Analysis of Classic Tomato Mutants Reveals Influence of Leaf Vein Density on Fruit BRIX

Tomato bipinnate (bip) is a classic leaf mutant, with highly increased leaf complexity resulting from the loss of function of a BEL-LIKE HOMEODAMAIN (BELL) gene. Here, we analyzed several bip mutants and their isogenic wildtype backgrounds for a suite of leaf morphology traits, ranging from leaf complexity, leaflet shape and size, to leaf vascular density to investigate how changes in leaf morphology influence fruit traits. Our analyses showed an unexpected relationship between leaf vein density and fruit sugar levels, where leaf vein density was negatively correlated with fruit BRIX. RNA-Seq analysis suggested variation in Glucose-6-phosphate translocator2 (GPT2) gene expression caused correlated changes in leaf vein density and BRIX when bip mutant and wildtype were compared, suggesting that the correlation between leaf vein density and fruit sugar may result from the genes regulating leaf vein development that are also involved in regulating leaf sugar biosynthesis. Our results provide a resource for further exploration of the genetic basis for the complex relationship between fruit quality and leaf traits in natural populations.


INTRODUCTION 41
Tomato (Solanum lycopersicum) is one of the highest-value and most extensively used 42 vegetable crops worldwide. However, to meet increasing demand, modern tomato cultivars have 43 been selected for qualities such as size and firmness instead of taste (Beullens et al., 2008;Wang 44 and Seymour, 2017). Consequently, most of modern commercial varieties have lost their flavor 45 and are often tasteless (Klee and Tieman, 2013;Tieman et al., 2017). 46 Flavor of fruit is the sum of interactions between taste and aroma, whereas sugars and 47 acids are the two of primarily components to activate taste receptors and aroma components such 48 as volatile compounds activate olfactory receptors (Malundo et al., 1996;Baldwin et al., 2008;49 Tieman et al., 2012). Though the relative contribution of taste and aroma to fruit flavor has not 50 been clearly defined (Malundo et al., 1996), plenty of studies have shown the importance of 51 sugars and acids in determining fresh fruit flavor (Malundo et al., 1996;Baldwin et al., 2008; 52 Beckles et al., 2012). For tomato, the levels of sugars and acids not only contribute to tomato 53 taste (sweetness and sourness), but also are major factors affecting tomato overall flavor intensity 54 (Allen Stevens, 1979;Jones and Scott, 1984), and increasing sugar content of the fruit will 55 enhance tomato flavor (Malundo et al., 1996;Tieman et al., 2017). Recent studies have shown 56 that fruit sugar accumulation in modern tomato is two to three-fold less than that in wild species 57 (Beckles et al., 2012), which can account for the decline in flavor quality of tomato fruit. 58 Fruits are the primary photosynthetic sinks and over 80% of sugars in the fruit are 59 produced in the leaf through photosynthesis and subsequently translocated through the phloem 60 (Cocaliadis et al., 2014). Therefore, factors involved in regulating leaf photosynthesis, as well as 61 sugar biosynthesis and sugar transport would have an effect on sugar levels in fruit. Leaves are 62 the principle site of plant photosynthesis and leaf traits (e.g. shape and size) directly impact the 63 efficiency of light capture and photosynthetic carbon fixation (Smith et al., 1997;Sarlikioti et al., 64 9 in ML ( Figure S6). Thus, these genes might play an important role in regulating fruit BRIX and 251 leaf vascular density between bip0663, Lukullus, bip2, and M82. 252 In addition, many DEGs detected between the HB and LB groups were involved in regulating 253 both development and carbohydrate metabolism (Table S5). We therefore hypothesize that these 254 genes might be co-expressed during leaf vascular development and involved in regulating leaf 255 sugar metabolism either directly or indirectly through vascular transport processes. To validate 256 this hypothesis, we performed correlation analysis and constructed a gene coexpression network 257 across all genotypes using DEGs enriched in "transcription and development" and "carbohydrate 258 and biosynthesis" related GO terms. Correlation analysis showed most of "transcription and 259 development" related DEGs and "carbohydrate and biosynthesis"-related genes are highly 260 correlated, whereas GPT2 was shown to be positively correlated with many genes related to 261 carbohydrate and biosynthetic process ( Fig.10C and  Table S5 and is connected to other two communities 266 (C1 and C2), which compose a core network (with hub genes with >100 edges). C1 was enriched 267 for genes related to transcription initiation and cell fate specification GO terms, and C2 268 contained genes involved in glucose metabolic process, carbohydrate biosynthetic process, plant 269 organ development, hormone-mediated signaling pathway and shoot system development ( caused not only high leaf complexity but also rounder leaves. Results also show that leaflet 286 roundness of bip0663 and Lukullus (HB genotypes) was significantly higher than bip2 and M82 287 (LB genotypes). These results agree with previous research that showed tomato with rounder and 288 more circular leaves tend to have the highest sugar content in their fruit (Chitwood et al., 2013;289 Rowland et al., 2019). Although bip2 has rounder leaves than M82 it does not have measurably 290 higher fruit BRIX (Fig. 2B), suggesting that the roundness caused by the bip mutation does not, 291 by itself, lead to increased BRIX. 292 A previous study (Chitwood et al., 2013) suggested the correlation between leaf shapes 293 and fruit sugar content may be due to the impact of leaf shape on photosynthetic capacity. 294 However, we found that across the whole season, average photosynthetic rates of bip2 and M82 295 (LB genotypes) were higher than those of bip0663 and Lukullus, perhaps resulting from high 296 vein density (vein length per unit area) in the LB genotypes (Sack and Scoffoni, 2013). Thus, the 297 variation in fruit BRIX between these genotypes was likely due to aspects other than leaf 298 photosynthesis. In addition, while mutations in the BIP gene cause increased leaf complexity, 299 only bip0663 has increased fruit BRIX compared with Lukullus. This suggests that leaf 300 complexity is not the cause of changes in fruit BRIX between bip0663 and Lukullus. This is 301 consistent with the results from path modeling connections between leaf traits and fruit BRIX in 302 multiple heirloom cultivars (Rowland et al., 2019). The same path modeling also indicated that 303 photosynthesis was not a major contributor to fruit BRIX (Rowland et. al., 2019). 304 In leaves, sugars synthesized through photosynthesis are first loaded into leaf veins and 305 then transported out of the leaves to the rest of the plant sinks, such as fruit. Thus, leaf veins also 306 play an important role in fruit sugar accumulation by contributing to transport of carbohydrates 307 (Cataldo, 1974 and Lukullus leaves ( Fig.9), which can promote sugar export (Adams et al., 2007).

Seed germination and plant growth Conditions 387
Tomato seeds were treated with 50% bleach for 10 min and rinsed 3-5 times with water, 388 then placed on water dampened Phytatrays (Sigma Aldrich). Seeds were moved to the dark and 389 incubated at room temperature for 3 days, then transferred to a growth chamber set at 25°C with 390 16:8 photoperiod until seedlings had expanded cotyledons (approximately 4-7 days). The 391 seedlings were then transplanted to 72 Seedling Propagation trays and grown in the chamber for 392 7 days. After that, seedlings were transferred to the greenhouse or grown for 2 weeks and then 393 transplanted to field. The greenhouse plants were watered from the top to encourage hardening. 394 Field plants were watered with furrow irrigation once weekly. 395

Analysis of leaf complexity and shape 396
Mature fully expanded leaves from adult nodes (leaf 5 and above) were used for leaf 397 complexity and shape analysis, and at least five leaves were collected from each plant for 398 analysis. Leaf complexity is defined as the number of all leaflets present on the leaf. For leaf 399 shape analysis, intercalary and secondary/tertiary leaflets were ignored due to their irregular 400 shapes. Leaf shape was analyzed using a method previously described (Ichihashi et al., 2014). 401 After leaf complexity was measured, the leaflet images were used for shape and size analysis.

Analysis of leaf vein density 417
For leaf vein analysis, leaf discs with an area of 0.28 cm 2 and were collected by a hole 418 puncher from the second lateral primary leaflets of each plant (the sampling sites were located 419 between second-order veins, seen in Figure S1A). Leaf discs were cleared using a modified 420 method from the Ainsworth lab (Bishop et al., 2018; Rowland et al., 2019). Leaf discs were 421 heated in 80% EtOH for 20 minutes at 80℃ and this process repeated twice or until leaf discs 422 turned white. Leaf discs were then placed in 5% NaOH and heated to 80℃ for 5 minutes and 423 were cooled by incubating at room temperature for 10 minutes. After that NaOH was removed 424 and leaf discs were treated with 50% bleach for approximately 30 seconds. Bleach treatment was 425 repeated until leaf discs were clear white. Then leaf discs were washed by ddH 2 O and vacuum 426 infiltrated with 50% glycerol for 20 minutes. Cleared leaf discs were placed on slides and leaf 427 veins of leaf disc were imaged using Eclipse C1 plus microscope (Nikon Instruments Inc., NY,

Root Grafts 456
For the root grafting control group, bip0663 scions were grafted onto bip0663 rootstocks, 457 and Lukullus scions were grafted onto Lukullus rootstocks. For the treatment groups, bip0663 458 scions were grafted onto Lukullus rootstocks and Lukullus scions were grafted onto bip0663 459 rootstocks. This grafting was done with 30 day-old plants, and the junction for these grafts was 460 in the internode between the cotyledons and first true leaves. These grafted plants were allowed 461 to recover for 2 weeks before being transferred outside into 2-gallon pots. The grafts were 462 allowed to flower and fruit freely and were regularly checked to assure that there were no leaves 463 emerging from the rootstock portion of the grafts. This allowed us to determine if differences in 464 root sink strength could lead to differences in growth of fruit. The yield and fruit sugar were 465 measured. 466

Leaf Sugar and Starch Measurements 467
16 In order to determine the sugar and starch content of bip0663 and Lukullus, leaf discs 468 with an area of 0.28 cm 2 and were collected with a hole puncher from the second lateral primary 469 leaflets of each plant (the sampling sites were located between second-order veins, seen in 470 Figure S1A). These samples were taken at dusk and immediately placed into microcentrifuge 471 tubes containing 500ul of 100% EtOH. Samples were heated at 80°C for 20 minutes. Supernatant 472 was immediately removed, stored at -20°C, and used for determination of sugar content 473 (including sucrose, glucose, and fructose). Sugar content was determined as previously described 474 (Rowland et al., 2019). Leaf discs were kept for further processing and starch extraction. 475 Residual sugar was removed from leaves by two additional rounds of heating in 500ul 100% 476 EtOH, discarding the supernatant each time. Leaf discs were resuspended in 500ul 5% NaOH 477 and heated to 80°C for 20 minutes. Samples were cooled and neutralized with 125ul 5M HCl. 478 The supernatant was removed, and leaf discs were rinsed 2X with ddH 2 O. Discs were 479 resuspended in 500μL 50mM Sodium Acetate Buffer and bead beaten by hand for 30 seconds 480 after the addition of a small metal bead. Samples were centrifuged for 1 minute at 13K rpm. 25ul 481 of a starch degradation mixture, containing a final concentration of 1.3U Amyloglucosidase and 482 220U α-Amylase, was added. Samples were incubated at RT for 1 hour, followed by 65°C for 23 483 hours. Enzymes were deactivated by heating to 80°C for 5 minutes. These samples, now 484 containing glucose were quantified using the same protocol as for sugar samples above. 485

RNA Extraction 486
The shoots or leaf primordia were sampled and frozen in liquid nitrogen or stored at -487 80°C. About 100 mg tissue was ground using a Bead-Beater (Bio Spec Products Inc., OK, USA) 488 and RNA was then extracted using a published protocol in routine use in the laboratory 489 (Townsley et al., 2015). 490

RNA-seq Library Preparation and Sequencing 491
Tissue (e.g. shoot apical meristem, young leaf, and mature leaf) used for RNA-seq 492 libraries for Illumina sequencing were collected using the method previously described 493 (Townsley et al., 2015). RNA-seq libraries were prepared from collected tissues using the BrAD-494 seq method (Townsley et al., 2015). Libraries were prepared from 5 replicates of each type of 495 tissue, collected at fruiting stage (week 17). These RNA-seq libraries were sequenced at the 496 Vincent Coates Genomics facility at University of California, Berkeley on a single lane of the 497 Illumina Hi-Seq 4000 platform and 50-bp single-end reads were generated. A total of 326 M raw 498 paired-end 100 bp reads were generated, ranging from 5.7 to 16.3 M reads per library. 499

Preprocessing of Illumina Reads 500
Illumina Reads were trimmed and mapped to reference using CLC Genomics Workbench 501 11 (https://www.qiagenbioinformatics.com/). Low-quality reads with average Phred quality 502 score <20 and low-quality bases from 3' end of the reads were trimmed using the Trim tool in 503 CLC Genomics Workbench. The clean reads were aligned to the tomato reference genomic 504 sequence (ITAG 3.0, Solanum lycopersicum Heinz 1706) by using the Large Gap Read Mapping 505 tool in CLC Genomics Workbench, which models the presence of introns in the tomato genome 506 reference sequence, but are expected to be absent from the corresponding transcriptome. We 507 used the Transcript Discovery Plugin in CLC Genomics Workbench to generate a consensus 508 CDS mapping track based on the existing reference genome (ITAG3.2; https://solgenomics.net). 509 The new transcripts tracks were used as reference for subsequent read mapping. 510

Differential Expression Analysis 511
Reads from individual libraries were mapped to the annotated transcripts reference using 512 default parameters in the RNA-seq mapping tool in CLC Genomics Workbench. Then 513 differential expression analysis between bip mutants and wildtype across three tissues was 514 FDR<0.05 were considered to be differentially expressed genes (DEGs). 517

Statistical Analysis 554
Statistical analyses of leaf shape, leaf complexity, and leaf vein density were all 555 performed using JMP (JMP Pro 14.0.0, 2018 SAS Institute Inc.) software. General Linear 556 Regression Model (GLM) and One-Way ANOVA followed by Tukey's-HSD were used to 557 determine statistical significance in measurements. The BRIX and yield measurements from 558 grafts were analyzed using JMP Pro software. Significance was tested using ANOVA and 559 19 Tukey's-HSD pairwise comparisons. Leaf sugar and starch data was analyzed using JMP Pro 560 software. Significance was tested using Student's t test. 561 562 SUPPLEMENTAL MATERIAL 563 Figure S1. Diagram of leaf disc sampling and grafting experiment. 564