Silver thiosulfate and Benzyladenine in combination with pruning additively feminizes cassava flowers and modulates transcriptome

Cassava, a tropical storage-root crop, is a major source of food security for millions in the tropics. Cassava breeding however is hindered by the poor development of flowers and female flowers in particular, since flower development is strongly skewed towards male flowers. Our objectives were to test plant growth regulator and pruning treatments for their effectiveness in field conditions in improving flower production and fruit set in cassava. Pruning the fork type branches that arise at the shoot apex immediately below newly formed inflorescences stimulated inflorescence and floral development. The anti-ethylene silver thiosulfate (STS) also increased flower abundance. Both pruning and STS increased flower numbers without influencing sex ratios. In contrast, the cytokinin benzyladenine (BA) feminized flowers without increasing flower abundance. Combining pruning and STS treatments led to an additive increase in flower abundance; with the addition of BA, over 80% of flowers were females. This three-way treatment combination of pruning+STS+BA also led to an increase in fruit development. Transcriptomic analysis of gene expression in tissues of the apical region and developing inflorescence revealed that the enhancement of female flower development by STS+BA was accompanied by the downregulation in of several genes associated with repression of flowering, including Tempranillo 1 (TEM1), GA receptor GID1b, and ABA signaling genes ABI1 and PP2CA. We conclude that treatments with pruning, STS and BA create widespread changes on the network of hormone signaling and regulatory factors beyond ethylene and cytokinin.

6 treatments to limit phytotoxicity. Concentration of BA used was 0.22 mM. Treatments are 143 summarized in Table 1. 144

PGR Experiment II: PGR and pruning effect on female flower development 145
In 2018 and 2019, experiments were conducted using a split-split-split plot design. Each experiment 146 comprised six plots, each of which were split into three subplots each with one of the genotypes. 147 Each genotype subplot was split into 5 PGR treatments (as shown in Table 2 were STS treatment (+STS, -STS) and BA timing, and the interaction between STS treatment and 164 BA timing. Plots were random effects. For Experiment II, sources of variation were PGR treatment, 165 pruning and the interaction between PGR treatment and pruning. Year was analyzed as a fixed effect 166 while plot as a random effect. The emmeans package (Lenth 2019) was used for post-hoc tests. 167 Multiple means comparison was carried out using the Tukey-HSD method. 168 169 2.9 Transcriptomic Analyses 170 7 IITA-TMS-IBA980002 was grown in a greenhouse at the Guterman Bioclimatic laboratory (Cornell 171 University, Ithaca, NY, USA) as described by Hyde et al (2020) and exposed to one of four treatment 172 combinations (a 2 × 2 matrix of treatments): a) either control or pruned at first inflorescence 173 appearance, and b) either a control or PGRs with a combination of petiole-fed STS, and apex-sprayed 174 BA. These were applied to plants at tier 1 of fork branching. Young inflorescence tissue, about 0.25 175 -0.5 cm, comprising the shoot apex and some bracts but excluding fork-type branches were 176 harvested from control, pruned and PGR treated plants. Samples  with similar benefit if the spray was localized to the young folded leaves and region of the shoot 212 apical meristem. However, when these PGRs have been used in preliminary studies on cassava with 213 spray application in the field, their effectiveness has not been clear cut (Setter, personal 214 communication). A new petiole feeding method was developed, similar to that reported by Lin et al. 215 (2011). This method introduced more modest quantities of PGR internally via xylem of the petiole 216 and delivered these substances to the shoot such that phytotoxicity was decreased (Hyde and Setter 217 personal communication). In the current studies we used petiole feeding as a method of STS 218 application in field studies. Further, we investigated whether BA applied as a spray to the immature 219 tissues, affects flower development as a sole treatment or when combined with STS delivered 220 through the petiole.

Transcriptomics 277
To advance our understanding of PGR and pruning effects on flowering regulatory processes, we 278 analyzed gene expression in response to PGR and pruning treatments. For this work, we used 279 treatments which had the largest effect in the field: a) STS+BA without pruning, b) STS+BA with 280 pruning, c) pruning without PGR treatment, and d) control (no PGRs and no pruning). This study 281 was conducted on the model genotype 0002, in a controlled environment green house. 282

Green house phenotype validation 283
To validate treatment responses that were observed in the field with those in a controlled 284 environment greenhouse, we evaluated flowering traits with select treatments in the greenhouse. 285 Similar to the field study, the controls initiated inflorescences, but did not produce any mature 286 flowers, while pruning or STS+BA as sole treatments produced a modest, number of flowers ( Figure  287 11 4a); in the pruning treatment all the flowers were male, but in STS+BA, about 80% were female 288 (Figure 4 b,c). Combining STS+BA with pruning increased total and female flower numbers by 289 about three-fold compared to the PGR-or pruning-only treatments. As with field studies, BA-290 containing treatments increased the number of female flowers and the proportion of flowers that were 291 female (Figure 4 b,c). These findings confirmed a consistent response to treatments under field and 292 greenhouse conditions. 293

Identification of differentially expressed genes and enrichment analysis 294
Transcriptome analysis was conducted for tissues of the shoot apical region and proceeded in three 295 phases (i) examining transcriptome changes due to pruning alone (design = ~pruning), (ii) examining 296 transcriptome changes due to PGR alone (design = ~pgr), (iii) examining transcriptome changes due 297 to the combination of pruning and PGR (design = ~pruning + pgr). This analysis revealed that the 298 majority of the changes in the transcriptome was due the influence of the PGR treatments. Principle 299 component (PC) analysis indicated that expression was not clearly grouped according to pruned 300 versus unpruned treatments but was clearly grouped according to PGR-treated versus PGR untreated 301 samples ( Figure 5a). This grouping was along the first PC axis, which accounted for 54% of the 302 variance. It appeared from the PC analysis that pruning had an intermediate effect in the positive 303 direction along the first principal component whereas PGR had a more substantial effect in the same 304 direction along PC1 axis; the combination of PGR and pruning gave the largest effect in the positive 305 direction of PC1 axis. Analysis of the full model including pruning and PGR revealed that 5440 306 genes were differentially expressed at a 5% false discovery rate (FDR) correction; 2448 genes were 307 up regulated while 2952 genes were down regulated. Functional analysis revealed that in the PGR-308 treated versus controls, PGR-upregulated genes were enriched in pathways involving cell 309 proliferation, cell maintenance, and biosynthetic processes; while PGR-down regulated genes were 310 enriched in pathways involved in plant hormone signal transduction, photosynthesis and degradation 311 metabolism, among others (Figure 6 a,b). Enrichment analysis indicated that this cluster, consisting of 61 genes, was enriched with genes 322 involved in abscisic acid metabolism and response, terpenoid metabolism, abiotic stress response and 323 response to chemicals (Figure 7 b,c). This cluster suggests that regulation associated with the 324 treatments that stimulate flowering was a decrease in stress-associated genes which were expressed at 325 moderate levels in the control apical region. 326 327

Pruning 328
Twenty-one genes were differentially expressed in response to pruning as the principal treatment 329 ( Figure 8a). This set was enriched in genes involved in response to wounding, herbivory and 330 jasmonic acid signaling (Figure 8b). Also upregulated in pruning were genes involved in terpene, 331 lipid and hormone metabolic pathways. Pruning increased expression of these genes in the presence 332 or absence of PGR treatments. Given that the tissues for this analysis were harvested 4 d after 333 pruning, it is not surprising that metabolic and signaling factors involved in wounding response were 334 expressed abundantly 335

Hormone signaling genes 336
We examined the expression profile of 115 hormone signal-transduction genes that were 337 differentially expressed (P adj < 0.05) in our samples (Figure 9; expression data for the full set of 338 select hormones signaling genes is available in Supplementary Figure 4). While many genes differed 339 only modestly between treatments, there was a cluster of genes with relatively high expression in the 340 control and substantially lower expression in the PGR treatments. This cluster included a GA 341 signaling gene (GID1b), and two repressors of ABA signaling (ABI1 and PP2CA) (Figure 9b). We 342 also found an auxin response gene (IAA16), and a JA response gene (TIFY10B) that were expressed 343 at a low level in the control, but they had slightly higher expression in PGR treatments, and the 344 highest expression in the pruned without PGR treatment (Figure 9c). Thus, the expression data 345 indicated that PGR treatments comprising anti-ethylene treatment STS and cytokinin treatment BA 346 influenced expression of genes in hormone signaling pathways other than cytokinin and ethylene 347 13 pathways as shown above, suggesting these hormones have considerable breadth of impact in the 348 network of hormone signaling. 349

Flowering genes 350
Among genes known to be involved in various aspects of flowering, we identified 217 genes that 351 were differentially expressed (P adj ≤0.05) in response to our experimental treatments (Figure 10a). 352 From the two clusters with the largest fold changes, two members of the GA flowering pathway 353 (GID1b, GA2ox2) and the TEM1 gene, a known flowering repressor, had the highest fold change 354 with relatively high expression in controls, and low expression in the PGR treatments (Figure 10 b,c). production and retention (i.e., lifespan) by five-fold relative to the control, leading to a larger number 377 of total flowers. The present study is in line with this as STS alone increased the total number of 378 flowers by two-fold relative to the no STS controls; STS treatment increased production of both 379 female and male flowers, but it did not significantly alter the female to male ratios compared to the 380 no STS control. In addition to effects on flowering, the current study also showed that STS increases 381 fruit numbers (Figure 2, Figure 3). 382 In other plant systems, ethylene is widely recognized as playing a role in fruit development, 383 especially fruit ripening (Pech et al. 2018). Ethylene hastens flower senescence, and anti-ethylene 384 treatments including STS have been developed to increase flower longevity (Serek et al. 2006). 385 Ethylene has also been shown to affect early stages of fruit development and fruit set. In Pea and 386 Arabidopsis, failure to develop fruit in the absence of pollination has been associated with ovary 387 senescence arising from increased ethylene biosynthesis in ovaries (Orzáez and Granell 1997;  until the current study these treatments were not applied together. Current study showed that 427 combining these three factors substantially improved reproductive development in cassava. While 428 STS and pruning acted additively to increase the total number of flowers, BA increased the female to 429 male ratio of flowers. This in turn led to greater fruit development than in each of the factors applied 430 singly (Figure 3 and 4). The largest fold changes in the transcriptome in response to STS+BA treatments were the down 435 regulation of genes which were relative to their expression in the control. Among the cluster of such 436 genes with the largest fold change (Figure 7), pathway enrichment analysis indicated that pathways 437 16 related to stress were over-represented compared to their frequency in the genome. These and other 438 examples of signaling by PGRs in our study will be further discussed below. 439 Genes responding to pruning as main treatment on the other hand were upregulated relative to 440 treatments with no pruning and were enriched in processes generally related to wounding (Figure 8) 441 (Reymond et al. 2000). This is sensible since pruning involves the excision of young fork-type 442 branches and can thus be perceived as a type of wounding response by the plant. These genes,