Spatial control of potato tuberization by the TCP transcription factor BRANCHED1b

The control of carbon allocation, storage and usage is critical for plant growth and development and is exploited for both crop food production and CO2 capture. Potato tubers are natural carbon reserves in the form of starch that have evolved to allow propagation and survival over winter. They form from stolons, below ground, where they are protected from adverse environmental conditions and animal foraging. We show that BRANCHED1b (BRC1b) acts as a tuberization repressor in aerial axillary buds, which prevents buds from competing in sink strength with stolons. BRC1b loss of function leads to ectopic production of aerial tubers and reduced underground tuberization. In aerial axillary buds, BRC1b promotes dormancy, abscisic acid responses and a reduced number of plasmodesmata. This limits sucrose accumulation and access of the tuberigen protein SP6A. BRC1b also directly interacts with SP6A and blocks its tuber-inducing activity in aerial nodes. Altogether, these actions help promote tuberization underground. In potato, the TCP transcription factor BRANCHED1b represses aerial tuber formation in the axils of the leaves. It functions through limiting the number of plasmodesmata, reducing sucrose levels and repressing the tuberigen protein SP6A.

A central question in plant biology is how carbon partitioning is regulated in response to changing environmental conditions, and how this control is integrated with developmental programs. Plant meristems (groups of undifferentiated, actively dividing cells) are sink organs that use sucrose for respiration, growth and development. Under low sucrose availability, they remain quiescent or dormant. Shoot axillary meristems (meristems located in the leaf axils) can give rise to aerial branches with orthotropic growth that produce leaves, inflorescences and flowers. They can also generate stolons, specialized branches that grow horizontally on the soil surface or below ground and have the potential to generate new clonal plants. Notably, stolons differ from regular branches in their responses to external stimuli, such as day length (see below).
Solanum tuberosum (potato) is one of the best-studied stolon-forming species, as these specialized branches develop energy-rich, starch-accumulating tubers, which constitute a large fraction of the dietary intake for many people worldwide. In the potato landrace ssp. andigena, the stolons become strong sinks when winter approaches and form tubers at their sub-apical region. Tubers overwinter buried in soil, where they are protected from adverse environmental conditions and animal foraging, to generate entirely new plants in spring. Tuberization is controlled by the FLOWERING LOCUS T (FT)-like SELF-PRUNING 6A (SP6A) gene 1 . In potato ssp. andigena, the activation of this gene is fully dependent on short days (SD). In long days (LD), the CONSTANS-like1 (COL1) transcription factor represses SP6A expression through the activation of another FT-like gene, SP5G 2,3 . In SD, the lack of SP5G allows SP6A activation in leaves, and the SP6A protein is then transported via phloem from leaves to stolons 1 . At the stolon apices, SP6A forms a tuberigen activation complex that promotes tuber formation 1,4 .
Unlike stolons, aerial branches and branch primordia (that is, axillary buds) do not undergo tuberization in SD; instead, they remain dormant. The molecular mechanisms driving these distinct responses remain unclear. In most plant species, SD promote shoot bud dormancy, which involves the induction of abscisic acid (ABA) responses and a carbon-starvation syndrome 5,6 . In perennials, SD-induced ABA signalling leads to a blockage of symplasmic intercellular communication by the closure of plasmodesmata (PD), which prevents growth-promoting signals (such as FT-like proteins 7 ) from entering the shoot apex 8,9 . Bud dormancy is also controlled by the widely conserved TCP transcription factor BRANCHED1 (BRC1) 10 . Arabidopsis BRC1, expressed in axillary buds, activates ABA signalling and downregulates genes associated with bud activation, including cell division and protein synthesis genes [11][12][13] . In poplar, SD-induced BRC1-like factors inhibit shoot apical bud growth by binding the FT-like protein FT2 and antagonizing its growth-promoting activity 14 . Potato BRC1a, one of the two BRC1 paralogues, also promotes bud dormancy in both aerial and stolon axillary buds 15,16 . The role of the second BRC1 paralogue, BRC1b, remained to be studied.
We have now found that BRC1b acts as a multimodal tuberization repressor in aerial axillary buds: it promotes bud dormancy and ABA-related responses, and causes a reduction in PD number. This in turn limits sucrose accumulation and SP6A access to axillary buds. In addition, the BRC1b factor interacts directly with the tuber-promoting SP6A protein, thus inhibiting its activity in aerial nodes. These combined activities of BRC1b prevent aerial axillary buds from becoming strong sugar sinks in conflict with stolons, which may facilitate sucrose allocation and SP6A accumulation in stolon tips. All these actions help restrict tuberization to stolons in the SD of winter as a critical plant survival strategy. The aerial tuber phenotype reported in lines overexpressing the microRNA 156 (miR156), an upstream regulator of BRC1b, can also be explained by a strong downregulation of BRC1b mRNA levels.

BRC1b is expressed in axillary buds and developing stolons.
To elucidate the role of potato BRC1b, we first studied its expression patterns. Quantitative real-time PCR (qPCR) analyses in different tissues (Fig. 1a) revealed that, like BRC1a, BRC1b mRNA accumulated in aerial and stolon axillary buds. However, unlike BRC1a, it also accumulated, at high levels, in developing stolons. Potato ssp. andigena and cv. Désirée transgenic lines carrying the BRC1b promoter fused to the β-GLUCURONIDASE (GUS) coding sequence (BRC1bpro-GUS) and RNA in situ hybridizations using an antisense BRC1b-mRNA-specific probe confirmed that BRC1b was active in aerial, stolon and tuber axillary buds ( Fig. 1b-i,k,l and Supplementary Fig. 1a-e,l). GUS activity and BRC1b mRNA were also detectable at the shoot apex and in the vasculature of growing stolons and tubers (Fig. 1d-j and Supplementary Fig. 1d,e,k,m) and flower meristems ( Supplementary Fig. 1n). In meristems, BRC1b expression was restricted to the inner layers. In the leaves of ssp. andigena, BRC1b expression was initially widely distributed but became progressively restricted to leaf margins and vasculature in more mature leaves ( Supplementary Fig. 1f-i), where it was also observed in Désirée ( Supplementary Fig. 1j). These results support a potential role of BRC1b in the development of aerial and underground buds and during stolon, flower and leaf development. and BRC1b relative mRNA levels in different organs quantified by qPCR. n = 3 biologically independent samples examined. The error bars indicate ±1 s.e.m. The letters above the bars indicate significant differences among means (one-way analysis of variance (ANOVA) plus Tukey's honestly significant difference (hSD), P = 0.0422). b-h, GUS histological activity in transgenic plants carrying a 1.7-kb BRC1bpro-GUS construct. The images show an aerial axillary bud (b), a stolon axillary bud (c), a stage-1 stolon (d), a close-up of a stage-1 stolon apex (e), a stage-3 developing tuber (f), a close-up of a tuber apical region (g) (the apical (asterisk) and axillary buds (arrowheads) show GUS activity) and a tuber section showing GUS signal in the apical and axillary buds and vasculature (h). The tubers in g and h are from mature senescent potato plants. i-l, BRC1b RNA in situ hybridizations with a probe complementary to BRC1b mRNA. The images show a longitudinal section of a stage-4 developing tuber (i), a shoot apical meristem close-up of the developing tuber shown in i (j), the axillary bud of a stage-1 stolon (k) and an aerial axillary bud (l). The arrowheads indicate axillary buds, and the asterisks indicate shoot apices; v, vascular tissue. Scale bars, 1 mm. n = 3 biologically independent samples (from three different plants) were hybridized and examined, with similar results as shown in b-l. Fig. 2a), in which BRC1b mRNA levels were strongly reduced relative to the wild type (WT) (Supplementary Fig. 2b). We compared the phenotypes of RNAi lines with those of the WT and of plants bearing the empty vector, in non-inducing (LD, 16 h day/8 h night) and tuber-inducing (SD, 8 h day/16 h night) conditions. In LD, some lines developed a few more aerial branches than the controls ( Supplementary Fig. 2c), and all had fewer, shorter and less branched stolons ( Supplementary Fig. 2d,e,f). After six weeks in SD, RNAi lines developed more aerial branches ( Supplementary  Fig. 2g), whereas their stolon numbers remained lower than in the controls ( Supplementary Fig. 2h,i). In addition, RNAi lines had simpler, more convex leaves than the controls ( Supplementary Fig. 3).
The most dramatic phenotypes were seen for tuber formation. Under non-inductive LD, some RNAi lines developed underground tubers, a phenotype rarely observed in the controls ( Fig. 2a and Supplementary Fig. 4a,f), and line 10 also displayed aerial tubers ( Fig. 2d and Supplementary Figs. 4f and 5a). The aerial nodes of some RNAi plants were enlarged compared with the controls and seemed to accumulate starch ( Fig. 2e and Supplementary Fig. 5d,g). Under inductive SD, all control plants developed underground tubers at the tips of stolons ( Fig. 2b and Supplementary Fig. 4b,c,f). In contrast, some RNAi individuals failed to tuberize ( Supplementary Fig. 4f), and the others produced fewer underground tubers ( Fig. 2b) with lower yield than the controls ( Supplementary Fig. 4d). Remarkably, all RNAi lines developed aerial tubers at axillary buds after four weeks in SD, a phenotype not observed in the controls (Fig. 2c,f,g and Supplementary Figs. 4c,e,f and 5b,c). Some RNAi tubers developed secondary tubers from tuber eyes (Fig. 2g). Aerial tubers were observed in the uppermost and, less frequently, in the lowermost nodes ( Supplementary Fig. 4e). Ssp. andigena brc1b CRISPR mutants also developed aerial tubers (Supplementary Figs. 5e,f,h,i and 6). We also confirmed the aerial tuber phenotypes in BRC1b RNAi lines generated in the day-length-independent cv. Désirée ( Supplementary Fig. 7). These results indicate that potato . Results are shown for the WT, EMPTY (control plants carrying an empty RNAi vector) and three BRC1b RNAi lines (5, 10 and 19). n = 10 biologically independent samples examined. The error bars indicate the highest and lowest values, the box indicates the middle 50% and the centre line indicates the median. The letters designate significant differences among means (one-way ANOVA plus Tukey's hSD, P = 0.337 (a), 0.00000108 (b) and 0.000166 (c)). d,e, Aerial tuber (d) and swollen node (e) of BRC1b RNAi plants grown in LD. f,g, Aerial tubers of BRC1b RNAi plants grown in LD followed by six weeks in SD. Scale bars, 1 cm.
BRC1b prevents aerial tuberization in SD. It also controls leaf development and (to a limited extent) shoot branching.
BRC1b RNAi aerial buds have high levels of sucrose in SD. We next investigated potential alterations in the source/sink status of RNAi plants. For this, we quantified sucrose content in leaves and aerial buds in LD and after exposing plants to a week of SD (7SD), in the WT and RNAi. Leaves are the major source of sucrose, which is phloem-transported to sink tissues 17 . In the leaves of both genotypes, sucrose levels were ~2.5× higher in LD than in SD, indicating that leaves have a stronger photosynthetic activity and/or lower sucrose export in LD than in SD conditions (Fig. 3a). WT and RNAi leaves displayed similar sucrose contents, indicating that BRC1b does not control photoassimilate biosynthesis. Conversely, axillary buds (sink organs) of both genotypes accumulated greater amounts of sucrose in SD than in LD (Fig. 3b,c), which may reflect a movement of sucrose from leaves to buds in SD. However, sucrose levels were significantly higher in RNAi than in the WT in SD, which suggests that RNAi buds have an increased ability to import sucrose in SD.

BRC1b
RNAi aerial buds display reduced dormancy. To understand in more detail the molecular mechanisms leading to the formation of aerial tubers in the RNAi lines, we studied the global transcriptional responses underlying this process. For this, we compared the transcriptomes of upper buds in RNAi and WT plants grown in LD, and after 2SD, 7SD and 14SD, using RNA sequencing (RNA-seq; Fig. 4a and Supplementary Data 1). Next, to investigate the responses of the gene sets of interest, we performed a gene set enrichment analysis (GSEA 18 ). From 2SD onwards, WT buds displayed responses typical of buds entering dormancy: a significant upregulation of gene sets of bud dormancy markers of Arabidopsis 5 and potato 19 ; the activation of gene sets of sugar-repressed genes, typical of a carbon starvation syndrome 5,20,21 ; and the downregulation of gene sets related to protein synthesis, cell division and bud activation (Fig. 4b, columns 5-7). These results were in agreement with the observation that SD promote bud dormancy in perennials 22 . In contrast, the buds of RNAi plants displayed significantly reduced dormancy-related responses relative to the WT, even in LD, which suggests that RNAi buds were partially active (Fig. 4b, columns 1-4).
ABA signalling has been associated with bud dormancy, reported to be induced in SD 14,23,24 and controlled by BRC1 genes in other species 13,25,26 . We therefore analysed a gene set of ABA-response markers in the WT and RNAi 5,27 . ABA responses were reduced in RNAi relative to the WT from 2SD onwards ( Fig. 4b and Supplementary  Fig. 8, columns 2-4). Consistently, direct quantifications of ABA revealed significantly lower hormone concentrations in RNAi than in WT buds (Fig. 4c).

BRC1b
RNAi aerial buds express tuberization markers in SD. We next investigated the expression of early tuberization marker genes during the time course. SP6A, a gene essential for tuber formation 1 , was induced in RNAi from 7SD (Fig. 4d, columns 9 and 10). SWEET11, which encodes a sucrose channel whose expression is increased in response to SP6A during tuberization 28 , was also upregulated in RNAi at 7SD (Fig. 4d, column 9). GIBBERELLIN 2-OXIDASE1 (Ga2OX1), involved in gibberellin catabolism and upregulated in stolons before swelling 1,29 , was induced in RNAi buds from 2SD (Fig. 4d, columns 8-10). In contrast, the tuberization marker BEL5 (ref. 30 ) was not differentially expressed in RNAi buds (Fig. 4d), in agreement with evidence suggesting that BEL5 is an upstream transcriptional regulator of the BRC1-like genes 31,32 . In summary, aerial axillary buds lacking BRC1b display an ectopic induction of tuberization pathway genes in aerial axillary buds as early as at 2SD.

BRC1b
RNAi aerial buds accumulate SP6A mRNA in SD. In the WT, the central tuberization regulator SP6A is expressed in leaves in SD, and the SP6A protein is then phloem-transported from leaves to stolons, where it triggers tuberization 1 . A positive feedback loop by the SP6A protein enhances SP6A expression in stolons 1 .
We studied in more detail the expression patterns of SP6A in WT and RNAi leaves and aerial buds, under LD and SD. In WT leaves, SP6A was induced at 2SD, and its expression continued to increase at 7SD, 14SD and 21SD (Fig. 4e). SP6A mRNA levels were similar in RNAi, which suggests that BRC1b was not involved in the transcriptional regulation of SP6A in leaves. In WT aerial buds, SP6A mRNA was almost undetectable in LD and at 2SD, 7SD and 14SD. In contrast, RNAi buds displayed significant amounts of SP6A transcripts from 7SD onwards. At 21SD, SP6A mRNA levels were around four times higher than in the WT (Fig. 4f). The induction of SP6A in RNAi buds from 7SD (subsequent to that in leaves at 2SD) could be due to movement of the SP6A protein from leaves to buds, and to positive regulation of the SP6A gene by its own protein.
BRC1b RNAi buds have more PD than WT buds in SD. The abnormal accumulation of sucrose and SP6A in RNAi buds in SD could reflect an increased movement of both molecules into the buds. To test whether this was due to increased symplasmic trafficking across PD, we quantified PD and studied PD-related gene expression in WT and RNAi buds.
To determine PD number, we analysed ultrathin (70 nm) plastic sections of WT and RNAi aerial buds at 7SD, using transmission electron microscopy (Fig. 5). Cells of RNAi buds (Fig. 5b,d-f) had a significantly higher number of PD than the WT (Fig. 5a,c,f). They often displayed twinned PD (Fig. 5b) and showed more clear connections with the endoplasmic reticulum (Fig. 5d).
We then studied, in our RNA-seq dataset (Supplementary Data 1), whether PD-related genes were differentially expressed between the WT and RNAi. In particular, we studied the expression levels of the potato orthologues of 115 Arabidopsis genes encoding PD-associated proteins identified by refined semi-quantitative proteomic analysis 33 . Several members of the gene family multiple C2 domains and transmembrane region proteins (MCTPs) were differentially induced in RNAi ( Supplementary Fig. 8, columns 2-4, and Supplementary Data 2). Interestingly, a member of this family (MCTP1 or FT INTERACTING PROTEIN 1) mediates the intercellular transport of FT and of RICE FLOWERING LOCUS T1 in Arabidopsis and rice, respectively 34,35 . RNAi buds also displayed relative downregulation of CALLOSE SYNTHASE3, a key gene involved in callose biosynthesis and deposition, and upregulation of genes encoding β-1,3-GLUCANASES, enzymes involved in callose degradation ( Supplementary Fig. 8, column 2, and Supplementary Data 2). This could result in reduced callose deposition in PD and thus increased PD pore size and permeability. Altogether, these results suggest that trafficking through PD may be more intense in RNAi buds.

Symplasmic movement is enhanced in RNAi aerial buds.
To evaluate the symplasmic movement of molecules across cells in RNAi buds, we performed dye-loading studies using the mobile compound 5,6-carboxyfluorescein diacetate (CFDA). CFDA is a membrane-permeable molecule that, upon entering cells, is converted into the fluorescent, membrane-impermeable carboxyfluorescein (CF) tracer, whose movement is exclusively symplasmic 36,37 . WT and RNAi plants were grown for 14SD, and stems of those plants, bearing the upper axillary buds, were fed with a CFDA solution. CF fluorescence was traced for several hours and quantified in axillary buds. RNAi buds contained significantly more fluorescence signal than WT buds ( Fig. 3d-h), suggesting that movement was enhanced in the RNAi lines.
BRC1b interacts with the tuberigen protein SP6A. In Arabidopsis and hybrid aspen, the BRC1/BRC1-2 proteins interact directly with FT/FT2 to prevent their flowering-/apical-growth-promoting activities, respectively 14,38 . We hypothesized that BRC1b could similarly bind and antagonize SP6A in potato aerial buds. To test this, we conducted yeast two-hybrid (Y2H) assays between the proteins BRC1b, SP6A and its paralogue SP5G. BRC1b interacted with both SP6A and SP5G, and the TCP domain was necessary for this    interaction (Fig. 6a). The BRC1b paralogue, BRC1a L , also interacted with both proteins ( Supplementary Fig. 9), but the binding affinity of BRC1a L -SP6A seemed lower than that of BRC1b-SP6A. Co-immunoprecipitation assays confirmed these interactions in planta. Constructs driving the expression of BRC1b-HA or BRC1a L -HA were co-agroinfiltrated in Nicotiana leaves with constructs expressing either SP6A-GFP or SP5G-GFP, and the presence of these proteins was analysed after pull-down of protein extracts with anti-HA magnetic beads. These assays confirmed a strong interaction of BRC1b with SP6A (Fig. 6b), whereas interactions were much weaker for BRC1b-SP5G and BRC1a L -SP6A, and no interaction was observed for BRC1a L and SP5G.
These interactions were further confirmed by acceptor photobleaching fluorescence resonance energy transfer (APB-FRET) assays of fluorescent protein fusions of BRC1b co-expressed with SP6A or SP5G in Nicotiana leaves. The E FRET (percentage of GFP expression change) values for BRC1b-mCherry/SP6A-GFP were almost as high as those for BRC1b-mCherry/BRC1b-GFP homodimers and significantly higher than those of BRC1b-mCherry/ SP5G-GFP (Fig. 6c).
These findings demonstrate that BRC1b specifically interacts with SP6A and probably antagonizes (via protein-protein interactions) its tuber-inducing activity in axillary buds, where BRC1b accumulates in the WT. miR156 negatively controls BRC1b. Lastly, we investigated the relationship between BRC1b and other genes whose altered expression cause aerial tuber phenotypes. Tomato lines overexpressing the cytokinin biosynthesis gene LONELY GUY1 (LOG1) display aerial tubers 39 . However, our transcriptomic data showed no induction of any LOG gene in the RNAi buds ( Supplementary Fig. 8, columns 2-4), suggesting that the phenotype observed in tomato is caused by a genetic pathway unrelated to or upstream of BRC1b.
Likewise, altered expression of the chromatin-remodelling-related genes MULTICOPY SUPPRESSORS OF IRA1 (MSI1) and BMI1 (both of which regulate miR156) cause similar phenotypes 40 . The expression of these genes is not significantly altered in RNAi buds ( Supplementary Fig. 8).
As the expression of Teosinte Branched1 (the monocot orthologue of BRC1) is controlled by the miR156/SPL module in rice 42 , we further investigated whether the overexpression of miR156 could indirectly lead to reduced BRC1b expression levels, which would result in a phenocopy of the BRC1b RNAi phenotype. We quantified BRC1b and BRC1a mRNA in axillary buds of the miR156-OE lines and observed that BRC1b (but not BRC1a) was significantly downregulated in these plants, both in LD and in SD ( Supplementary Fig.  10d,e).
Altogether, these findings indicate, first, that miR156 negatively controls BRC1b expression in potato axillary buds, and second, that the aerial tuber phenotype of miR156-OE can be explained, at least in part, by a reduced activity of BRC1b. Moreover, the loss of BRC1b function in RNAi buds causes an induction of miR156 and downregulation of both miR156-dependent and independent SPL expression, perhaps not directly, but rather associated with the development of aerial tubers.

Discussion
In potato, all axillary meristems (aerial and underground) have the potential to remain dormant or develop into a branch, stolon or tuber. The acquisition of either of these fates requires the activation of specific genetic programs in response to positional and environmental cues. In the ssp. andigena, tubers are formed only in the SD of winter, and usually at stolon tips, below ground. This spatial restriction probably evolved to protect tubers from animal foraging and unfavourable environmental conditions over the winter and to ensure vegetative propagation the following spring.
In this study, we have shown that BRC1b is essential to preventing aerial tuberization and promoting efficient underground tuber formation: BRC1b silencing leads to ectopic aerial tuber formation and reduced below-ground tuberization. This suggests that in the WT, BRC1b prevents aerial buds from becoming strong sinks for sucrose in conflict with stolons and probably facilitates stolon sink dominance and tuberization (Fig. 7).
BRC1b loss of function causes a collection of interconnected physiological, cellular and molecular alterations in aerial axillary buds, mainly detectable in SD: impaired bud dormancy, reduced ABA levels and responses, cells with an increased number of PD, enhanced symplasmic movement, and abnormal accumulation of sucrose and SP6A mRNA (Fig. 7).
One of the earliest features of RNAi aerial buds, detectable as early as 2SD, is their failure to enter dormancy, consistent with the active-bud phenotype of brc1 mutants in other species 43 . Nevertheless, although active axillary buds are necessary 39 , they are not sufficient to promote tuber development: BRC1a loss of function also causes bud activation, but plants lacking BRC1a do not display aerial tubers 15 . This indicates that BRC1b plays a divergent, more specific role in tuberization. A key differential feature of BRC1b (compared with BRC1a) is the unique capability of its encoded protein to directly bind the SP6A protein. In other species, the interaction between BRC1-like and FT-like proteins results in the inactivation of the latter 14,38 (see below). Thus, in potato, BRC1b-SP6A interaction may help prevent SP6A tuber-promoting activity in axillary buds.
RNAi buds also have reduced ABA levels and responses relative to the WT, in agreement with observations in brc1/tb1 mutants of other species 13,25 . Reduced ABA responses may contribute to a reduced dormancy of RNAi buds, as ABA is necessary to maintain bud growth arrest 9,12,24,[44][45][46] .
In hybrid aspen, SD-induced ABA signalling leads to a blockage of symplasmic intercellular communication that prevents FT-like proteins from entering the shoot apex to promote growth 9,14 . Likewise, SD-induced ABA signalling may lead to reduced PD-mediated transport of SP6A into WT potato aerial buds. Indeed, WT bud cells (with higher ABA levels than RNAi) have fewer PD and probably more restricted symplasmic trafficking into buds, as indicated by the reduced movement of CFDA in the WT compared with RNAi. Furthermore, several MCTPs (PD-associated proteins found in Arabidopsis and rice to be involved in cell-to-cell trafficking of FT-like proteins and other developmental factors 34,47,48 ) become more downregulated in the WT than in RNAi at the LD-to-SD transition. This could promote the increased selective symplasmic transport of SP6A into the RNAi buds.
We have also observed significantly increased levels of sucrose in RNAi buds. An enhanced symplasmic movement of sucrose into the buds could also explain this abnormal accumulation. This situation would mirror that of WT stolon tips, which undergo a switch in sucrose unloading from apoplastic to symplasmic during tuberization that enhances their sink potential 36 . This is a key step in the regulation of resource allocation during tuber formation that could take place ectopically in RNAi aerial buds, leading to major alterations in photoassimilate distribution 36 . Indeed, changes in the strength of individual sinks can greatly impact the final assimilate partitioning 49 , which in this case would result in a poor allocation of assimilates to stolons. Remarkably, aerial tubers are more frequently formed in the topmost nodes (where flowers are usually formed), indicating that sucrose fails to be transported basipetally as it occurs during tuberization. In the WT, sucrose promotes the expression of SP6A in phloem and developing tubers 28,50 . In RNAi buds, the increased sucrose levels at 7SD coincide with a strong accumulation of the SP6A mRNA, which could in part be sucrose-induced. The feedback regulation of SP6A by its own protein may account for the additional increase in SP6A mRNA levels 1 . SP6A can, in turn, boost sucrose unloading into the buds by a mechanism similar to the one taking place in stolons: SP6A directly interacts with and antagonizes the sucrose transporter SWEET11b in the plasma membrane, thus preventing sucrose apoplastic leakage and facilitating symplastic transport 28 . Due to this cross-talk, the levels of sucrose and SP6A, key players in tuberization, could escalate and trigger tuber formation in RNAi aerial buds (Fig. 7). In the WT, the direct interaction of BRC1b with SP6A would be a buffering mechanism to prevent the initiation of this positive feedback. Furthermore, BRC1b activates SWEET11b (Fig. 4d, columns 1 and 2), which facilitates apoplastic transport 28 .
The aerial tuber phenotype of lines overexpressing miR156 (a microRNA that negatively controls BRC1-like genes) can be due to the observed downregulation of BRC1b in these lines. Likewise, the phenotype of other potato lines displaying aerial tubers, such as the strigolactone-deficient CAROTENOID CLEAVAGE DIOXYGENASE 8 (CCD8) RNAi lines 51 , could also be explained by reduced BRC1b activity. Indeed, strigolactone mutants have low expression levels of BRC1-like genes in most angiosperms 6 . Furthermore, the overexpression of miR156 causes a reduction in strigolactone content in potatoes 41 .
A remarkable degree of genetic conservation is found between the pathways that control flowering time in annuals, apical growth (versus dormancy) in perennials and tuber formation in pot ato 1-3,14,52,53 (this work). Notably, all are sugar-demanding processes that take place in strong sink organs. In those sinks, sucrose supports growth and activates FT-like genes, which ultimately regulate these developmental pathways 28 . In contrast, BRC1-like genes, repressed by sugar 5,6 and active under energy-limiting conditions, promotes ABA signalling and, directly or indirectly, causes a reduction in the number of PD. This restricts symplasmic transport of sucrose and SP6A into the buds, which in turn are preferentially transported basipetally to the stolons. In buds, any residual SP6A protein is bound and antagonized by direct interaction with BRC1b proteins. The green arrows indicate the positive feedback loops between sucrose and SP6A and between SP6A and its own gene. In RNAi (right), the lack of BRC1b leads to bud activation. ABA signalling is reduced, and PD number is increased; this facilitates the movement of sucrose and SP6A into the buds. Moreover, the SP6A proteins accumulated in the buds are not bound by BRC1b. Axillary buds become strong sinks that compete with stolons for sink dominance. This leads to alterations in the distribution of photoassimilates, which preferentially enter the aerial buds instead of being transported to stolons.
promote energy-saving mechanisms and antagonize these developmental pathways in part by negatively modulating the activity of FT-like genes via direct protein-protein interactions, but probably also through other mechanisms 14,38 (this work). This reveals the existence of a widely conserved FT-BRC1 genetic module, perhaps evolved in early angiosperms, in which FT promotes processes requiring sucrose and energy, whereas BRC1 genes activate energy-saving programs 5,6 . The positive (FT) and negative (BRC1) regulation of these genes by sucrose or sucrose derivatives, and the direct interaction and mutual regulation of their proteins, would directly integrate the metabolic status of the plant with developmental decisions that require a tightly regulated use and allocation of photoassimilates. Plants used for phenotypic or expression analysis (qPCR or RNA-seq) in LD versus SD were first grown in the greenhouse and then transferred to a growth chamber with similar temperature and light conditions and an adequate photoperiod. For the co-immunoprecipitation assays, Nicotiana benthamiana plants were grown in the greenhouse in similar conditions. Plant phenotyping. Phenotypic analyses in LD and SD were done as follows: ten synchronized plants of each genotype, with six visible nodes, were transferred to soil; five nodes were buried. For LD phenotyping, plants were grown for 16 weeks in the greenhouse; for SD phenotyping, plants were grown for 10 weeks in the greenhouse (LD) and then transferred to a growth chamber in SD (8 h light, 16 h dark) for 6 weeks. Branches and stolons were counted when longer than 1 cm. All tubers were considered in the analyses. Phenotypes were analysed when the WT plants became senescent. At this time, the transgenic plants were still green. For the leaf phenotyping, plants were grown for six weeks in the greenhouse, and then leaves from even nodes were collected starting from the lowest nodes.
For the proBRC1b-GUS construct, a genomic fragment comprising 1.7 kb upstream of the ATG was amplified from genomic DNA of Désirée cv. plants with Phusion polymerase (New England BioLabs; the primers are given in Supplementary Data 3). The PCR fragment was cloned in pGEM-t Easy (Promega), re-amplified by PCR with primers with attB tails and BP-cloned into pDONR207 (Invitrogen). The fragment was then mobilized into the destination vector pGWB3 54 with LR-clonase.
To generate the BRC1b RNAi constructs, BRC1b-specific PCR products (168 base pairs) of a coding region non-conserved with other TCPs were obtained with the primers indicated (Supplementary Data 3) and cloned into the vector pHANNIBAL (CSIRO) as previously described 16 . The pHANNIBAL cassettes were then digested with NotI and subcloned in the NotI site of the binary vector pART27 as previously described 15 . For the transgenic control lines, an empty pHANNIBAL cassette subcloned in pART27 was used. The miR156-OE lines were made using the previously described plasmid 35SCaMV-miR156a (ref. 39 ), kindly provided by Y. Eshed.

Generation of transgenic plants. Binary vectors were transformed into the
Agrobacterium tumefaciens strain AGL-0. To generate stable transgenic potato plants, leaves were transformed as described in ref. 15 . BRC1b silencing and PDK intron expression (intron of the hairpin) were confirmed by qPCR in axillary buds using the primers indicated in Supplementary Data 3. BRC1b RNAi lines 5, 10 and 19 were selected in the ssp. andigena background, and lines 13, 27, 36 and 40 were selected in Désirée cv.
CRISPR-Cas9 mutagenesis. BRC1b mutagenesis by CRISPR-Cas9 was performed by transforming plants with the pK7mG34GW,0 destination vector containing two BRC1b-specific guide RNAs (CAGCTGTCAGCTAAGACGTG and GAAGGCCGAGGCGGAAACAA) and the Cas9-TPC 55 under the Ubiquitin10 promoter. Potato (ssp. andigena) plants were transformed as described above. To avoid chimaeric mutated plants, the leaves of transgenic plants were used for de novo callogenesis and new plant regeneration. BRC1b mutagenesis was confirmed by amplicon sequencing (Genewiz.com). An amplicon comprising both single guide RNAs was obtained by PCR using the primers indicated in Supplementary Data 3. Allelic frequency was analysed using the FASTQ files with Crispresso2 (ref. 56 ). We analysed 57,643, 37,206, 46,803 and 274,903 paired-end reads for brc1b CRISPR mutant lines A6, A10, and A12 and the WT, respectively. GUS staining. GUS staining was conducted as previously described 13 .
RNA in situ hybridization. RNA in situ hybridization and image acquisition were performed as previously described 57 . Tissues were collected at the end of the day. The samples were fixed with formaldehyde-acetic acid-ethanol, dehydrated with ethanol and embedded with paraffin (Paraplast, Leica). After dehydration, the samples were processed in an embedding station (EG1160, Leica). Then, 8-µm sections were obtained with a rotary microtome (RM2265, Leica) and transferred to polylysine-coated slides (Roth). A BRC1b full-length probe was generated from its cDNA and cloned into the pGEM-T Easy Vector (Promega). The probes were synthesized with a DIG RNA Labeling Kit (Roche).

Stem section.
Stems from the basal part (between the second and fourth nodes) of one-month-old BRC1b RNAi plants were cut and placed in cold 90% acetone and fixed in 4% glutaraldehyde. Dehydration, staining, resin inclusion and sectioning were done as previously described 58 .
Sucrose measurements. Plants were grown in the greenhouse for six weeks and transferred to an LD growth chamber for one week to set up dawn time. Leaves and axillary buds were sampled one hour after dawn and flash-frozen in liquid N 2 . After the LD samples were collected, the chamber photoperiod was set to SD conditions, and samples were collected one week later. Material termed 'upper axillary buds' was from the four aerial axillary buds below the third visible node counting from the apex. Material termed 'lower axillary buds' was from the first four nodes starting from ground level. The leaf samples were from these lower nodes. Biological replicates were pooled material from four plants. The leaves were ground in the mortar and the axillary buds in Qiagen TissueLyser. An aliquot of the powder was transferred to a 2-ml-cap tube and weighed in a precision scale for normalization of sucrose content. Then, 1 ml of 80% ethanol was added to the ground tissue, and the tubes were placed in a thermo-shaker at 80 °C for one hour. The supernatants of the first wash containing the reducing sucrose were dried in a Speed-vac, and the pellet was resuspended in 200 µl of milli-Q water. Sucrose was quantified with the Starch Saccharose/D-Glucose kit (Roche/r-Biopharm) following the NAD+/NADH technique.

Experimental design of the RNA-seq time course and qPCR experiments.
Plants were grown in the greenhouse for six weeks and then placed in an LD growth chamber for one week, after which samples for the LD time point were collected. The growth chamber photoperiod was then set to SD by bringing the night-time forwards. The following day corresponds to the first day in SD. Samples were collected at two days (2SD), one week (7SD), two weeks (14SD) and three weeks (21SD) (only for qPCR). All samples were obtained at one hour after dawn. Upper axillary buds (see the sugar measurements, those most likely to produce tubers) were collected. Twenty-four samples of S. tuberosum ssp. andigena were sequenced on a BGISEQ-500 Platform, generating about 7.01 Gb per sample. The average genome-mapping rate was 83.28%, and the average gene-mapping rate was 69.20%. RNA-seq paired-end reads were filtered using SOAPnuke v.1.5.2 (ref. 59 ) to remove reads with adaptors, having more than 5% unknown bases or with poor base quality (more than 20% of bases having quality <15). The cleaned reads were then aligned to the PGSC_DM_v4.03_pseudomolecules reference genome 60 (available at http://solanaceae.plantbiology.msu.edu/pgsc_download.shtml) using HISAT v.2.0.4 (ref. 61 ) with the sensitive mode, options no-discordant and no-mixed, minimum fragment size 1 and maximum fragment size 1,000. Reads per gene quantification for each sample was performed using the FeatureCounts function implemented in the R package Rsubread v.2.0.0 (ref. 62 ) with the PGSC_DM_V403_genes.gff file from http://solanaceae.plantbiology.msu.edu/pgsc_download.shtml as genome annotation and parameter isPairedEnd = TRUE. Differentially expressed genes were then calculated for each comparison between conditions using the R package DESeq2 v.1.26.0 (ref. 63 ) with the default values. The RNA-seq data are available in the NCBI Gene Expression Omnibus database with accession code GSE155774.

GSEA.
The tests of overrepresentation of the different gene sets were carried out with the GSEA 18 method (version 3.0) as described in ref. 5 . All the gene sets defined for this study were previously used 5 , except the gene sets for tuber eye dormancy and sprouting 19 . All gene sets are available in Supplementary Data 4. Potato putative orthologues of A. thaliana genes were determined according to the potato genome annotation computed by PHYTOZOME (Stuberosum_448_v4.03. annotation_info; based on genome version v.4.03, best hits in BLASTP alignment to A. thaliana TAIR10). When several potato gene loci were associated with a single Arabidopsis gene, all of them were included in the gene set. The GSEA method evaluates whether these genes occur preferentially towards the top or bottom of a ranked list. The complete results are available at https://bioinfogp.cnb. csic.es/files/projects/nicolas_et_al/.
Real-time qPCR. We used 1 µg of total RNA of axillary buds for reverse transcription using the High-Capacity cDNA Reverse Transcription Kit from Applied Biosystems following the manufacturer's instructions. We diluted cDNA seven times before performing qPCR. GoTaq qPCR Master Mix (Promega) and the Applied Biosystems 7500 real-time PCR system were used to perform qPCR. The primers used are in Supplementary Data 3. StACTIN8 was used as a reference gene 1, 15 . Three technical replicates were performed for each of the three biological replicates. The gene relative expressions were calculated as previously described 15 .
We determined miR156 expression levels by using the stem-loop qPCR method 64 . We used 300 ng of total RNA for the cDNA synthesis. U6 small nuclear RNA was used as a reference gene. Three technical replicates were performed for each of the three biological replicates.
ABA quantification. Plant endogenous ABA was analysed using high-performance liquid chromatography-electrospray-high-resolution accurate mass spectrometry using an Orbitrap Exploris 120 Mass Spectrometer (Thermo Fisher Scientific). The extraction of this hormone was carried out as previously described 65 . The detection and quantification of ABA were carried out using a Product Ion Scan experiment in the negative-ion mode, employing a multilevel calibration curve with the internal standard (2H6-(+)-cis,trans-ABA). Instrument control and data processing were carried out by TraceFinder v.5.1 EFS software.
PD transmission electron microscopy imaging and quantification. Meristems were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) and 2.5% glutaraldehyde (TAAB Laboratories) in phosphate-sucrose buffer for 4-5 h at room temperature and overnight at 4 °C in fresh fixative buffer. The samples were washed with the same buffer, post-fixed with 1% osmium tetroxide (TAAB Laboratories) in 0.8% potassium ferricyanide (Sigma) for 1 h at 4 °C and incubated with 2% aqueous uranyl acetate (Electron Microscopy Sciences) for 1 h at the same temperature. After washing with distilled water, the samples were dehydrated with increasing concentrations of acetone (VWR) and embedded in epoxy resin TAAB 812 (TAAB Laboratories). Polymerization was carried out for two days at 60 °C.
The resin blocks were trimmed, and semi-thin sections were obtained and stained with toluidine blue 0.1% (Merck) to select the area of interest. Next, ultrathin 70 nm-thick sections were obtained using a Leica EM UC6 ultramicrotome, transferred to 200 mesh formvar/carbon-coated nickel grids (Gilder) and stained with saturated uranyl acetate and 0.2% lead citrate at room temperature. The sections were visualized on a JEOL JEM 1011 electron microscope (operating at 100 kV). Micrographs were taken with a Gatan Erlangshen ES1000W CCD camera. Six WT and six RNAi buds were analysed, and 10-50 images were obtained for each bud, from cells located close to the developing vasculature, below the meristem dome, where BRC1b is expressed. PD were quantified in the micrographs and normalized by μm of plasma membrane in each image. The plasma membrane length in each photograph was calculated using the Measure Segmented Line tool in ImageJ/FIJI (https://imagej.net/software/fiji/). CFDA mobility assays. WT and RNAi plants grown in vitro in LD for a month were transferred to an SD growth chamber. Two weeks later, the uppermost part of each stem containing four nodes was cut and transferred to a tube containing a solution of 1 mg ml −1 CFDA (Sigma) in distilled water (from a 30 mg ml −1 DMSO-dissolved stock solution). Four hours later, individual axillary-bud-containing nodes were excised, hand-sectioned in half, mounted and observed under an epifluorescence microscope (Leica DMR) using a green (530-550 nm) emission filter and bright field optics. Microphotographs were taken with an Olympus DP70 digital camera. The quantification of CF in buds was determined by the relative fluorescence intensity in the axillary bud normalized by the fluorescence intensity in the stem below using the tool Measure Integrated Intensity in ImageJ/FIJI (https://imagej.net/software/fiji/). Four to six buds per plant were analysed for 10 WT plants and 12 RNAi plants.
Y2H assays. Vectors were transformed in yeast strain AH109. Yeast interaction assays were carried out in selective medium deficient in leucine, tryptophan, adenine and histidine (SD-LWAH) with 20 mM or 50 mM 3-AT as previously described 66 . Yeast colony growth was compared with their equivalent in SD-LW plates (positive growth controls). Each combination was analysed three times.
Co-immunoprecipitation assays. Constructs were transformed in the Agrobacterium tumefaciens strain AGL-0. Overnight cultures of Agrobacterium were resuspended in 10 mM MES (pH 5.5), 10 mM MgCl 2 and 150 mM acetosyringone at OD 600 nm = 0.5 and incubated for three hours at room temperature. Identical volumes of each resuspended culture were mixed for co-infiltrations in N. benthamiana leaves. Samples were collected 24 hours after infiltration. Co-immunoprecipitation was performed as previously described 66 .
APB-FRET assays. The APB-FRET assays were performed as previously described 15 . FRET efficiency (E FRET ) was calculated as the percentage of increase in donor fluorescence after acceptor photobleaching; E FRET = 100 × (D post − D pre )/D post , where D pre and D post are the fluorescence intensity of the donor before and after photobleaching, respectively. D pre and D post were measured with a secondary region of interest (ROI) inside the bleached region of the first ROI (4-mm-diameter ROI for nuclei). A 1.2-kb truncated luciferase protein fused with either Citrin or mCherry was used for the negative control assays. BRC1b was also fused with Citrin or mCherry. The SP5G and SP6A proteins were fused to the MYC nuclear localization signal and Citrin (SP5G-MYC-Citrin and SP6A-MYC-Citrin) to do the APB-FRET assays in the nucleus. Proteins fused with Citrin were under a 2X35S promoter. Proteins fused to mCherry were under an estradiol-inducible promoter (pABindmCherry 15 ). Gene expression was induced the night before the APB-FRET assays.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus under accession no. GSE155774. Source data are provided with this paper. 1 nature portfolio | reporting summary

March 2021
Corresponding author(s): Pilar Cubas and Michael Nicolas Last updated by author(s): Feb 2, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection no software was used for data collection

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy -RNA-seq data are available at the NCBI GEO database with Accession Code GSE155774.
-Complete results of the GSEA are in https://bioinfogp.cnb.csic.es/files/projects/nicolas_et_al/ Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
For axillary bud q-RT-PCR 3-4 biological replicates that comprised 16 axillary buds (4 buds from 4 individuals) and 3 technical replicates for each biological replicate, were performed as indicated in the manuscript. For in situ hibridization, three biological replicates (i.e. tissue from three independent plants) was used. Plus we always generated consecutive sections, which were all transferred to the respective slide -these also serve as controls.
For sugar quantification analyses results are mean of three biological replicates comprising either 4 leaves or 16 axillary buds (4 buds from 4 individuals).
For phenotypic analyses, 10 plants of each genotype/condition/time-point were analysed, as indicated in the manuscript. For RNA-seq, three biological replicates were analysed that comprised 16 axillary buds each (4 buds from 4 individuals).
For CFDA mobility assay, 10 WT plants and 12 BRC1b RNAi plants were analysed (4-6 buds each). For ABA content measurement, three biological replicates were analysed for each genotype/time-point that comprised 16 axillary buds each (4 buds from 4 individuals). For Plasmodesmata number counting, pictures from 2 buds from 3 different plants of each genotype were taken (at least 10 pictures for each bud) Sample size (number of plants used for each analysis) was mainly determined by the limitations of our growth chambers. Neverthless they had been proven sufficient to provide robust and significant differences (with very low P values) in reproducible experiments. These numbers are also similar to those classically used by other research groups in other published work with satisfactory results.

Replication
For the phenotypic analyses, 4 experiments were performed that gave similar results with comparable conclusions. The last one, which contained the highest number of individuals, was included in the manuscript. The co-IP experiments were done three times.
The CFDA experiments were done three times.
No replication problems were found in any experiment.
Randomization For all studies, individual plants of different genotypes were randomized in the growth chambers to avoid environmental effects.

Blinding
No blinding was performed in the experiments as the control and experiment plants are very different phenotypically and impossible to mistake for each other Reporting for specific materials, systems and methods We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.