Transcripts switched off at the stop of phloem unloading highlight the energy efficiency of sugar import in the ripening V. vinifera fruit

Transcriptomic changes at the cessation of sugar accumulation in the pericarp of Vitis vinifera were addressed on single berries re-synchronised according to their individual growth patterns. The net rates of water, sugars and K+ accumulation inferred from individual growth and solute concentration confirmed that these inflows stopped simultaneously in the ripe berry, while the small amount of malic acid remaining at this stage was still being oxidised at low rate. Re-synchronised individual berries displayed negligible variations in gene expression among triplicates. RNA-seq studies revealed sharp reprogramming of cell-wall enzymes and structural proteins at the stop of phloem unloading, associated with an 80% repression of multiple sugar transporters and aquaporins on the plasma or tonoplast membranes, with the noticeable exception of H+/sugar symporters, which were rather weakly and constitutively expressed. This was verified in three genotypes placed in contrasted thermo-hydric conditions. The prevalence of SWEET suggests that electrogenic transporters would play a minor role on the plasma membranes of SE/CC complex and the one of the flesh, while sucrose/H+ exchangers dominate on its tonoplast. Cis-regulatory elements present in their promoters allowed to sort these transporters in different groups, also including specific TIPs and PIPs paralogs, and cohorts of cell wall-related genes. Together with simple thermodynamic considerations, these results lead to propose that H+/sugar exchangers at the tonoplast, associated with a considerably acidic vacuolar pH, may exhaust cytosolic sugars in the flesh and alleviate the need for supplementary energisation of sugar transport at the plasma membrane.

S2a). They were further grouped according to their expression profiles (Fig. S3), and the cluster 180 number was restricted to four upon visual analysis. Cluster A and B included the majority of the 181 genes increasing or decreasing in expression from G to S (Fig. 3a). The genes commonly 182 8 Other important differentially expressed genes (DEGs) could be identified in Syrah and 226 only one microvine due to discrepancies at P stage. Since the variance among P stage samples 227 was certainly exacerbated by abrupt changes in physiology and gene expression at phloem arrest, 228 worsened by sampling delay, data were tested in pairwise mode analysing stage S versus stage 229

G. 230
A total number of 6585, 6453 and 5621 DEGs were identified in Syrah, MV032, and 231 MV102 ( Table S2b). The commonly modulated genes increased to 899 up-regulated genes and 232 1155 down-regulated ones, yielding to 2054 DEGs (Table S3). Other 411 genes were modulated 233 in the three genotypes but with a discordant gene trajectory. Compared to the previous analysis, 234 this second comparison gave a more comprehensive list of commonly modulated DEGs (before 235 they were 1448). These two lists shared 809 genes. In this new DEGs list, there were down-236 regulated genes like VviTMT2, VviHT2, VviGIN1, VviH + ATPase, VviEXP19 and other cell wall-237 related genes (Table S3). 238 239

Rates of gene expression: ranking genes by transcription priority 240
Gene expression burdens cells by consuming resources and energy for transcription, 241 translation, and the synthesis and degradation of hydrophobic proteins like transporters is 242 particularly expensive 29 . In this respect, the transcription of transporters must stop very fast once 243 they are no longer required. As a proxy for energy cost, we merged and ranked the two lists of 244 differentially expressed genes according to their absolute change in expression (CPM, i.e. 245 uncorrected for transcript length) during the stop of water, sugar and K + influx in the berry (S 246 versus G), as summarised in Table 1 (full list in Table S3). This straightforward procedure 247 proved extremely efficient in the detection of channels and transporters virtually turned off 248 simultaneously with phloem together with cell wall-related genes. Amazingly, in Syrah, among 249 solute transporters, VviHT6 exhibited the largest decay in expression from G to S stages with 250 80% inhibition, indicating tight synchronism with phloem arrest. This was also verified for 251 VviSWEET10, VviPIP1.3 and VviPIP2.5, which decreased by 90% and 83%. Moreover, a pectate 9 (+244%), metallothionein (+213%), and polyubiquitin (+73%) appeared as the most induced 255 ones. 256 257

Promoter analysis 258
Sixty selected genes related to water, sugar, and growth down-regulated at the stop of 259 phloem were distributed in different regulatory networks when clustered according to the Z-260 score 30 which accounts for length, number and position of their promoter motifs (Fig. 4) For the first time, non-destructive monitoring of single berry growth allowed us to sample 286 synchronised fruits within an original population marked by a significant delay in the 287 development of individual fruits. Sorting berries according to their net growth (or water import) 288 rate warranted that berries still growing and importing water were not mixed with shrivelling 289 (over-ripe) ones. Real-time individual growth monitoring outperformed our previous sampling 290 procedure which just relied on solute concentration 4,17 upon adding the lacking kinetic 291 dimension. Therefore, the simultaneous use of four distinct variables (malate, glucose, fructose 292 and volume) empowered us to discriminate individual berry stages and obtain a recognizable 293 transcriptomic signature. We showed here that the net imports of water, sugar, and K + 294 simultaneously stop when the ripening berry reaches its maximum volume, which confirms that 295 sugar accumulation at constant volume indicating intense xylem back-flow during late 296 ripening 3,16 resulted from an averaging artefact on asynchronous samples 4 . This was verified in 297 contrasted thermo-hydric conditions (field -higher temperature and water demand, versus 298 greenhouse -non-limiting water supply) and when phloem unloading arrested at 0.8 M of sugar 299 in microvines (as already observed for these genotypes 33 ) and at 1.2 M in Syrah. At phloem 300 arrest, with a malate concentration below 100 mEq (39 in Syrah, 48 and 93 mEq in MV032 and 301 MV102), the berry already consumed between 60 and 80% of the malic acid accumulated before 302 softening. These differences in acidity that would principally originate from the green stage 303 accumulation period 27 did not apparently interfere with phloem arrest. 304 Our study, performed in two environments and in three genotypes displaying different 305 ripening features, reports for the first time robust gene trajectories associated with the end of the 306 sugar and water accumulation processes, with the most intensively down-regulated genes related 307 to cell wall modification (arrest of growth), aquaporins and sugar transporters (arrest of phloem 308 unloading). The prevalence of plasma-membrane channels and transporters genes virtually 309 turned off at phloem arrest re-emphasizes the central role of the apoplasmic pathway in ripening 310 berry. It clearly illustrates that a strategy addressing the single berry level can lead to more 311 comprehensive insights on fruit developmental biology than random samplings. 312

313
The arrest of growth is modulated by the down-regulation of cell wall associated genes 314 The large cohort of cell wall related genes being repressed at growth arrest yields an 315 extremely dynamic image of the rearrangements accompanying cell wall extension and cellular arrest where many genes were down-regulated. The present single berry pericarp profiling 318 greatly affirms the connection between aquaporins, cell wall and sugar transport. No obvious 319 marker of cell death was detected here, which is often described later in the shrivelling process 320 of Syrah berries 34 . 321 Here we reported that genes involved in cellulose metabolism (15 cellulose synthases), 322 hemicellulose metabolism (4 endo-1,4-beta-glucanases, 7 xyloglucan endo-transglucosylase / 323 hydrolase proteins), pectin metabolism (4 pectate lyase, 4 polygalacturonases, 6 ß-galactosidases, 324 6 fasciclin-like arabinogalactan protein), and expansins genes (9) were strongly down-regulated 325 at the arrest of phloem (Table S3). In particular, we emphasise that precisely such expansins 326 (VviEXPA19, VviEXLA1, VviEXPB4, VviEXPA14) previously linked to growth and cellular 327 expansion during the second growth phases both in the flesh and in the skin 35 were down-328 regulated concomitantly with growth cessation. A link between these expansins and aquaporins 329 has been envisaged in a pre-genomic study reporting a down-regulating trend two or three weeks 330 before the attainment of maximum berry size on asynchronous berry samples 36 . This so-called 331 "unexpected" link was recently confirmed through a gene coexpression network analysis 37 . 332

333
The arrest of phloem is linked with strongly down-regulated aquaporins 334 Aquaporins are transmembrane channels facilitating the transport of water and small 335 solutes from cell to cell and between cell compartments. Noticeably, seven of the ten plasma 336 membrane intrinsic proteins (PIPs) identified in V. vinifera 37 were drastically inhibited (Table 1,  337   Table S3)  Petit Manseng, which was not followed by a marked decrease at phloem arrest 41 . Our RNA-seq 375 data show that VviSWEET15 displays a similar expression level than VviSWEET10 during the 376 active phloem unloading without being inhibited at phloem stop (Table S4)  The probable PM H + hexose symporter VviHT2 (Vitvi18g00397 -LOC100232961) 44 also 404 exhibited a considerable increase during ripening, followed by a 35% decrease in two weeks 53 . 405 Synchronised berries show here that more than 50% inhibition occurs at growth arrest, reaching 406 after that 84% decrease in Syrah. VviHT2 was already associated with the induction of VviHT6 at 407 veraison both in flesh and skin 54 and in ABA, and GA 3 treated berry 45,55 , but data were lacking 408 regarding its inhibition at ripe stage. To the best of our knowledge, VviHT2 localisation is still 409 unknown. Unfortunately, more information is available for VviHT1, but it was not commonly 410 modulated in the genotypes studied here. 411 accumulates one hexose 4 . Thus, a phloem unloading pathway passing through H + -coupled sugar 415 transporters on the three membrane interfaces from SE/CC complex to flesh vacuoles, combined 416 with the functioning of apoplasmic invertase, would waste within 50% of cellular ATP for H + 417 recycling, in a pure loss, as the sucrose gradient between first and terminal compartments is 418 thermodynamically favourable in V. vinifera. Present results (DEGs, promoter analysis) lead to 419 propose a more efficient design of sugar import, which refines the previous ones 4,56 upon 420 integrating key genes/functions observed in this study (Fig. 5). 421 VviSWEET10, located on the SE/CC plasma membrane, would be the major player in the 422 export of sucrose from the phloem into flesh apoplasm. Sucrose hydrolysis in the apoplasm 423 would require its resynthesis in flesh cells to allow sucrose vacuolar transport by TST (see 424 discussion above) which could be compatible with the huge induction of SPS during ripening 23,56 425 and with SPS, SuSy, AI, NI in vitro activities 41 in five thousand excess, at least, with sugar 426 accumulation rate. However, except for AI, the corresponding genes were not strongly inhibited 427 at phloem arrest, which may indicate that sucrose resynthesis could be also involved in 428 intercellular sugar exchanges needed for respiration in fruit periphery, which continues after 429 phloem arrest. In fact, Vitvi07g00353 and Vitvi11g00542 annotated as sucrose synthase and 430 sucrose phosphate synthase (Table S4) were rather constant during the three stages in the three 431 genotypes. 432 Sugar transport by SWEETs is energetically silent and the sink strength for sugar 433 accumulation is driven by the key electrogenic antiporters (proton-coupled, i.e. energy 434 consuming) VviHT6 and VviTMT2 that transport sucrose across the tonoplast. At the beginning 435 of ripening, the vacuolar acidity within 400 and 450 mEq (Table S1) at pH 2.7, provides a 436 considerable proton motive force for the transport of sucrose into the vacuole, as long as the 437 counterion malate is available. In tandem with vacuolar invertase, this system is susceptible to 438 restrict cytosolic sucrose below the nanomolar range. Although less dramatic, a similar 439 conclusion holds for hexose, in case VviHT6 and VviTMT2 do not transport sucrose (Table S5). 440 Once in the cytoplasm, the exchanged malic acid will serve as a substrate for respiration or 441 gluconeogenesis, according to the reactions: 2 H + + malate + 3 O 2 = 4 CO 2 + 3 H 2 O and 2 malate 442 + 4 H + → sucrose at the tonoplast (Table S1) are scavenged from the cytoplasm through these reactions, 445 but the remaining ones need to be pumped back in the vacuole at the expense of a progressive 446 activation of glycolysis, aerobic fermentation (ADH2), vacuolar H + /ATPase and H + /PPiase 57 . It 447 is noticeable that also VviADH2 is inhibited at phloem arrest, along with vacuolar invertase. 448 Results on the respiration of single grapevine berry, which provide strong arguments in this 449 direction, will be detailed in a next publication. The only PM H + symporter encountered here 450 (VviHT2) displayed the highest expression in Syrah, which also needed more ATP at the 451 tonoplast (Table S1), when compared to microvines. This suggests that energy-requiring H + 452 symporters would be activated on the plasma membrane to increase sink strength in 453 unfavourable environmental conditions, while VviSWEET15 would allow the bulk of sugar 454 import on flesh cells, even though further work is needed to precise its substrate specificity. 455 The energetically-optimised model proposed here alleviates the opening of an AKT2-like 456 channel that would compensate the energy deficit during phloem unloading 5 , fully inherent to the 457 sugar/H + symporters dogma, that can be discussed on thermodynamic and transcriptional 458 grounds in grape berry. Such a mechanism would require a simultaneous activation of 459 quantitatively similar flows of sugars and K + in ripening berries, which must be clearly excluded Single berry growth and development were monitored through biweekly pictures of 483 selected clusters (up to 8 for each genotype) starting a few days before softening day until two 484 weeks after maximum berry growth. Pictures were taken using a Lumix FZ100 camera 485 (Panasonic), keeping the focal range and cluster to camera lens distance (30cm) constant. The 486 volume increment of selected berries was calculated by analysing the pictures with ImageJ 60 . The 487 software, after calibrating the images using the 1 cm scale present in each of them, automatically 488 counted the pixels enclosed in each targeted berry area, measured as an ellipsoid. The estimated 489 berry volume was then calculated accordingly using the radius of the previously calculated area. 490 To eliminate the intra-and inter-cultivars variability in berry sizes and compare the changes in 491 volume among berries of different clusters, vines, and genotypes, each berry growth profile was 492 normalised to the softening volume, set to 1. 493 According to their own volume changes, 10 to 15 berries were sampled for each genotype 494 at different dates during the ripening growing phase (stage G), closest to the berry growth peak 495 (stage P) and two weeks after the maximum growth (stage S) in the shrivelling phase. 496 Concerning stage P, one must notice that a-posteriori it was found that phloem and, obviously, 497 growth, were already blocked at this stage, which should, technically speaking, be called P+1. 498 Berries were sampled at the same time of the day, between 9 am to 11 am, to avoid circadian 499 cycle influences. Berries without pedicel were rapidly deseeded before freezing (1-2 min after 500 harvest) in liquid N 2 and stored at -80 until further analysis. Single berries were ground to a fine 501 powder under liquid nitrogen using a ball mixer mill (Retsch MM400).  (Table S6). Aligned 523 reads were counted using the last available annotation VCost.v3 with HTSeq-count (version 524 0.9.1) with the "nonunique all" flag 64 . Genes were filtered by applying an RPKM>1 cut-off in at 525 least one experimental condition (Table S4), and the variance stabilising transformation was 526 applied for data visualisation. 527 Genes (TMM normalised) were tested for multi time-series significance for finding genes 528 with significant (P<0.05) temporal expression changes using the R package MaSigPro 65 , with 529 Syrah dataset used as a control. A quadratic regression model (rqs=0.7) was applied. Significant 530 genes over time were clustered using STEM (short time-series expression miner) software 66 with 531 the maximum number of model profiles set to 24. Filtering or normalisation was not applied by 532 the software since input data were already filtered and normalised, but a mean-centred scale was 533 applied to overcome cluster mismatches due to different expression levels. Genes showing the 534 same profiles pattern among the significant clusters were grouped as follows: cluster A = cluster 535 Promoter sequences (1.5 kb upstream of TSS) were extracted using bedtools from the 543 12X2 genome assembly 62 . Gene correspondence file at 544 https://urgi.versailles.inra.fr/content/download/5723/43038/file/list_genes_vitis_correspondences 545 V3_1.xlsx and in the related gff3 shows that the 5'UTR was frequently lost in the VCost.v3 546 annotation. Therefore, we decided to take the longest sequence between VCost.v3 and NCBI 547 (Vitvi and LOC references, respectively) as the most probable TSS. Cis-regulatory elements 548    The figure is divided into three panels: early ripening, late ripening, and shrivelling with the stop 822 of phloem between the two latter (marked by a black circle). Genes written in red or blue italics 823 refer to those respectively expressed or repressed at specific stages, those in black are Table 1. List of thirty-five selected genes related to cell growth, water and sugar transport sorted 828 by absolute difference in Syrah G versus P. Gene ID is reported as VCost.V3, V1 and RefSeq 829 code. Gene expression is expressed as count per million (CPM). 830 Table 1 List of thirty-five selected genes related to cell growth, water and sugar transport sorted by absolute difference in Syrah G versus P. Gene ID is reported as VCost.V3 and RefSeq code. Gene expression is expressed as count per million (CPM).