaf (rho =20.352, q =0.031), and positively correlated with Glu dehydrogenaseactivity (rho = 0.367, q = 0.017) and seed galactinol content(rho = 0.342, q = 0.048).LMA is associated with a distinct suite of traits from

leaf thickness. In addition to a positive correlation withthe content of some enzymes (GAPDH and ShikimateDE) and metabolites (Glu), LMA is significantly nega-tively correlated with the accumulation of Na and Mgin all leaflets tested. LMA, but not leaf thickness, also issignificantly positively correlated with total plant weight,reflecting vegetative biomass accumulation.

Thick IL Leaves Have Elongated PalisadeParenchyma Cells

Leaf cross sections of field-grown M82 and select ILswith increased leaf thickness, as well as greenhouse-grown S. pennellii leaves, were stained with propidium

iodide to assess the anatomical changes that lead to in-creased leaf thickness. We observed that, relative to theM82 parent, the S. pennellii parent and several ILs havean elongated palisade mesophyll cell layer correspond-ing to the adaxial layer of photosynthesizing cells in to-mato leaves (Fig. 3). Palisade parenchyma elongation isespecially dramatic for IL1-3, IL2-5, IL4-3, and IL10-3.Both leaf thickness and palisade elongation phenotypesare attenuated for vegetative leaves of greenhouse-grown plants (Supplemental Figs. S2 and S3A).

Anatomy and Early Leaf Development in Select ILs withThick Leaves

To capture an overall view into the core mechanismsof leaf thickness patterning, we further analyzed linesIL2-5 and IL4-3. We selected IL2-5 due to its dramaticanatomy in field conditions (Fig. 3) and its lack of othercharacterized leaf morphology phenotypes (Chitwoodet al., 2013), while IL4-3 leaflets are both significantlythicker and less serrated than those of the domesticatedparent (Fig. 2; Supplemental Data Set S1; circularity, theratio between leaflet area and the square of its perim-eter, reflects lobing and serration). To further investi-gate the relationships between genetic determinants of

Figure 2. QTL for leaf thickness in tomato. A, Leaflet thickness values across the S. pennellii IL panel. Colors indicate the level ofsignificance in comparisons of each ILwithM82 (arrow). B, Significant correlations (Spearman’s rho) between leaf thickness or LMAand a suite of other traits across the S. pennellii IL panel (q , 0.05). Traits are grouped by type: ION, elemental profile; MOR,morphological; DEV, developmental; ENZ, enzyme activity; SED, seed metabolite content (Supplemental Data Sets S3 and S4).








leaf thickness in these ILs, we generated a double ho-mozygous line combining the entire S. pennellii seg-ments of IL2-5 and IL4-3.

Double homozygotes (IL2-5/IL4-3) have signifi-cantly thicker leaves than M82 at both vegetative (Fig.4A; P = 0.019) and postflowering stages (SupplementalFig. S2B) in greenhouse conditions. Additionally, IL2-5/IL4-3 plants have significantly smoother marginsthan either of the IL parents (Fig. 4B), suggesting ad-ditive genetic interactions for both of these traits. Wenext compared the dimensions of the mesophyll celllayers in each IL and the double homozygote line todetermine the contribution that each cell layer makes tothe observed increase in leaflet thickness.We found thatpalisade mesophyll cells are significantly larger in IL2-5/IL4-3 than in M82 leaves (Supplemental Fig. S4).Furthermore, the ratios of palisade cell length to bothtotal leaf thickness and the length of the spongy me-sophyll are significantly larger in IL2-5/IL4-3 than inM82 leaves (Supplemental Fig. S4). IL2-5 shows similar,albeit less pronounced, trends to the double homozy-gote line, while in IL4-3, both spongy and palisademesophyll cell layers are longer than in M82, with thespongy mesophyll layer making the most significantcontribution to leaf thickness.

Since increases in cell size often are driven by en-dopolyploidy, we performed flow cytometry on fullyexpanded vegetative leaves of each genotype and ob-served increased ploidy profiles in all lines relative tothe domesticated parent (Fig. 4C). Notably, the doublehomozygote line exhibited higher ploidy levels thanboth single ILs and the S. pennellii parent (Fig. 4C;Supplemental Fig. S4). Notably, we also observed atrend to increased ploidy in several greenhouse-grownthick ILs (IL7-4-1 and IL8-1; Supplemental Fig. S3B).

To understand if alterations in leaf size occur duringearly stages of leaf ontogeny in these lines, we quanti-fied P3 organ dimensions and compared them with theM82 parental line. For this, we assembled 3D confocalreconstructions of vegetative shoot apices, calculatedthe surface mesh, extracted P3 leaf primordia, and quan-tified their total volume, length, and mean diameter. Wefound that IL4-3 P3 leaf primordia are significantly larger

than M82 primordia in terms of overall volume (P =0.0179) as well as both length (P = 0.0035) and diameter(P = 0.0230). In IL2-5 P3, volume (not statistically sig-nificant) and diameter (P = 0.0116) are increased, whilelength is comparable to that in M82. Although P3 pri-mordia of double homozygote plants were statisticallyindistinguishable from those of M82 plants except forshorter arc length (P = 0.0411; Fig. 4D), our observationsalso suggest that double homozygote leaves increase insize dramatically between P3 and P4 stages (SupplementalFig. S5).

Transcriptomic Signatures of Early Leaf Development inThicker ILs

To investigate the molecular events that define thepatterning of IL2-5 and IL4-3 leaves, we isolated leafprimordia from each IL and the two parents (M82 or S.lycopersicum and S. pennellii) at four successive stages ofdevelopment: P1 (containing the shoot apical meristem[SAM] and the youngest leaf primordium), P2 and P3(characterized by leaflet emergence), and P4 (typicallythe onset of cell differentiation; Fig. 5A). For S. pennellii,P1 sampleswere composed of the SAM, P1, and P2, sincethese organs were not separable by hand dissection.Thus, the S. pennellii transcriptomic data set includessamples designated as P1, P3, and P4. Principal com-ponent analysis of the resulting RNA-Seq data, afternormalization and filtering, shows that samples groupclearly by organ stage (Fig. 5B, PC2). In addition, PC1separates S. pennellii samples from all other genotypes.To investigate how IL leaves are similar to the S. pennelliiparent, we looked for genes that are differentiallyexpressed (DEGs) between corresponding stages of eachIL and the M82 parent while also being differentiallyexpressed between M82 and S. pennellii. In other words,we identified the set of DEGs for each organ stage that iscommon to each IL and S. pennellii relative to M82. ForP2, we considered only the comparisonwithM82, as ourS. pennellii data set did not include independently dis-sected P2 stage primordium samples (Supplemental Fig.S6; Supplemental Data Set S5).

Figure 3. Anatomical manifestations of thickerleaves. A, Confocal images of propidium iodide-stained cross sections of field-grown M82, selectILs, and S. pennellii grown in greenhouse condi-tions. Bars = 50 mm. B and C, Representative leafthickness plots (B) and leaflet binary images offield-grown plants (C) as for A.


Coneva et al.














We identified a total of 812 DEGs across the P1 to P4stages in IL2-5, and of these, 544 are up-regulated in atleast one organ stage while 269 are down-regulated (Fig.5C). In IL4-3, we detected 632 DEGs, 361 of which areup-regulated and 271 are down-regulated in the IL (Fig.5C). Many of the DEGs are differentially expressed atmore than one stage (Fig. 5C; Supplemental Data Set S5).Additionally, based on tomato transcription factor (TF)annotation by Suresh et al. (2014), we identified putativeTF-encoding genes among each IL’s DEG sets. Myb-related, Ethylene Responsive, MADS, and WRKY arethe abundant classes of TF-encoding DEGs in IL2-5,while in IL4-3, TFs belonging to bZIP and Myb-relatedare highly represented families (Supplemental Fig. S7).We identified differentially expressed TF-encoding

genes that are common to the two ILs and the S. pennelliiparent (Fig. 6), reasoning that some of these can be reg-ulators of leaf thickness. Five of the seven shared TF-encoding genes are up-regulated in the ILs relative toM82. A MADS-box TF (Solyc12g087830) is up-regulatedat all stages in both ILs, while two additional inflores-cence meristem-related TFs, LFY-like (Solyc03g118160)and AP2-like (Solyc07g049490), are differentiallyexpressed at corresponding stages in both ILs. TheSHORTROOT-like (SHR-like) GRAS TF Solyc08g014030is up-regulated at P2 in both ILs, while its expressionincreases at each progressive stage and peaks at P4 in all

genotypes. A putative JASMONATE ZIM-domain pro-tein (JAZ1; Solyc12g009220) also is up-regulated at P2 inboth ILs, while a LIM domain protein (Solyc04g077780)is up-regulated in the ILs at P3 (in IL4-3) and P4 (bothILs; Fig. 6A).

Next, we compared the expression profiles of genesknown to be involved in tomato leaf development(Ichihashi et al., 2014). We selected only genes that aredifferentially expressed in the samedirection in each IL andS. pennellii relative to the domesticated parent M82 andhighlighted genes that are common to both thick ILs toarrive at a set of entities thatmaybe core to thepatterningofleaf thickness (Fig. 6B). A gibberellin 20-oxidase-encodinggene (GA 20-ox; Solyc03g006880) is up-regulated at P3in both ILs and throughout the P1-P3 interval in IL4-3.A set of two closely related ULTRAPETALA1 genes(Solyc12g010360 and Solyc12g010370) is down-regulatedat all leaf developmental stages in both ILs. A number ofleaf development regulators are additionally differentiallyexpressed in either of the ILs. Some noteworthy classesinclude entities related to auxin metabolism or transport(auxin efflux carrier, IAA-carboxymethyltransferase, andYUCCA-like monooxygenase), leaf complexity, lobing,and serrations (three BEL1-like TFs, CUC2-like, andBOP2-like), meristem maintenance or patterning (twoBAM1-like receptor kinases and an AP2-like TF), and celldivision and expansion (GRF1 and ROT3-like TFs).

Figure 4. Leaf morphology and ploidy of IL2-5/IL4-3 double homozygote plants. A, Representative propidium iodide-stainedleaflet cross sections (left) and thickness measurements (right) for the seventh leaf of greenhouse-grown M82 and double ho-mozygous IL2-5/IL4-3 plants (n = 10). **, P, 0.01. Bars = 200 mm. B, Circularity (ratio of area to the square of the perimeter) ofdistal lateral leaflets as in A. Silhouettes of representative M82 and IL2-5/IL4-3 leaflets are shown above the bars. Letters indicatestatistical significance in each pairwise genotype comparison (P , 0.05). C, Distribution of relative nuclear sizes reflectingendoreduplication in leaflets as in A and B (n = 5). Letters denote statistical significance between pairwise genotype comparisonsat each ploidy level. Sp, S. pennellii. D, Leaf plastochron P3 dimensions calculated from 3D surface reconstructions of vegetativeshoot apices (n = 9; *, P , 0.05 and **, P , 0.01 relative to M82).







Similarly, we also queried DEG sets for entities an-notated as cell cycle or endoreduplication to assesswhether these two thick ILs share a common trajectory ofcellular events during leaf ontogeny (Fig. 6C). Overall,we observed distinct expression profiles for these genesin IL2-5 and IL4-3.

Finally, to broadly characterize the types of processesthat may regulate the molecular networks of early leafdevelopment in the ILs, we applied Gene Ontology(GO) enrichment analysis (agriGO; Du et al., 2010;Supplemental Data Set S6) and identified statisticallyenriched promoter motifs among the organ-specificDEG sets (Supplemental Data Set S7). Importantly, weobserved that, at P4, the set of up-regulated genes inIL2-5 is enriched for biological process terms relating tophotosynthesis (GO:0015979) and translation (GO:0006412),while down-regulated genes at this stage are enrichedfor terms relating to DNA binding (GO:0003677). Ourpromoter motif analysis showed that motifs associatedwith regulation by abiotic factors such as light, circadianclock, water availability, and temperature are prominentamong IL2-5 DEGs. In addition, binding sites for devel-opmental regulators, hormone-associated promotermotifs, and a cell cycle regulator are among the listof significant motifs. Among development-associatedmotifs, CArG (MADS-box), BEL1-like, and SBP-boxTF-binding sites also are enriched significantly in

both IL2-5 and IL4-3 DEG sets. (Supplemental Fig. S8;Supplemental Data Set S7).

Gene Regulatory Networks of Early Leaf Development inThick ILs

To detect regulators of early leaf development thateach IL (IL2-5 and IL4-3) shares with the S. pennelliiparent, we inferred dynamic Bayesian networks usingthe IL and S. pennellii overlapping DEG sets describedin the previous section (de Luis Balaguer et al., 2017).Additionally, we only allowed putative TF-encodinggenes (Suresh et al., 2014) as source nodes (genes thatcontrol the expression of other coexpressed genes). First,we constructed individual networks for each leaf devel-opmental stage, for which an overlap with S. pennellii datais available (P1, P3, and P4) and then combined the resultsto visualize the overall S. pennellii-like leaf developmentalnetworks (Fig. 7; Supplemental Data Set S8). The IL2-5network (Fig. 7A) contains two major regulators, whichare central to more than one developmental stage: aSQUAMOSA promoter-binding protein-like domain gene(SBP-box 04g, Solyc04g064470) and a CONSTANS-likezinc finger (Zn-finger CO-like 05g, Solyc05g009310;Supplemental Data Set S8). Similarly, the IL4-3 network(Fig. 7C) features two central regulators: a BEL1-like

Figure 5. Comparative transcriptomics of leaf development in two thick ILs and their parents. A, Successive stages of leaf de-velopment (plastochrons P1–P4 colored as in the legend to B) were dissected fromM82, S. pennellii (Sp), and thick IL2-5 and IL4-3. B, Principal component analysis of normalized RNA-Seq read counts. C, Venn diagrams (not to scale) depict an overview ofDEGs (q, 0.05) that are shared in each IL and the S. pennellii parent relative toM82. The number of DEGs unique to each organ isshownwithin the ellipses, and those common to all organs are shown in the center. The total number of DEGs at each plastochronstage is shown outside the ellipses.


Coneva et al.









homeodomain TF gene (BEL1 04g, Solyc04g080780) anda MADS-box domain-containing gene (MADS-box 12g,Solyc12g087830; Supplemental Data Set S8). Impor-tantly, few nodes are shared between the organ-specificnetworks of IL2-5 and IL4-3. We surveyed each networkfor shared differentially expressed leaf developmentgenes and found that GA 20-ox 03g (Solyc03g006880) ispresent in both networks but is regulated by differentsets of TFs in each IL (Fig. 7, B and D).We also inferred a second set of networks for each of

the ILs by identifying DEGs using similar criteria tothose described above. However, in contrast to theprevious set of networks, where genes were separatedinto organ stages based on differential expression ateach discrete stage, we used a clustering approach togroup regulators and select coexpressed gene setsaccording to expression profiles. For these analyses,we also included P2 DEGs (IL versus M82) to ensurethe continuity of expression profiles (SupplementalData Set S9). This approach allowed us to examine amore dynamic view of early developmental processes.The resulting networks (Supplemental Data Set S9)

feature a putative auxin-responsive TF, AUX/IAA 12g(Solyc12g096980), for both ILs (Fig. 7, E and F). More-over, the AUX/IAA 12g subnetwork or IL2-5 includesthe SHR-like GRAS domain TF that is up-regulatedduring leaf development in both ILs (GRAS 08g,Solyc08g014030; Figs. 6A and 7E).

DISCUSSION

Leaf Thickness Has a Complex Genetic Architecture inDesert-Adapted Tomato and Is Associated with OverallLeaf Shape, Desiccation Tolerance, and Decreased Yield

While extensive progress has been made dissectingthe molecular-genetic patterning of two-dimensionalleaf morphology, relatively little is known about thethird dimension of leaf shape: thickness. Here, weused a custom-built dual confocal profilometer toobtain direct measurements of leaf thickness acrossthe S. pennellii 3 S. lycopersicum IL panel (Eshed andZamir, 1995; Fig. 1; Supplemental Fig. S1) and iden-tified QTL for this trait (Fig. 2A). We found that

Figure 6. Comparative expression profiles ofgenes in three functional categories across leafdevelopment (P1–P4) in thick IL2-5 and IL4-3.A, Transcription factors common to both ILs. B,Genes involved in leaf development in tomato(as in Ichihashi et al., 2014). C, Genes anno-tated to encode components of the cell cycle orubiquitin proteasome pathway (they containone of the terms cell cycle, cyclin, ubiquitin,E2F, mitosis, mitotic, or SKP). Plastochron stageswith statistically significant DEGs (q , 0.05)relative to M82 are marked with asterisks.Genes that are differentially expressed in at leastone stage in both ILs are marked in boldface.










nearly half of the ILs have significantly thicker leavesthan the domesticated parentM82,while a small numberhave transgressively thinner leaves. The broad-senseheritability for leaf thickness in this experiment is mod-erate (39%). Collectively, these observations point to acomplex genetic basis for this trait. A previous quanti-tative genetic analysis of a suite of desert-adaptive traitsin the same S. pennellii IL panel found fewer significantlythicker lines and lower heritability (12%) for this trait(Muir et al., 2014). However, the previous study esti-mated thickness as the ratio of LMA to leaflet drymatter

content, while we measured thickness directly. Fur-thermore, our study was conducted in field conditions,while Muir et al. (2014) measured the trait usinggreenhouse-grown plants. Given that environment signif-icantly affects the magnitude of this trait (SupplementalFig. S2), it is not surprising that these studies report onlypartially overlapping outcomes.

In order to understand how variation in leaf thicknessrelates to other traits, particularly to LMA, we calculatedpairwise correlation coefficients among all leaf shapeand elemental profile traits as well as a collection of

Figure 7. Select leaf development gene regulatory subnetworks for IL2-5 (A–C) and IL4-3 (D–F). Subnetworks for regulatorscentral to more than one plastochron stage are shown in A andD. AGA 20-ox gene (Solyc03g006880) and its regulators in each ILare shown in B and E. Subnetworks of dynamic gene regulatory networks, showing interactions of an AUX/IAATF (AUX/IAA 12g,Solyc12g096980)with other source nodes, are shown in C and F. Gene identifiers of the highlighted nodes are as follows: SBP-box04g, Solyc04g064470; BEL1-like 04g, Solyc04g080780; MADS-box 12g, Solyc12g087820; MYB TF 05g, Solyc05g007710;bHLH TF 04g, Solyc04g074810; GRAS 08g, Solyc08g014030; Myb 07g, Solyc07g052490; WRKY 05g, Solyc05g015850; AUX/IAA 06g, Solyc06g008580;Myb 03g, Solyc03g005570;Myb 08g, Soly08g005870; andWRKY 02g, Solyc02g080890. Nodes andedges are colored according to the legend.


Coneva et al.





previously published traits (summarized by Chitwoodet al., 2013; Supplemental Data Sets S3 and S4). As ex-pected, leaf thickness and LMA are correlated signifi-cantly across the IL panel. However, the two traits havedistinct sets of significant trait correlations (Fig. 2B).Taken together with our finding that only a proportionof ILs harbor QTL for both leaf thickness and LMA(Supplemental Data Set S1), these data suggest thatthickness and LMA are likely patterned by partiallydistinct mechanisms and that direct measurements ofleaf thickness are necessary to further dissect the geneticbasis of this trait.Leaf thickness is correlated significantly with leaf

shape traits such as aspect ratio and the first two prin-cipal components of elliptical Fourier descriptors ofoverall shape. However, our data do not establishwhether this correlation reflects a common patterningmechanism or developmental and/or mechanical con-straints among these traits. Alternatively, the relativelymodest correlations (rho values between 0.33 and 0.41)could reflect independent variation in these traitsresulting in considerable flexibility in final leaf mor-phology, as suggested by Muir et al. (2017).Leaf thickness is negatively correlated with yield-

related traits, which suggests a tradeoff between in-vestments in vegetative and reproductive biomass thatis further substantiated by the positive correlation be-tween LMA and plant weight (Fig. 2B). Some studiessupport the hypothesis of a tradeoff between LMA andrapid growth (Smith et al., 1997; Poorter et al., 2009),while others find poor coordination between growthrate and LMA (Muir et al., 2017). Finally, leaf thick-ness is correlated significantly with leaf stomatal ratio,Glu dehydrogenase activity, and galactinol content inseeds, a suite of traits associated with desiccation tol-erance in plants (Taji et al., 2002; Lightfoot et al., 2007).We also observed negative correlations between LMAand the accumulation of several elements in leaves,most notably Na and Mg (Fig. 2C). This finding sup-ports the idea that LMAand thickness are distinct traits,and that LMA reflects the material composition ofleaves, while leaf thickness is a developmentally pat-terned trait.

Thicker S. pennellii IL Leaves Have Elongated PalisadeMesophyll Cells

The observed elongated palisade mesophyll cells inthe leaves of several field-grown ILs with significantlythicker leaves (Fig. 3A), as well as in the desert-adaptedS. pennellii parent, suggest that the dorsiventral ex-pansion of palisade mesophyll cells contributes mostprominently to increased leaf thickness. This hypothe-sis is supported by the fact that palisade cell heightincreases more significantly than the total height of thespongy mesophyll in thick leaves of double homozy-gous IL2-5/IL4-3 lines (Supplemental Fig. S4). Palisadecell height is positively correlated with photosyntheticefficiency (Niinemets et al., 2009; Terashima et al., 2011)

and water storage capacity in succulent Crassulaceanacid metabolism plants (Nelson et al., 2005). Our dataalso indicate that the magnitudes of palisade cell elon-gation, as well as overall leaf thickness, are modulatedby environmental inputs (Fig. 2; Supplemental Fig. S2).High light has been shown to mediate increased leafthickness (Poorter et al., 2009; Wuyts et al., 2012; Kalveet al., 2014) as well as specifically palisade cell elon-gation (Kozuka et al., 2011) in Arabidopsis, whileelongated palisade cells promote a more efficient dis-tribution of direct light throughout the photosyntheticmesophyll compared with shorter cells (Brodersenet al., 2008; Brodersen and Vogelmann, 2010). Thus,thicker leaves composed of elongated palisade cellsmay be an adaptation to desert-like dry, direct lightenvironments, whereby the magnitude of these traitsis responsive to these environmental cues. Consistentwith this hypothesis, we observed that IL2-5 DEGpromoters are enriched inmotifs that reflect sensitivityto abiotic stimuli, prominently light and water status(Supplemental Fig. S8; Supplemental Data Set S7).

Mechanisms of Cell Enlargement in Thick ILs: IncreasedPloidy and Alterations in Cell Cycle-RelatedGene Expression

We compared the size of palisade mesophyll cells inleaf cross sections of thick IL2-5, IL4-3, and a homozy-gous line combining both introgression segments andobserved larger palisade cells compared with M82(Supplemental Fig. S4), suggesting a link between leafthickness and cell size in tomato. Furthermore, weshowed significantly higher ploidy levels in the leavesof these lines relative to the domesticated parent (Fig.4C), indicating that increased endoreduplication mayunderpin larger cells and, ultimately, thicker leaves. Apartially overlapping series of cell division, cell ex-pansion, and cell differentiation events underlie leafdevelopment (Efroni et al., 2008). These processes aretightly coordinated to buffer perturbations in overallorgan shape and size (Beemster et al., 2003; Tsukaya,2003). Thus, the relative timing and duration of any ofthese events can impact leaf size and morphology.Additionally, different tissue types in the leaf can havedistinct schedules of cellular events during leaf ontog-eny; for example, in Arabidopsis, palisade mesophyllcells have a shorter window of cell division comparedwith epidermal cells, and thus an earlier onset of cellexpansion and endoreduplication, resulting in differ-ences in cell volumes and geometry (Wuyts et al., 2012;Kalve et al., 2014). Given the prominent contribution ofspecific cell types to leaf thickness (palisade mesophyllcells in IL2-5, for example, versus both palisade andspongy mesophyll cells in IL4-3; Supplemental Fig. S4),kinematic studies to capture the timing and extent oftissue-specific cell division and endoreduplication areneeded to fully address the dynamic cellular basis ofleaf thickness patterning. The observed increase in P3organ volume and thickness in IL4-3 and, to a lesser













extent, IL2-5 relative toM82 (Fig. 4D) support the notionthat differences in the trajectory of cellular events duringearly leaf ontogeny may underpin leaf thickness.

Comparative gene expression profiles of early leafontogeny in IL2-5 and IL4-3 show evidence of S. pennellii-like alterations in cell proliferative activity in thesethick ILs. Specifically, among a small set of shareddifferentially expressed genes, the GRAS-domain TFGRAS 08g (Solyc08g014030) is up-regulated at P2 inboth lines (Fig. 6A; Supplemental Data Set S5). Thisgene is closely related to the Arabidopsis gene encodingSHR (Huang et al., 2015), which together with anotherGRAS-domain TF, SCARECROW (SCR), regulates theduration of cell proliferation in leaves (Dhondt et al.,2010). Moreover, consistent with previous reports, IL2-5 and IL4-3 DEGs are enriched for E2F binding sitemotifs (Supplemental Data Set S7; Ranjan et al., 2016).E2F TFs act downstream of SHR and SCR to regulateprogression through the S-phase of the cell cycle(Dhondt et al., 2010). These data support the notion thatthe extent and/or duration of cell proliferation under-pin increased thickness in these lines. Another set ofDEGs that distinguish the thick ILs and the S. pennelliiparent from domesticated tomato include three geneswith predicted functions in regulating the cell cycleand cell expansion activities: a LIM-domain protein(Solyc04g077780), a JAZ1 TF (Solyc12g009220), and aGA20-ox (Solyc03g006880; Fig. 6). LIM-domain proteinshave been implicated in a variety of functions, includingregulationof the cell cycle andorgan size inArabidopsis (Liet al., 2008). GA 20-ox encodes a key GA biosynthetic en-zyme, which acts to promote cell elongation (Hisamatsuet al., 2005; de Lucas et al., 2008) and, thus, determinacyduring leaf morphogenesis of compound leaves, such asthose of tomato (Hay et al., 2002). Moreover, JAZ proteinsact as transcriptional repressors and are a central hub in thesignaling circuit that integrates environmental cues, such aslight quality, to balance growth and defense (for review,see Hou et al., 2013). Finally, it is noteworthy that abioticcues such as light quality and abscisic acid have beenshown to interact and modulate the activity of GA 20-oxand JAZ, and the Arabidopsis LIM-domain protein DA1,respectively, thereby establishing a conceptual meansof environmental regulation of leaf thickness pattern-ing. Taken together with higher endopolyploidy levels,the shared expression patterns for these genes betweenboth thick ILs and the S. pennellii parent suggest that leafthickness may bemodulated by changes in the trajectoryof cellular events during leaf ontogeny, specifically, theduration of cell proliferation, and the timing and extentof cell expansion. Further validation of this hypothesis isnecessary to evaluate the contribution of dynamic, en-vironmentally responsive changes in cell division andendoreduplication to leaf thickness.

Gene Expression Networks Point to Distinct LeafOntogeny in IL2-5 and IL4-3

Since we observed a set of shared DEGs in lines IL2-5and IL4-3, we hypothesized that general patterns of leaf

ontogeny also may be shared between these lines, sug-gesting a core shared trajectory of leaf thickness pattern-ing. However, we found that dynamic Bayesian networksof gene coexpression in IL2-5 and IL4-3 are largely distinct(Fig. 7, A and D; Supplemental Data Sets S8 and S9).

For example, central to the organ-specific network of IL2-5 is an SBP-box domain gene, SBP 04g (Solyc04g064470),which is highly expressed throughout leaf development inIL2-5 (Fig. 7, A and B; Supplemental Fig. S6). SBP TFsregulate various aspects of plant growth by controlling therate and timing of developmental events, including leafinitiation rate (for review, see Preston andHileman, 2013).Furthermore, the promoters of IL2-5 DEGs are enrichedfor SBPmotifs (Supplemental Data Set S7), supporting thecentral role of this group of TFs during IL2-5 leaf ontog-eny. Interestingly, GO terms for photosynthesis andtranslation are enriched among P4 up-regulated genes.This observation suggests that processes associated withcell differentiation (i.e. photosynthetic gene function andprotein translation) are precociously activated in IL2-5relative to domesticated tomato and supports a hypothe-sis whereby the overall schedule of leaf developmentalevents may be hastened in IL2-5.

In contrast, a central node in the IL4-3 coexpressionnetwork is a BEL1-like 04g (Solyc04g080780). BEL1-likehomeodomain proteins interact with class I KNOX TFsto pattern the SAM and lateral organs, including leafcomplexity (Kimura et al., 2008; Hay and Tsiantis, 2010)and the extent of lobing and serrations (Kumar et al.,2007). Like S. pennellii, IL4-3 leaflets have significantlysmoother margins (fewer serrations) than M82, asreflected in increased circularity (Fig. 4B; Holtan andHake, 2003; Chitwood et al., 2013).

These distinct dynamic patterns of leaf ontogeny thateach IL shares with the desert-adapted parent may reflectaspects of leaf development unrelated to the patterning ofleaf thickness, such as the patterning of leaf complexityand leaflet shape in IL4-3. Alternatively, it is also possiblethat the core mechanism of leaf thickness patterning isachieved by regulation of the timing and extent of cellularactivities, such as the balance between cell proliferationand the onset of cell expansion and endoreduplication,with a number of potential molecular networks needed toaccomplish these roles. An observation supporting thismodel is the fact that IL2-5 and IL4-3 have nonoverlap-ping sets of cell cycle-related DEGs. This hypothesis isconsistent with the additive phenotypes of IL2-5/IL4-3double homozygotes (Fig. 4; Supplemental Fig. S4),whereby IL-specific regulators may converge and act ad-ditively to modulate the expression of a common set oftargets (such as GA 20-ox and SHR-like) that regulate cellsize and shape and, eventually, leaf thickness. Extendingthese analyses to a broader set of thick lines will yieldfurther insight into the validity of this hypothesis.

CONCLUSION

Leaf thickness is a functional trait associated withthe ability of plants to occupy arid environments. Our


Coneva et al.









understanding of how this trait is patterned is poor,due in part to challenges in measuring leaf thickness in aprecise and high-throughputmanner.Here,we combine anovel tool, a custom-built confocal profilometer designedto measure leaf thickness directly and efficiently, withanatomical and transcriptomic observations across a panelof S. pennellii ILs to assemble a broad and comprehensiveframework of the genetic architecture, anatomical mani-festations, and molecular patterning processes associatedwith thick leaves in tomato. We find a complex geneticbasis for this trait and a prominent role for environmentalcues in modulating it. Thick leaves are associated withelongation of the palisade parenchyma and an increase inleaf ploidy, suggesting that increased cell size underpinsthick leaves in tomato. Finally, transcriptomic data showthat relative gene expression trends for several putativeregulators of cell elongation and meristem determinacyare shared among two thick ILs and their desert tomatoparent, implicating these processes as contributors to leafthickness pattering during development. Ultimately,given the complex genetic architecture of this trait, itsresponsiveness to environmental factors, and its de-pendence on plant age, further experiments comparingmore thick lines, working with smaller introgressionsegments to reduce the contribution of epistasis, andexplicitly addressing the roles of environment and age,are necessary to understand the mechanistic basis of leafthickness. This work establishes the necessary founda-tion to further dissect this highly functional develop-mentally patterned trait.

MATERIALS AND METHODS

Plant Material and Growth Conditions

Seeds for 76 Solanum pennellii ILs (LA4028–LA4103; Eshed and Zamir, 1995)and the Solanum lycopersicum domesticated variety M82 (LA3475) wereobtained either from Dr. Neelima Sinha (University of California, Davis) orfrom the TomatoGenetics Resource Center (University of California, Davis). Allseeds were treated with 50% (v/v) bleach for 3 min, rinsed with water, andgerminated in Phytatrays (P1552; Sigma-Aldrich). Seeds were left in the darkfor 3 d, followed by 3 d in light, and finally transferred to greenhouse conditionsin 50-plug trays. Hardened plants were transplanted to field conditions at theBradford Research Station in Columbia, Missouri (May 21, 2014) with 3 m be-tween rows and about 1 m spacing between plants within rows. A nonexper-imental M82 plant was placed at both ends of each row, and an entire row wasplaced at each end of the field to reduce border effects on experimental plants.The final design had 15 blocks, each consisting of four rows with 20 plants perrow. Each of the 76 ILs and two experimental M82 plants were randomizedwithin each block. IL6-2 was excluded from final analyses due to seed stockcontamination. For the analysis of leaf primordia by confocal microscopy andRNA-Seq, IL2-5, IL4-3, M82, and S. pennellii seeds were germinated as aboveand transferred to pots in controlled growth chamber conditions: irradiance at400 mmol m22 s21, 23°C, and 14-h days. Growth conditions for the droughtphenotyping experiment were irradiance of 200 mmol m22 s21 at a daytimetemperature of 22°C and 18°C at night.

Whole-Plant Phenotyping under Drought

The LemnaTec Scanalyzer plant phenotyping facility at theDonaldDanforthPlant ScienceCenter (LemnaTec)wasused tophenotype approximately 3-week-old S. lycopersicum and S. pennellii plants (n = 8 per genotype) subjected to one ofthree watering regimes: 40% field capacity, 20% field capacity, and no watering(0% field capacity). Top-view images of each plant taken every second night

over 16 nights were analyzed using custom pipelines in Lemna Launcher(LemnaTec software) to extract total plant pixel area (a proxy for biomass).

Trait Measurements

After flowering (July 2014), four fully expanded adult leaves were harvestedfrom each plant; the adaxial (upper) surfaces of distal lateral leaflets harvestedfrom the left side of the rachis were scannedwith a flatbed scanner to obtain rawJPGfiles. Themiddle portion of each leafletwas then attached on a custom-builddual confocal profilometer device (Supplemental Fig. S1), and the thickness ofeach leaflet was measured across the leaflet surface at a resolution of 1 mm2.Median thickness was calculated across each leaflet using values in the range0 to 2mm, and these median values were averaged across four leaflets per plantto arrive at a single robust metric of leaf thickness. Finally, entire leaflets weredried and their dry mass was used to calculate LMA for each leaflet. Leafletoutline scans were processed using custom macros in ImageJ (Abràmoff et al.,2004) to segment individual leaflets and to threshold and binarize each leafletimage. Shape descriptors area, aspect ratio, roundness, circularity, and solidity(described in detail by Chitwood et al., 2013) were extracted from binary im-ages. Additionally, elliptical Fourier descriptors for leaflet outlines were de-termined using SHAPE (Iwata and Ukai, 2002). For this analysis, 20 harmonicswith four coefficients each were used to derive principal components that de-scribe major trends in the shape data.

Elemental Profiling (Ionomics)

Distal lateral leaflets of fully expanded young and old leaves of the sameplants as above were collected from five individuals of each genotype. Wholeleaflets were weighed and digested in nitric acid at 100°C for 3 h. Elementalconcentrations were measured using an inductively coupled plasma massspectrometer (Elan DRC-e; Perkin Elmer) following the procedure described byZiegler et al. (2013). Instrument-reported concentrations were corrected forlosses during sample preparation and changes in instrument response duringanalysis using yttrium and indium internal standards and a matrix-matchedcontrol run every 10th sample. Final concentrations were normalized to sampleweight and reported in milligrams of analyte per kilogram of tissue.

Statistical Analyses and Data Visualization

All statistical analyses and visualizationwere carried out using R packages (RCore Team, 2013). QTL were identified using the mixed-effect linear modelpackages lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al., 2015)withM82as intercept, IL genotype as afixed effect, andfield position attributes (block, row,and column) as random effects. Only effects with significant variance (P, 0.05)were included in the final models. For elemental composition data, leaf age(young and old) also was included as a random effect unless the variance due toage was the greatest source of variance; in such cases, young and old sampleswere analyzed separately. Heritability values represent the relative proportion ofvariance due to genotype. For the quantification of organ volumeparameters andphotosynthesis measurements, linear models were used to test the effect ofgenotype. All plots were generated with the package ggplot2 (Wickham, 2009).

Trait Correlations and Hierarchical Clustering

For trait correlation analyses, we included all traits reported in this study andmeasured on the same set of field-grown IL individuals (leaf thickness, LMA,leaflet shape traits, and elemental profiles).We also included several sets ofmeta-data detailed in Supplemental Data Set S3, including developmental, morpho-logical, fruit pericarp metabolite content, enzyme activity, and seed metabolitecontent traits (fromChitwood et al., 2013, and refs. therein). Spearman correlationcoefficients (rho) were calculated between each pair of traits using the rcorrfunction in Hmisc (Harrell et al., 2015), and P values for the correlations werecorrected for false discovery rate using the Benjamini-Hochberg procedure(Supplemental Data Set S4). Hierarchical clustering and visualization of signifi-cant correlations (q , 0.05) of leaf thickness and LMA were clustered (hierar-chical ward algorithm) and visualized using pheatmap (Kolde, 2015).

Estimation of Nuclear Size Profiles by Flow Cytometry

Distal lateral leaflets were harvested from the seventh leaf of greenhouse-grown 6-week-old plants and immediately chopped in 1 mL of ice-cold buffer








LB01 as described by Dole�zel et al. (2007). The resulting fine homogenate wasfiltered through a 30-mm Partec CellTrics filter (5004-004-2326) and incubatedwith 50 mg mL21 propidium iodide (Thermo Fisher; P21493) and 50 mg mL21

RNase A (Qiagen; 19101) for 20 min on ice. Fluorescence scatter data werecollected without gating using a BD Acuri CS6 instrument (BD Biosciences).Plots of event count as a function of fluorescence area were used to estimate theproportion of nuclei of sizes corresponding to 2C, 4C, and 8C in each genotype.

Confocal Microscopy, 3D Reconstructions, and OrganVolume Quantification

For mature leaf cross sections, field-grown leaves were fixed in 4% (w/v)formaldehyde, 5% (v/v) glacial acetic acid, and 50% (v/v) ethanol, vacuuminfiltrated, dehydrated through an ethanol series, rehydrated to 100% water,stained in 0.002% (w/v) propidium iodide (Thermo Fisher; P21493) for 2 h,dehydrated to 100% ethanol, and finally cleared in 100% methyl salicylate(Sigma; M6752) for 7 d. Hand sections were visualized with a Leica SP8 laserscanning confocalmicroscope usingwhite light laser excitation at 514 nmwith a203 objective. Two partially overlapping images were captured for each crosssection and merged into a single image using the Photomerge function inAdobe Photoshop CC 2014 (Adobe Systems). For the quantification of P3 leafprimordium dimensions, shoot apices (shoot apical meristem and P1–P4) of14-d-old seedlings grown in controlled conditions were excised, fixed, pro-cessed, and stained as detailed for leaf cross sections above. Confocal stackswere obtained at software-optimized intervals and exported as TIFF files. Rawstack files were imported into MorphoGraphX (Barbier de Reuille et al., 2015).After Gaussian filtering, the marching cube surface reconstruction functionwasused (cube size = 5 mm and threshold = 7,000). The resulting surface mesh wassmoothed, subdivided twice, and exported as a PLY file. Tominimize the effectsof trichomes on P3 volume, all meshes were trimmed in MeshLab (Cignoniet al., 2008). The volume, length, and diameter of processed P3 meshes werecalculated using custom scripts in MatLab (MathWorks). Briefly, first, wedetected the boundary of each hole and calculated its centroid point. We con-nected boundary points of each hole to its centroid and filled the triangle faces.After filling all the holes, 3D mesh represents the closed surface. Then, wecalculated the volume based on the divergence theorem, whichmakes use of thefact that the inside fluid expansion equals the flux ( F

!) of the fluid out of the

surface (S).When the flux is F!¼ ðx; 0; 0Þ, the volume isV ¼ ∯ð F!$ n!Þ dS, where

n! is normal vector. Thus, for each triangle, we computed the normal vectorn!¼ ðxn; yn; znÞ, the areaA, and the centroid point P ¼ ðxp; yp; zpÞ. The volumeVis the summation of Axnxp for all triangles. To estimate organ arch length, wemade use of the fact that the Laplace-Beltrami eigenfunctions are deformation-invariant shape descriptors (Rustamov, 2007). We thus employed its firsteigenfunction, which is associated with the smallest positive eigenvalue, anddiscretized the eigenfunction values into 50 sets to compute the centroid pointto each set. We fit a cubic function by fixing two end-point constraints to thosecentroid points to get a smooth principal median axis. Note that the two endpoints were adjusted manually to correct for artifacts. The length of this axis isused to quantify the length of the organ. Finally, we calculated mean organdiameter as

d ¼ 2

ffiffiffiffiffiffiVpL

r

RNA-Seq Library Preparation and Sequencing

Apices of 14-d-old IL2-5, IL4-3, M82, and S. pennellii plants grown in arandomized design under controlled growth conditions were hand dissectedusing a dissecting microscope to separate plastochrons P4, P3, P2, and P1+SAMorgans corresponding approximately to leaves L5 to L8. For S. pennellii plants,wewere not able to separate P2 primordia from the apex, so we obtained P4, P3,and SAM+P1+P2 samples. Dissected organs were removed from the apex inless than 60 s and immediately fixed in 100% ice-cold acetone to preserve theintegrity of RNA in the sample. Each biological replicate is a pool of 10 indi-viduals, and a total of five biological replicates were obtained for each geno-type/organ combination. RNAwas extracted using the PicoPure RNA IsolationKit (Thermo Fisher) according to the manufacturer’s protocol with the optionalon-column DNase treatment. RNA integrity was assessed by running allsamples on an Agilent RNA 6000 Pico chip (Agilent Technologies), and threebiological replicates with RNA integrity . 7 were selected for further pro-cessing. Double-stranded cDNA amplified using the Clontech SMARTer PCRcDNA synthesis kit (634926; TaKaRa Bio) was fragmented for 15 min using

Fragmentase (M0348; New England Biolabs) and processed into Illumina se-quencing libraries as follows: the ends of 1.53 AMPure XP bead (A63880;Agencourt)-purified fragmented DNA was repaired with End Repair EnzymeMix (E6050; New England Biolabs) andKlenowDNAPolymerase (M0210; NewEngland Biolabs), followed by dA tailing using Klenow 39-59 exonuclease(M0212; New England Biolabs). The Illumina TruSeq universal adapter dimerwas ligated to library fragments with rapid T4 DNA Ligase (L6030-HC-L; En-zymatics) followed by three rounds of 13 AMPure XP bead purification toremove unligated adapter. Finally, libraries were enriched and indexed by PCRusing Phusion HiFi Polymerase mix (M0531; New England Biolabs). Illuminalibraries were quantified using a nanodrop, pooled to a final concentration of20 nM, and sequenced as single-end 100-bp reads on Illumina HiSeq2500 at theSchool ofMedicine Genome TechnologyAccess Center, Washington Universityin St. Louis (https://gtac.wustl.edu/).

RNA-Seq Data Analysis

Adapters and low-quality bases were removed using Trimmomatic (Bolgeret al., 2014) with default parameters. Trimmed reads were mapped tothe ITAG2.3 S. lycopersicum genome (https://solgenomics.net/organism/Solanum_lycopersicum/genome; Tomato Genome Consortium, 2012) usingbowtie2 (Langmead and Salzberg, 2012) to obtain SAM (Sequence AlignmentMap) files. After sorting and indexing of SAM files, BAM (Binary SAM) fileswere generated using samtools commands (Li et al., 2009). The BEDtoolsmulticov tool (Quinlan and Hall, 2010) was then used to obtain read counts perannotated gene for each sample. Subsequent analysis was done with the Rpackage edgeR (Robinson et al., 2010). After normalization for library size,20,231 genes with at least one count per million reads across three samples wereretained for further analysis. Lists of DEGs were generated between pairwisesample combinationswith q, 0.05. For IL2-5 and IL4-3 at P1, P3, and P4 stages,we identified genes that are differentially expressed relative to M82 in both theIL and the S. pennellii parent to interrogate S. pennellii-like changes in geneexpression in the ILs. For P2, the list of DEGs in each IL reflects changes relativeto M82 only (Supplemental Data Set S5).

GO, MapMan, and Promoter Motif Enrichment Analyses

Lists of IL organ-specific DEGs were interrogated for enrichment of GOterms using agriGO (http://bioinfo.cau.edu.cn/agriGO/; Du et al., 2010) withdefault parameters (Fisher’s exact significance test and Yekutieli false discoveryrate adjustment at q , 0.05). We further divided DEG gene lists into ILup-regulated and down-regulated genes and report significant terms inSupplemental Data Set S6. We tested IL organ-specific DEGs for the enrichmentof annotated promoter motifs using a custom R script (Julin Maloof, personalcommunication). Briefly, functions in the Bioconductor Biostrings package(Pages et al., 2009) were implemented to count the frequency of 100 knownmotifs in the promoters of DEGs (1,000-bp upstream sequence) and calculateP values for enrichment based on these counts. We report exact matches ofknown motifs and motifs with up to one mismatch in IL up-regulated anddown-regulated organ-specific gene sets (Supplemental Data Set S7).

IL Organ-Specific Gene Network Inference

To infer IL organ-specific networks (Fig. 7, A–D; Supplemental Data Set S8),we selected DEGs between IL2-5/M82 (IL4-3/M82) and S. pennellii/M82 foreach organ (P1, P3, and P4; q, 0.05). Since coexpression analysis can inform thelikelihood that genes interact, or participate in the same functional pathway, theselected genes for each IL (IL2-5 or IL4-3) and each organ were clustered basedon their coexpression across genotypes. To perform clustering, the Silhouetteindex (Rousseeuw, 1987) followed byK-means (MacQueen, 1967)were applied.After clustering, networks were inferred as described by de Luis Balaguer et al.(2017). Briefly, for each DEG, we identified a set of potential regulators andmeasured the likelihood of gene-target regulation using a Bayesian Dirichletequivalence uniform (Buntine, 1991). Genes that had the highest value of theBayesian Dirichlet equivalence uniformwere chosen as regulators, and of these,only TFs (as annotated by Suresh et al., 2014) were further considered as reg-ulatory (source) nodes. To obtain the final IL2-5 and IL4-3 organ-specific net-works, the networks for each cluster were connected. For this, we foundregulations among the cluster hubs (node of each individual network with thelargest degree of edges leaving the node) by using the same Bayesian Dirichletequivalence uniform metric. In addition, we implemented a score to estimatewhether the inferred interactions were activations or repressions. The score was


Coneva et al.


https://gtac.wustl.edu/

https://solgenomics.net/organism/Solanum_lycopersicum/genome

https://solgenomics.net/organism/Solanum_lycopersicum/genome


http://bioinfo.cau.edu.cn/agriGO/





calculated for each edge, and it measured the ratio between (1) the conditionalprobability that a gene is expressed given that its regulator was expressed in theprior time point and (2) the conditional probability that a gene is expressedgiven that its regulator was not expressed in the prior time point. If the firstconditional probability is larger than the second one, then the parent was foundto be an activator, and vice versa. In the case of a tie, the edgewas found to havean undetermined sign. Networks for each organ were jointly visualized inCytoscape (Shannon et al., 2003).

Dynamic IL Network Construction

To infer dynamic IL networks (Fig. 7, E and F; Supplemental Data Set S9), weselected DEGs between IL2-5/M82 or IL4-3/M82 and S. pennellii/M82 for eachorgan (P1, P3, and P4; q, 0.05 or fold change. 2 and q, 0.2). All DEGs in IL2-5 or IL4-3 were clustered in four groups, corresponding to the four develop-mental stages: each gene was assigned to the developmental stage where itshowed the maximum expression. A network was then inferred for each de-velopmental stage as described for the IL organ-specific networks. To ensurethat all potential regulators of each gene were considered, genes from thepreceding developmental stage were included in the inference of the network ofeach developmental stage. The final network for each IL was visualized inCytoscape (Shannon et al., 2003).

Accession Numbers

RNA-Seq data have been deposited in the National Center for Biotechnol-ogy Information Short Read Archive under BioProject identifier PRJNA396585containing 45 BioSamples.

Supplemental Data

The following supplemental materials are available.

Supplemental Figure S1. Dual confocal profilometer device used to mea-sure leaf thickness.

Supplemental Figure S2. Comparison of leaf thickness of select ILs as afunction of shoot position and field versus greenhouse conditions.

Supplemental Figure S3. Representative leaf cross sections and flow cy-tometry of leaf 6/7 for 10 ILs harboring leaf thickness QTLs grown ingreenhouse conditions.

Supplemental Figure S4. Mean dimensions of palisade and spongy meso-phyll cell layers in select thick leaf ILs and representative flow cytometryhistograms of leaf 7 and post-flowering leaves from each genotype.

Supplemental Figure S5. Representative shoot apex reconstructions high-lighting the appearance of early- and late-stage leaf primordia for eachgenotype in Figure 4.

Supplemental Figure S6. Summary of differentially expressed genes inIL2-5 and IL4-3.

Supplemental Figure S7. Expression profiles of differentially expressedputative TFs in IL2-5 and IL4-3.

Supplemental Figure S8. Summary of enriched promoter motifs amongdifferentially expressed genes in IL2-5 and IL4-3.

Supplemental Data Set S1. Trait value estimates and heritability for leafthickness, LMA, and leaflet shape across the IL panel.

Supplemental Data Set S2. Trait value estimates and heritability for ele-mental concentration across the IL panel.

Supplemental Data Set S3. Summary of all measured and meta-data traitsused in the correlation matrix.

Supplemental Data Set S4. Pairwise trait correlation matrix including sig-nificance values.

Supplemental Data Set S5. List of differentially expressed genes (q , 0.05)in each organ (P1–P4) for the comparison M82/IL overlapping withM82/S. pennellii.

Supplemental Data Set S6. List of significantly enriched (q , 0.05) GOterms for gene sets listed in Supplemental Data Set S5.

Supplemental Data Set S7. List of enriched (q , 0.05) promoter motifs forgene sets listed in Supplemental Data Set S5.

Supplemental Data Set S8. List of organ-specific (P1, P3, P4) gene inter-actions for IL2-5 and IL4-3.

Supplemental Data Set S9. List of dynamic gene interactions for IL2-5 andIL4-3.

ACKNOWLEDGMENTS

S. pennellii IL panel seedswere provided byDr. Neelima Sinha (University ofCalifornia, Davis) and the Tomato Genetics Resource Center (University ofCalifornia, Davis). We thank Dr. Ivan Baxter and Dr. Greg Ziegler (DonaldDanforth Plant Science Center) for generating elemental profile data and Dr.Julin Maloof (University of California, Davis) for sharing custom promoter en-richment analysis scripts. We acknowledge the advice and assistance of Dr.Noah Fahlgren, Dr. Malia Gehan, and Melinda Darnell (Donald DanforthPlant Science Center) with drought phenotyping experiments. We thankDr. Elizabeth Kellogg (Donald Danforth Plant Science Center) for insightfuldiscussions and comments on the article.

Received June 14, 2017; accepted August 1, 2017; published August 9, 2017.

LITERATURE CITED

Abràmoff MD, Magalhães PJ, Ram SJ (2004) Image processing with ImageJ.Biophoton Int 11: 36–42

Bates D, Maechler M, Bolker B, Walker S (2014) lme4: Linear mixed-effectsmodels using Eigen and S4. R package version, 1: 1–23

Barbier de Reuille P, Routier-Kierzkowska AL, Kierzkowski D, BasselGW, Schüpbach T, Tauriello G, Bajpai N, Strauss S, Weber A, Kiss A,et al (2015) MorphoGraphX: a platform for quantifying morphogenesisin 4D. eLife 4: 05864

Becker B (2007) Function and evolution of the vacuolar compartment ingreen algae and land plants (Viridiplantae). Int Rev Cytol 264: 1–24

Beemster GT, Fiorani F, Inzé D (2003) Cell cycle: the key to plant growthcontrol? Trends Plant Sci 8: 154–158

Bohmert K, Camus I, Bellini C, Bouchez D, Caboche M, Benning C (1998)AGO1 defines a novel locus of Arabidopsis controlling leaf develop-ment. EMBO J 17: 170–180

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer forIllumina sequence data. Bioinformatics 30: 2114–2120

Brodersen CR, Vogelmann TC (2010) Do changes in light direction affectabsorption profiles in leaves? Funct Plant Biol 37: 403–412

Brodersen CR, Vogelmann TC, Williams WE, Gorton HL (2008) A newparadigm in leaf-level photosynthesis: direct and diffuse lights are notequal. Plant Cell Environ 31: 159–164

Buntine W (1991) Theory refinement on Bayesian networks. In BD D’Ambrosio,P Smets, PP Bonissone, eds, Proceedings of the Seventh Conference onUncertainty in Artificial Intelligence. Morgan Kaufmann Publishers, LosAngeles, CA, pp 52–60

Chitwood DH, Kumar R, Headland LR, Ranjan A, Covington MF, IchihashiY, Fulop D, Jiménez-Gómez JM, Peng J, Maloof JN, et al (2013) A quan-titative genetic basis for leaf morphology in a set of precisely defined tomatointrogression lines. Plant Cell 25: 2465–2481

Cignoni P, Callieri M, CorsiniM, DellepianeM, Ganovelli F, Ranzuglia G (2008)Meshlab: an open-source mesh processing tool. In V Scarano, R De Chiara, UErra, eds, Eurographics Italian Chapter Conference, 2008. pp 129–136

de Lucas M, Davière JM, Rodríguez-Falcón M, Pontin M, Iglesias-PedrazJM, Lorrain S, Fankhauser C, Blázquez MA, Titarenko E, Prat S (2008)A molecular framework for light and gibberellin control of cell elonga-tion. Nature 451: 480–484

de Luis Balaguer MA, Fisher AP, Clark NM, Fernandez-Espinosa MG,Moller B, Weijers D, Lohmann JU, Williams C, Lorenzo O, Sozzani R(2017) An inference approach combines spatial and temporal gene ex-pression data to predict gene regulatory networks in Arabidopsis stemcells. PNAS https://doi.org/10.1101/140269

Dhondt S, Coppens F, De Winter F, Swarup K, Merks RM, Inzé D,Bennett MJ, Beemster GT (2010) SHORT-ROOT and SCARECROWregulate leaf growth in Arabidopsis by stimulating S-phase progressionof the cell cycle. Plant Physiol 154: 1183–1195
























https://doi.org/10.1101/140269


Dole�zel J, Greilhuber J, Suda J (2007) Estimation of nuclear DNA contentin plants using flow cytometry. Nat Protoc 2: 2233–2244

Du Z, Zhou X, Ling Y, Zhang Z, Su Z (2010) agriGO: a GO analysis toolkitfor the agricultural community. Nucleic Acids Res 38: W64–W70

Efroni I, Blum E, Goldshmidt A, Eshed Y (2008) A protracted and dynamicmaturation schedule underlies Arabidopsis leaf development. Plant Cell20: 2293–2306

Eshed Y, Zamir D (1995) An introgression line population of Lycopersiconpennellii in the cultivated tomato enables the identification and finemapping of yield-associated QTL. Genetics 141: 1147–1162

Fridman E, Carrari F, Liu YS, Fernie AR, Zamir D (2004) Zooming in on aquantitative trait for tomato yield using interspecific introgressions.Science 305: 1786–1789

Garnier E, Laurent G (1994) Leaf anatomy, specific mass and water contentin congeneric annual and perennial grass species. New Phytol 128: 725–736

Gibson AC (1982) The anatomy of succulence. In IP Ting, M Gibbs, eds,Crassulacean Acid Metabolism. American Society of Plant Physiologists,Rockville, MD, pp 1–17

Hali�nski ŁP, Kalkowska M, Kalkowski M, Piorunowska J, TopolewskaA, Stepnowski P (2015) Cuticular wax variation in the tomato (Solanumlycopersicum L.), related wild species and their interspecific hybrids.Biochem Syst Ecol 60: 215–224

Harrell, FE Jr, Dupont, C, et al (2015). Hmisc: Harrell Miscellaneous. Rpackage version 3.16-0. http://CRAN.R-project.org/package=Hmisc

Hay A, Kaur H, Phillips A, Hedden P, Hake S, Tsiantis M (2002) Thegibberellin pathway mediates KNOTTED1-type homeobox function inplants with different body plans. Curr Biol 12: 1557–1565

Hay A, Tsiantis M (2010) KNOX genes: versatile regulators of plant develop-ment and diversity. Development 137: 3153–3165

Hisamatsu T, King RW, Helliwell CA, Koshioka M (2005) The involve-ment of gibberellin 20-oxidase genes in phytochrome-regulated petioleelongation of Arabidopsis. Plant Physiol 138: 1106–1116

Holtan HE, Hake S (2003) Quantitative trait locus analysis of leaf dissectionin tomato using Lycopersicon pennellii segmental introgression lines.Genetics 165: 1541–1550

Hou X, Ding L, Yu H (2013) Crosstalk between GA and JA signaling me-diates plant growth and defense. Plant Cell Rep 32: 1067–1074

Huang W, Xian Z, Kang X, Tang N, Li Z (2015) Genome-wide identifica-tion, phylogeny and expression analysis of GRAS gene family in tomato.BMC Plant Biol 15: 209

Ichihashi Y, Aguilar-Martínez JA, Farhi M, Chitwood DH, Kumar R, MillonLV, Peng J, Maloof JN, Sinha NR (2014) Evolutionary developmentaltranscriptomics reveals a gene network module regulating interspecific di-versity in plant leaf shape. Proc Natl Acad Sci USA 111: E2616–E2621

Iwata H, Ukai Y (2002) SHAPE: a computer program package for quantitativeevaluation of biological shapes based on elliptic Fourier descriptors. J Hered93: 384–385

John GP, Scoffoni C, Sack L (2013) Allometry of cells and tissues withinleaves. Am J Bot 100: 1936–1948

Kalve S, Fotschki J, Beeckman T, Vissenberg K, Beemster GT (2014)Three-dimensional patterns of cell division and expansion throughoutthe development of Arabidopsis thaliana leaves. J Exp Bot 65: 6385–6397

Kimura S, Koenig D, Kang J, Yoong FY, Sinha N (2008) Natural variationin leaf morphology results from mutation of a novel KNOX gene. CurrBiol 18: 672–677

Koenig D, Jiménez-Gómez JM, Kimura S, Fulop D, Chitwood DH,Headland LR, Kumar R, Covington MF, Devisetty UK, Tat AV, et al(2013) Comparative transcriptomics reveals patterns of selection in do-mesticated and wild tomato. Proc Natl Acad Sci USA 110: E2655–E2662

Kolde R (2015) pheatmap: Pretty Heatmaps. R package version 1.0.2Kozuka T, Kong SG, Doi M, Shimazaki K, Nagatani A (2011) Tissue-

autonomous promotion of palisade cell development by phototropin 2 inArabidopsis. Plant Cell 23: 3684–3695

Kumar R, Kushalappa K, Godt D, Pidkowich MS, Pastorelli S, HepworthSR, Haughn GW (2007) The Arabidopsis BEL1-LIKE HOMEODOMAINproteins SAW1 and SAW2 act redundantly to regulate KNOX expres-sion spatially in leaf margins. Plant Cell 19: 2719–2735

Kuznetsova A, Christensen RH, Bavay C, Brockhoff PB (2015) Automatedmixed ANOVA modeling of sensory and consumer data. Food Qual Prefer40: 31–38

Langmead B, Salzberg SL (2012) Fast gapped-read alignment with Bowtie2. Nat Methods 9: 357–359

Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G,Abecasis G, Durbin R (2009) The sequence alignment/map format andSAMtools. Bioinformatics 25: 2078–2079

Li Y, Zheng L, Corke F, Smith C, Bevan MW (2008) Control of final seedand organ size by the DA1 gene family in Arabidopsis thaliana. GenesDev 22: 1331–1336

Lightfoot DA, Mungur R, Ameziane R, Nolte S, Long L, Bernhard K,Colter A, Jones K, Iqbal MJ, Varsa A, et al (2007) Improved droughttolerance of transgenic Zea mays plants that express the glutamate de-hydrogenase gene (gdhA) of E. coli. Euphytica 156: 103–116

MacQueen J (1967) Some methods for classification and analysis of mul-tivariate observations. In LM Le Cam, J Neyman, eds, Proceedings of theFifth Berkeley Symposium on Mathematical Statistics and Probability.University of California Press, Berkeley, CA, pp 281–297

McDowell ET, Kapteyn J, Schmidt A, Li C, Kang JH, Descour A, Shi F, LarsonM, Schilmiller A, An L, et al (2011) Comparative functional genomicanalysis of Solanum glandular trichome types. Plant Physiol 155: 524–539

McHale NA (1992) A nuclear mutation blocking initiation of the lamina inleaves of Nicotiana sylvestris. Planta 186: 355–360

McHale NA (1993) LAM-1 and FAT genes control development of the leafblade in Nicotiana sylvestris. Plant Cell 5: 1029–1038

Muir CD, Conesa MÀ, Roldán EJ, Molins A, Galmés J (2016) Weak co-ordination between leaf structure and function among closely relatedtomato species. New Phytol 213: 1642–1653

Muir CD, Pease JB, Moyle LC (2014) Quantitative genetic analysis indi-cates natural selection on leaf phenotypes across wild tomato species(Solanum sect. Lycopersicon; Solanaceae). Genetics 198: 1629–1643

Nakazato T, Warren DL, Moyle LC (2010) Ecological and geographicmodes of species divergence in wild tomatoes. Am J Bot 97: 680–693

Nelson EA, Sage TL, Sage RF (2005) Functional leaf anatomy of plants withCrassulacean acid metabolism. Funct Plant Biol 32: 409–419

Nicotra AB, Leigh A, Boyce CK, Jones CS, Niklas KJ, Royer DL, TsukayaH (2011) The evolution and functional significance of leaf shape in theangiosperms. Funct Plant Biol 38: 535–552

Niinemets U, Díaz-Espejo A, Flexas J, Galmés J, Warren CR (2009) Role ofmesophyll diffusion conductance in constraining potential photosyn-thetic productivity in the field. J Exp Bot 60: 2249–2270

Ogburn R, Edwards EJ (2010) The ecological water-use strategies of suc-culent plants. Adv Bot Res 55: 179–225

Oguchi R, Hikosaka K, Hirose T (2005) Leaf anatomy as a constraint for pho-tosynthetic acclimation: differential responses in leaf anatomy to increasinggrowth irradiance among three deciduous trees. Plant Cell Environ 28: 916–927

Pages H, Aboyoun P, Gentleman R, DebRoy S (2009) String objects representingbiological sequences, and matching algorithms. R package version 2(2)

Pieruschka R, Poorter H (2012) Phenotyping plants: genes, phenes andmachines. Funct Plant Biol 39: 813–820

Poorter H, Niinemets U, Poorter L, Wright IJ, Villar R (2009) Causes andconsequences of variation in leaf mass per area (LMA): a meta-analysis.New Phytol 182: 565–588

Preston JC, Hileman LC (2013) Functional evolution in the plant SQUAMOSA-PROMOTER BINDING PROTEIN-LIKE (SPL) gene family. Front PlantSci 4: 80

Quinlan AR, Hall IM (2010) BEDTools: a flexible suite of utilities forcomparing genomic features. Bioinformatics 26: 841–842

R Core Team (2013) R: A language and environment for statistical com-puting. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/

Ranjan A, Budke JM, Rowland SD, Chitwood DH, Kumar R, Carriedo L,Ichihashi Y, Zumstein K, Maloof JN, Sinha NR (2016) eQTL regulatingtranscript levels associated with diverse biological processes in tomato.Plant Physiol 172: 328–340

Robinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductorpackage for differential expression analysis of digital gene expressiondata. Bioinformatics 26: 139–140

Roderick ML, Berry SL, Noble IR, Farquhar GD (1999) A theoretical ap-proach to linking the composition and morphology with the function ofleaves. Funct Ecol 13: 683–695

Rousseeuw PJ (1987) Silhouettes: a graphical aid to the interpretation andvalidation of cluster analysis. J Comput Appl Math 20: 53–65

Rustamov RM (2007) Laplace-Beltrami eigenfunctions for deformation in-variant shape representation. In Proceedings of the Fifth EurographicsSymposium on Geometry Processing. Eurographics Association, Barcelona,Spain, pp 225–233


Coneva et al.


http://CRAN.R-project.org/package=Hmisc

http://www.R-project.org/

http://www.R-project.org/


Sack L, Frole K (2006) Leaf structural diversity is related to hydraulic ca-pacity in tropical rain forest trees. Ecology 87: 483–491

Schauer N, Semel Y, Roessner U, Gur A, Balbo I, Carrari F, Pleban T,Perez-Melis A, Bruedigam C, Kopka J, et al (2006) Comprehensivemetabolic profiling and phenotyping of interspecific introgression linesfor tomato improvement. Nat Biotechnol 24: 447–454

Semel Y, Nissenbaum J, Menda N, Zinder M, Krieger U, Issman N,Pleban T, Lippman Z, Gur A, Zamir D (2006) Overdominant quanti-tative trait loci for yield and fitness in tomato. Proc Natl Acad Sci USA103: 12981–12986

Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,Schwikowski B, Ideker T (2003) Cytoscape: a software environment forintegrated models of biomolecular interaction networks. Genome Res13: 2498–2504

Smith SD, Monson RK, Anderson JE (1997) Plant processes and responsesto stress. In Physiological Ecology of North American Desert Plants.Springer, Berlin, pp 45–71

Suresh BV, Roy R, Sahu K, Misra G, Chattopadhyay D (2014) Tomato ge-nomic resources database: an integrated repository of useful tomato genomicinformation for basic and applied research. PLoS ONE 9: e86387

Taji T, Ohsumi C, Iuchi S, Seki M, Kasuga M, Kobayashi M, Yamaguchi-Shinozaki K, Shinozaki K (2002) Important roles of drought- and cold-inducible genes for galactinol synthase in stress tolerance in Arabidopsisthaliana. Plant J 29: 417–426

Terashima I, Hanba YT, Tholen D, Niinemets Ü (2011) Leaf functionalanatomy in relation to photosynthesis. Plant Physiol 155: 108–116

Tomato Genome Consortium (2012) The tomato genome sequence pro-vides insights into fleshy fruit evolution. Nature 485: 635–641

Tsuge T, Tsukaya H, Uchimiya H (1996) Two independent and polarizedprocesses of cell elongation regulate leaf blade expansion in Arabidopsisthaliana (L.) Heynh. Development 122: 1589–1600

Tsukaya H (2003) Organ shape and size: a lesson from studies of leafmorphogenesis. Curr Opin Plant Biol 6: 57–62

von Willert DJ (1992) Life Strategies of Succulents in Deserts: With SpecialReference to the Namib Desert. Cambridge University Press, Cambridge,United Kingdom

Wickham H (2009) ggplot2: Elegant Graphics for Data Analysis. Springer,New York

Wright IJ, Reich PB, Westoby M, Ackerly DD, Baruch Z, Bongers F,Cavender-Bares J, Chapin T, Cornelissen JH, Diemer M, et al (2004)The worldwide leaf economics spectrum. Nature 428: 821–827

Wuyts N, Massonnet C, Dauzat M, Granier C (2012) Structural assessmentof the impact of environmental constraints on Arabidopsis thaliana leafgrowth: a 3D approach. Plant Cell Environ 35: 1631–1646

Ziegler G, Terauchi A, Becker A, Armstrong P, Hudson K, Baxter I (2013)Ionomic screening of field-grown soybean identifies mutants with al-tered seed elemental composition. Plant Genome 6: http://doi.org/10.3835/plantgenome2012.07.0012




http://dx.doi.org/http://doi.org/10.3835/plantgenome2012.07.0012

http://dx.doi.org/http://doi.org/10.3835/plantgenome2012.07.0012


Date post:	24-Aug-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

Genetic Architecture and Molecular Networks Underlying ...endemic to coastalregions of the Atacama...

Documents