Probing the floral developmental stages, bisexuality and sex reversions in castor (Ricinus communis L.)

Castor (Ricinus communis L) is an ideal model species for sex mechanism studies in monoecious angiosperms, due to wide variations in sex expression. Sex reversion to monoecy in pistillate lines, along with labile sex expression, negatively influences hybrid seed purity. The study focuses on understanding the mechanisms of unisexual flower development, sex reversions and sex variations in castor, using various genotypes with distinct sex expression pattern. Male and female flowers had 8 and 12 developmental stages respectively, were morphologically similar till stage 4, with an intermediate bisexual state and were intermediate between type 1 and type 2 flowers. Pistil abortion was earlier than stamen inhibition. Sex alterations occurred at floral and inflorescence level. While sex-reversion was unidirectional towards maleness via bisexual stage, at high day temperatures (Tmax > 38 °C), femaleness was restored with subsequent drop in temperatures. Temperature existing for 2–3 weeks during floral meristem development, influences sexuality of the flower. We report for first time that unisexuality is preceded by bisexuality in castor flowers which alters with genotype and temperature, and sex reversions as well as high sexual polymorphisms in castor are due to alterations in floral developmental pathways. Differentially expressed (male-abundant or male-specific) genes Short chain dehydrogenase reductase 2a (SDR) and WUSCHEL are possibly involved in sex determination of castor.

Plant sexual diversity forms the basis of taxonomy and has evolved to achieve mating success in flowering plants [1][2][3][4][5][6] . Sex expression or sexuality in plants can be altered in response to changes in environment or age of plant, and such lability in sex expression is significant in long term survival and adaptation of a species 7,8 .
Castor (Ricinus communis L.) is an industrially important oilseed crop, the non-edible oil mainly used as a lubricant due to high viscosity 9 . The inflorescence of castor is monoecious, monoclinal, primarily wind-pollinated raceme, bearing unisexual flowers, with female flowers found on the apex and male flowers borne at the bottom of inflorescence 10 . Though a diploid (2n = 20), belonging to a monotypic genus of family Euphorbiacea, castor exhibits racial differences for sex tendency 11 . High polymorphism of sex expression in castor ranges from genotypes having spikes which are completely pistillate to completely staminate, with interspersed staminate flowers or ISF (male flowers interspersed throughout in the pistillate spike after capsule formation), monoecious with apical ISF (male flowers interspersed in the apical region of spike), or apically non-interspersed, or with terminal hermaphrodite flower and having monoecious sex variants showing variation in percentage as well as the relative position of male and female flowers within the genotype 10,12,13 . In addition, the highly unstable pistillate character reverts to monoecism 14 . The hereditary instability affecting sex, manifested as sex reversal in pistillate lines is a widely prevalent phenomenon in castor, which negatively influences hybrid seed purity. Such reversions may be early or late, when it occurs respectively at lower or higher than quaternary orders of spikes, later reversions resulting in more females in the progenies 10,15 . During hybrid seed production, the reverted pistillate lines themselves act as a source of pollen. Roguing of reverted spikes is cumbersome and increases the production costs. Conventional method of hybrid seed production retains 20-25% monoecists for maintenance of pistillate lines, resulting in high percentage of monoecists and early revertants in seed production plots. In a modified method

Results
Inflorescence development has distinct stages and inflorescence architecture varies with genotypes. Castor exhibits determinate growth, where main stem and subsequent branches terminate in a racemose monoecious inflorescence, but diverse sex patterns were observed. Different castor genotypes with distinct and diverse sex phenotypes ranging from completely pistillate to completely staminate and monoecious with different proportion and position of male and female flowers were used in the present study (Fig. 1).
The shoot apical meristem (SAM) was enveloped by many whorls of bracts and leaves and not visible externally. During vegetative growth, the dome-shaped shoot apical meristem produced leaf primordia. Each leaf at a node was subtended by a bract. The vegetative stage terminated with initiation of inflorescence primordium. Based on ontogeny of distinct morphological events, eight morphological stages of inflorescence development, from floral initiation to capsule setting, were identified in castor ( Fig. 2A, Supplementary Table S1). At stage I, inflorescence primordium initiated, but was not distinct from vegetative shoot apex. At stage II, inflorescence primordium was externally visible as a bulge. The inflorescence increased in girth and had fully formed flower buds at stage III. Stage IV represented spike opening from tip downwards, with three sub-stages to mark beginning, half and three-quarter opened inflorescence. Increase in girth of inflorescence from stage I to stage IV are shown ( Supplementary Fig. S1). Spike emerged fully at stage V and elongated at stage VI. During elongation, flower buds were added to floral whorls. Anthesis or flower bud opening (stage VII) and capsule formation (after pollination) and maturation (stage VIII) occurred. Spike elongation continued during capsule maturation. For scoring the sex phenotype, end of stage VI, just before anthesis was ideal for most genotypes, since elongation and addition of the floral whorls would mostly be complete. Stage VIII was scored for ISF.
Time taken for transitions to each stage and anthesis pattern varied with the genotype (Supplementary Fig. S2; Supplementary Tables S2A and B). In monoecious DCS 107, a distinct stage of elongation (stage VI) existed before anthesis, but in other genotypes elongation was distinct during or after anthesis. In DPC 9, anthesis started immediately after emergence (3 days), when compared to other genotypes (7-10 days). Elongation was greater during capsule formation and maturation ( Supplementary Fig S2; Supplementary Tables S2C and D). Studies on anthesis pattern revealed that in monoecious DCS 107, anthesis first occurred in the lower whorl of female flowers, but in monoecious RG 156, 1-2 male flowers opened first, followed by lower whorls of female flowers and majority of male flowers remained unopened even during capsule formation. In pistillate DPC 9 and DPC 21, bottom whorl of female flower buds opened first and continued up. In DPC 17-S3 (monoecious apical ISF), the bottom whorls of female flowers opened first followed by middle whorl. The male flowers at the bottom of the inflorescence opened only after anthesis of 75% of female flowers and mostly during capsule formation stage.
The inflorescence architecture in terms of number of floral whorls, sex and relative position of flower buds varied in different genotypes ( Supplementary Fig. S3). Bottom whorls were more branched and with more number of buds than top whorls. In monoecious genotypes (DCS 107 and RG 156) proportion of male flowers gradually decreased, while that of female buds gradually increased to completely female whorls towards top. Monoecious genotypes also differed in inflorescence architecture. In DCS 107, the female flower bud appeared with in first to third whorl from bottom ( Fig. 2B), whereas in RG 156, not a single female bud was observed in lower whorls (Fig. 2C). In pistillate lines DPC 9 and DPC 21, female buds alone were seen in all the whorls, although male flowers were observed during sex reversion or ISF in DPC 9 and capsule formation (ISF) in DPC 21. In M 574-OS1, only male flowers were observed from quaternary orders but the terminal flower was occasionally bisexual in some spikes. In DPC 17-S3 (with apical ISF spike), male flowers were found throughout the inflorescence, but completely female whorls occurred occasionally in between (Supplementary Table S3). In top whorls with triplet buds, one of the two lateral buds were male and the terminal bud was mostly female. Also, few slightly pointed male flowers towards the spike tip were observed in DPC 17-S3 and DCS 107, while some female flowers of RG 156 and DPC 9 were round than elongated ( Supplementary Fig. S4), which were more prominent during summer. For a given length, the number of floral whorls decreased with branch order (tertiaries had less whorls than secondaries). Every spike order in a genotype had definite number of floral whorls during emergence which did not increase significantly later, but increase in floral whorls during inflorescence elongation and development was noticed in monoecious DCS 107 and staminate M 574-OS1 (p < 0.05) (Supplementary Table S3; Supplementary Fig. S5) indicating potential of sex lability with environmental conditions during growing season in these genotypes. The number of flowers in a floral whorl (especially in lower whorls) may increase, due to continuous development of flower primordia in triplets, resulting in more branching.
Apical meristem shows chronology of organ developmental stages during inflorescence initiation and development. Inflorescence primordia initiation at stage I, though not distinct externally and inflorescence development at stage II were identified by histology sections and SEM. The much-condensed internodes near apical meristem, enveloped by bracts were revealed only after emergence of each leaf at a node, which takes approximately 3-4 days. Inflorescence primordia initation usually occurs at 8-12 nodes in the primary branch or later depending on genotype, growth conditions and seed quality (data not shown). Fully emerged inflorescence (stage V) appeared 3-5 nodes later than or 2-3 weeks after inflorescence initiation at stage I.  www.nature.com/scientificreports/ During the vegetative phase, the dome-shaped apical meristem enveloped within bracts had 2-to-3 protuberances (Fig. 3A,B), which later differentiated into leaf-like structures that emerged disrupting the bracts (Fig. 3C,D). The protrusion can be also the bract. The inflorescence primordia initiation was marked by elongation of apical meristem beyond the youngest leaf primordium, to become the main axis of inflorescence (Fig. 3E), followed by floral primordial differentiation, where lateral regions of meristematic activity or individual floral buds were differentiated on the axis (Figs. 3F,G). Growth and elongation of inflorescence axis occurred in a centripetal fashion from bottom to top, adding new floral whorls as it elongated (Fig. 3H). Meristematic growth continued at inflorescence tip while the lateral regions continued differentiating (Fig. 3I). The buds in the lower whorls matured first and individual flower buds were at different developmental stages (Fig. 3J). In monoecious genotypes, male buds differentiated first and female buds were observed after the inflorescence elongated further Bisexuality and reversion to bisexuality occur in castor flowers. Occurrence of bisexual flowers having fully developed male and female organs was rare and bisexual flowers did not occur in all castor genotypes under normal conditions. Spikes of DPC 16, DPC 9-OS2 and M 574-OS1 had terminal hermaphrodite flower. In DPC 16 and DPC 9-OS2, few bisexual flowers were also found near the spike tip. A gradation in bisexuality exists in castor, where three categories of bisexual flowers were identified such as, bisexual flower which was predominantly female, having well-developed ovary and few rudimentary stamens ( Fig. 6A-D), bisexual flower with equally and fully developed gynoecium and androecium ( Fig. 6E-H) and predominantly male bisexual flower with underdeveloped tubular ovary, rudimentary stigma or without bifid stigma ( Fig. 6I-L). The terminal bisexual flowers generally had equal development of male and female structures.
Out-crossing in pistillate line DPC 9 (DPC 9-OS2) resulted in monoecious inflorescence with bisexual flower(s) at and near tip. In lower branch orders, bisexual flowers had equally developed ovary and stamens, but transited to predominantly male bisexual flower, with rudimentary tubular ovary, and gradually to completely male flower. The terminal bisexual flower progressively became male and had rudimentary gynoecium during stages of transition, the progressive transition being noticed in bisexual flowers of different spikes of higher order and not of the same spike. Similar sex expression was observed in M 574-OS1 and in few plants of VP1 (data not shown).
At high day temperatures during summer, sex reversion was unidirectional towards maleness, at floral and inflorescence level, from female to bisexual, in female flowers, and from bisexual to male, in spikes with terminal bisexual flowers. Female flowers of both monoecious and pistillate lines reverted to bisexual flowers with www.nature.com/scientificreports/ rudimentary stamens, at most developmental stages from stages 6 to 12 ( Fig. 6M-Q). The aborted stamen primordia, which remained dormant as a layer of cells beneath the ovary (Fig. 6R,S), resumed development as small protrusions or bulges at the base of the ovary during reversion (Fig. 6T), to result in rudimentary stamens. Such reversions occurred even after complete pistil development, but simultaneous development of gynoecium and androecium were not observed in reverted female flowers. When reversion occurred during later stages of female flower development (after complete development of gynoecium), rudimentary stamens appeared without complete development, with less branching and development which was limited to filament bifurcation (Fig. 6U,V). Also, anthers in reverted stamens of bisexual flowers of DPC 9-OS2 were larger in size    www.nature.com/scientificreports/ (terminal), DPC 21, and DPC 17-S3 (terminal) and DPC 9 (reverted) (Fig. 7B). Sex reversions in pistillate lines resulted in monoecious racemes, but a monoecious raceme never reverted to a pistillate raceme. When temperatures dropped down to 28-33 °C (T max ) by June 4 th week , from 38-41 °C (T max ) during May 3 rd week of 2017, normal phenotype was restored in all genotypes (Fig. 7C). In M 574-OS1 and DPC 9-OS2, female flowers started appearing in terminal position in inflorescences at later orders with drop in temperature in June. The sex of the terminal flower in different inflorescences of M 574-OS1 and DPC 9-OS2 varied from female to bisexual to male with increase in temperatures during summer, and then from male to bisexual to completely female, when the temperatures dropped down after summer. This indicated that the transition of monoecious inflorescence with 20% female flowers and tip bisexual flower, to staminate inflorescence with bisexual flower and later to completely staminate inflorescence (from 4th or higher branching order onwards), was due to temperature, rather than the higher branch order or ageing that occurred during the advancement of growing season.
Weather parameters such as daily maximum temperatures (T max ), growing degree days (GDD), daily minimum temperatures (T min ) and daily average temperatures (T avg )  The temperature conditions prevailing during inflorescence bud initiation affect the sex phenotype and the proportion of male and female flowers in the inflorescence. Hence the mean temperatures of T max , T min and T avg existing for 2-3 weeks prior to actual observation of the altered phenotype in summer were calculated, and the altered phenotypes were observed at T max > 38 °C, T min > 23 °C and T avg > 30 °C (Fig. 7G,H). Alteration of sex phenotypes during summer towards maleness occurred in all spike orders, irrespective of the order of branching (Supplementary Table S4A). Bisexual flowers occurred in spikes of monoecious and pistillate genotypes in summer, while bisexuality and pistillateness returned in spikes of staminate genotypes with drop in temperature after summer (Supplementary Table S4B-D).
Sudden variations (rise or drop) in temperatures, rather than the absolute value of T max seem to alter sex expression. In addition to temperature, genetic factors also contribute to sex variations, since completely male inflorescences were observed even in the primaries of DPC 9-OS2 during normal temperatures of growing season by August-September 2017 (Data not shown).
The mean of weather parameters existing 2 weeks before sample collection for scanning electron microscopy (Supplementary Table S5) shows that, though the DPC 21 (ISF line) samples were collected during January-February (T max : 28-29 °C and CDD: 2650-2856), female flowers of DPC 21 exhibited distinct stamen primordia than that of DPC 9 and RG 156 which were collected during summer (April 2016, CDD = 4032), indicating that the conspicuousness of arrested organs depends on the genotype as well.
Differentially expressed genes may determine sexuality in castor flowers. Expression of 5 out of 17 candidate genes, were consistent and re-verified by semi-quantitative RT-PCR. Expression of 1-aminocyclopropane-1-carboxylate synthase (ACC synthase) genes, ACS and ACS-1 were upregulated by 1.1-1.2 times, in male buds of monoecious genotype RG 156 (collected by May, 2015). Expression level of ACS was more than that of ACS-1 in all the tissues (Fig. 8A). Expression of SDR) and ACS-1 were compared in monoecious DCS 107, staminate M 574-OS1 and pistillate DPC 9 in samples collected by February-March, 2017. Short-chain dehydrogenase reductase 2a (SDR), a homologue of Zea mays TASSELSEED2 (TS2), was expressed in all tissues such as vegetative and reproductive (differentiated) SAM, primordial leaves, male and female buds etc, but was highly expressed in male buds of monoecious genotypes DCS 107 ( Fig. 8B) and RG 156 ( Supplementary Fig. S7A). SDR expression was higher in male buds (by 1.4 times) than female buds of monoecious genotype DCS 107 and much higher in male buds of staminate genotype M 574-OS1 (by nearly 1.8 times) than that of monoecious genotypes, while feeble expression was noticed in female buds of pistillate DPC 9. Expression pattern of ACS-1 was also similar to SDR in DCS 107, expression being higher in male than female buds (Fig. 8B).Samples were collected during 2nd week of May, when temperatures were as high as 41 °C and 2nd week of July (after summer when temperature drops to 32 °C) 2017 from monoecious RG 156 and expression of 6 control genes were verified at two temperature conditions by semi-quantitative RT-PCR. Elongation Factor-1 Delta (EF1) gene was found to be uniformly expressing, without much fluctuations, in all tissues at higher and lower temperatures and therefore chosen as the control gene for temperature studies (Supplementary Fig. S7B). Expression of various candidate genes were verified in the same samples. Expression of SDR was higher in male buds of RG 156, (by nearly 1.2 times) than female buds, while WUSCHEL (WUS) was expressed only in differentiated shoot apical meristem (SAM) that has undergone floral differentiation and in male buds, but was totally absent in vegetative tissues such as undifferentiated SAM and leaves (Fig. 8C). DEFICIENS (DEF), also showed higher expression in differentiated SAM and male buds, like WUS, but unlike WUS, expression of DEF was not totally absent, but lower in other tissues. Expression of SDR in male buds and expression of DEF in differentiated SAM and male buds were also higher in July, when the temperatures dropped down, than during May. However, expression of WUS did not vary significantly with temperature fluctuations, though feeble expression noticed in female buds at high temperature conditions, was absent when temperatures lowered in July (Fig. 8C). The experiments were repeated and results re-confirmed ( Supplementary Fig. S7C). Expression of SDR and WUS were verified in various tissues of monoecious DCS 107 and pistillate DPC 9 collected by August-September 2017 (at normal temperature of 32 °C) from 3-months old crop. In monoecious DCS 107, expression pattern of SDR and WUS were similar to that of RG 156, while in DPC 9, expression of SDR was feeble in all tissues, and almost absent from female buds and WUS expression was totally absent from all tissues including differentiated SAM, (Fig. 8D). The complete gels and plot profiles generated by ImageJ are given in Supplementary Dataset 1: Supplementary Fig. S8). www.nature.com/scientificreports/  Table S6). The CREs in the 1 kb region for anther or pollen specific expression and temperature (cold)-responsiveness were found to associate in the SDR and DEF genes, but not for WUS (Fig. 8E).  8J) ACS and ACS-1 were closely related. But ACS of castor was closely related to that of Cucurbita sp and less related to other members of Euphorbiacea (Jatropha, Manihot and Hevea) and to ACS2 of Cucumis sativus and Arabidopsis thaliana (Fig. 8K).

Discussion
Genetic instability of females and unknown mechanisms of sex expression is a constraint in hybrid breeding programmes of castor 29 . To circumvent these constraints, inflorescence development and developmental mechanism of unisexual flowers of castor must be understood. The inflorescence development in castor has eight morphological stages from floral initiation to capsule setting and takes 15-20 days till anthesis. In Arabidopsis thaliana 12 organogenic stages were identified, while floral meristem initiation to anthesis was divided into 12 and 16 stages in cucumber (Cucumis sativus L.) and Fragaria × ananassa, respectively [38][39][40] . Castor is protandrous (male flowers are formed first in monoecious raceme), but female flowers open first indicating the preference for cross pollination in castor, thereby contributing to wide genomic variations.
The inflorescence architecture controls number and size of seeds thereby affecting success of both seed and pollen parents [41][42][43] . In castor, the inflorescence architecture varied with and within genotypes in response to high temperatures. Architectural variation along the inflorescence may be due to decline in vascular tissue along the inflorescence length 44 or due to hormonal gradient in the spike, determining the specific position of male, female or bisexual flowers in inflorescence. Auxin (IAA) has feminisation effect while ethylene, gibberellic acid (GA 4 ) abscisic acid (ABA) and jasmonic acid (JA) have masculinization effect in castor 26,28,29,45 . The variation within genotype is phenotypic plasticity in response to different levels of resources or developmental constraints 46 . The flower number and size decrease in distal inflorescences, in response to decline in resource availability over time. The main axis becomes longer and produces more lateral flowers in racemose inflorescence 47 .
In the present study, 8 and 12 developmental stages were identified in male and female flowers of castor respectively. Similarly, in unisexual flowers of Cedreleae and in Jatropha, 12 developmental stages were identified 48,49 . The male and female unisexual castor flowers were morphologically similar upto stage 4, unlike flowers of Jatropha which were similar till stage 6 49 or till stage 7 in Vernicia fordii 50   www.nature.com/scientificreports/ of bisexuality varied with genotypes, the pistil primordia being distinct in male flowers of M 574-OS1, but not in those of monoecious RG 156, while stamen primordia were distinct in female flowers of ISF line DPC 21, but not in those of pistillate line DPC 9. Earlier, rudimentary female organs were found to be absent in male flowers, but pistillate flowers were reported to have rudimentary stamens 51 . Female flowers of castor were believed to pass through bisexual stage 29 . We report here that both male and female flowers of castor have an initial bisexual state, the conspicuousness of which varied with genotypes and rudimentary female as well as male organs were observed in male and female flowers respectively. Rudimentary stamens occurred in female flowers, and rudimentary carpel was found in male flowers, but occurrence of bisexual flowers was occasional in castor.
In castor, sex is determined by selective growth arrest or abortion of either male or female reproductive organs, possibly through programmed cell death of opposite sex organs in bisexual primordia, similar to type 1 unisexual flowers 29,52 . Cells or primordia of both male and female organs arising initially, are selectively eliminated in female and male flowers respectively. Unisexuality but is the basic attribute in type 2 flowers 52 . Type 1 flowers are morphologically similar, while type 2 unisexual flowers, are morphologically distinct, where gynoecium and androecium are entirely absent from male and female flowers respectively 52 . Unisexual flowers of castor are morphologically distinct like that of type 2 flowers, with round male and elongated female flowers. Thus, unisexual castor flowers belong to an intermediate category between type 1 and type 2 flowers, having features of both. Castor flowers may be predominantly of type 1 category of flowers, since morphological distinctness is not precise in identifying sex of a flower, due to occurrence of slightly pointed male flowers and slightly round female flowers during summer.
In castor, sex determination takes place after an initial bisexual stage of each floral primordium, similar to maize and cucumber 2,53 . But whether programmed cell death (PCD) of preformed sex organs determines sex in castor 54 is not known, though PCD-related cysteine protease gene was expressed at peak of anther abortion in pistillate line 29,55 . DNA methylation also suppresses male flower-specific gene expression in pistillate inflorescences 29 .
In castor, growth arrest of pistil (in male flower) occurs earlier than that of stamens (in female flower), the stage of inhibition being dependent on genotypes. Pistil abortion occurred earlier than that of stamen in cucumber and Diospyros lotus 39,56 , but later than stamen in Asparagus officinalis 57 . In castor, stamen inhibition is clear in pistillate flowers and can be traced as a layer of cells or small bulges at base of ovary, unlike in Carica papaya where pistillate flowers do not show any traces of stamens 58 . Sex differentiation occurs at different stages after floral meristem development, such as flower organ primordia initiation (by developmental arrest of both stamen and pistil), organ differentiation, pre-meiotic and post-meiotic stages (by selective degeneration of opposite sex organ in unisexual flowers of only one sex type, but not the other) 36 . Sex differentiation in castor flowers occurs by growth arrest of opposite sex organs at stage 6 during organogenesis before meiosis.
Stamens can develop at a later stage after initial inhibition as seen in reversion to bisexuality in female flowers, but not pistil. After inhibition, pistil primordia do not develop into fully functional ovary, but can only develop simultaneously with stamens. Larger anthers in rudimentary stamens of bisexual flowers suggest incomplete divisions, where all developmental steps of male programming are not present. Similarly, rudimentary pistil with under-developed ovary and tubular style-like elongated structure results when female developmental programme is incomplete, altered and inhibited. Thus, growth arrest of the inappropriate organ does not uniformly affect the organ but occurs only in portions of the organs and all the arrested portions are spore-bearing parts 39 .
In monoecious plants, male as well as female programs, genes and signals co-exist in all cells and the sexual fate of cells is determined by sex-determining genes, their interaction within the cells and external factors, but there are no sex-determining loci 59,60 . Thus, all individuals are capable of developing flowers of either sex, where interaction and regulation of the sex-determining genes in response to environment results in sex-phenotype variations 60,61 . This explains for the sex lability and high sexual polymorphism observed in castor. Of the various factors affecting sex expression, high day temperature (T max ) was found to be major cause of sex alterations and reversions. Extreme temperature events (5 °C above the normal temperatures) of short-term durations may have adverse impact on plant productivity 62,63 . Low temperature favoured female flower differentiation, while high temperatures changed in inflorescence morphology and increased the proportion of staminate to pistillate flowers (4:1) in plants 64,65 . Plant growth and development are more associated to the thermal than chronological time and degree day's approach is widely used for thermal time quantification. In castor, sex phenotype alterations and sex reversions observed at extremely high day temperatures (above 33-34 °C) and degree days (above 19-20) during floral differentiation stage, was unidirectional towards maleness at high temperature, but femaleness was restored at lower or normal temperatures, indicating the role of temperature in determining sexuality and sex reversibility of castor flowers. High temperature causes maleness 22 . Reproduction is resource-intensive, females use more resources for reproduction than males 42 and female development pathway acts downstream to male developmental pathway. Therefore, female flower developmental program is entirely shut down and male flower development program requiring minimal resources is favoured, under high temperature stress conditions. Other factors such as order of branching, genotype (genetic factors) and physiological factors also influence sex expression. Ethylene, a stress hormone is produced under high temperature conditions. Auxin level at the differentiating apex which determines the sex balance of the flowers is in turn influenced by temperature, nutrition and photoperiodism 24 . Male flowers could be induced even at low temperatures of 30 ºC by removing female flowers 66 . The alteration in sexuality in castor is due to the effect of temperature rather than the higher branch order or ageing that occurs during the advancement of growing season, since femaleness is restored even at higher orders of branching, when temperature lowers after summer. The floral developmental pathway in unisexual flowers not only varies with temperature, but also with the genotype, since female flowers of DPC 21 had more distinct stamen primordia even at lower temperature conditions and later stages of development, when compared to DPC 9. Bisexuality is not a norm since bisexual flowers with equally developed male and female organs do not occur in all, but only in few genotypes, indicating that out-crossing results in genetic instability that alters www.nature.com/scientificreports/ developmental programming in inflorescence. Control of sex expression is thus frequently the control of floral organ development, stopping or starting an existing, functional developmental pathway 67 .
In monoecious plants, the successive induction of first male (protandrous) and then the female program, results from inverse gradients of male and female signals 59 . It is 'male first' programming in castor at inflorescence and individual flower level. Female organ development is more complex, requiring activation of more genes or regulatory networks for ovary, stigma and ovule development and for silencing genes of male organ development. More genes were expressed close to raceme formation in pistillate line when compared to monoecious line 29 . Genes involved in development of female organs act downstream to male organs but not simultaneously. A few critical or unique female development genes may act after stamen inhibition in female flowers or after activation of male developmental programme in bisexual flowers. Some genes in male and female flower development may have overlapping functions or crosstalk. Gene expression analysis shows that SDR, a short chain dehydrogenase reductase 2a was expressed in all tissue samples. but higher expression of SDR in male than female buds and in male buds of staminate than monoecious genotypes, indicates its significant role in pistil abortion in male flowers and maintenance of maleness, in otherwise bisexual floral primordia. SDR expression was lower at high temperatures of summer (May) and higher when temperature drops down in July, in male flowers. This may be because of general deregulation or lower expression levels of genes at extremely high temperature. Lower temperatures favour femaleness or pistil development, and hence SDR expression increases at low temperatures in male flowers, for abortion of the pistil primordium, thus maintaining maleness in male flowers and spikes of staminate genotypes. Even when bisexuality was observed in staminate genotypes at lower temperatures during June-July, the expression of SDR is still low in female flowers and increases in male flowers of monoecious genotype, indicting its significance in male flower development. Short-chain dehydrogenases/reductases (SDRs) constitute one of the largest and oldest protein superfamilies found in all domains of life, having 49 different families and are classified into 5 different types (classical, extended, intermediate, divergent and complex) , the diversified ones are involved in secondary metabolism and developmental processes, while the less diversified ones are involved in primary metabolism 68 . SDR2a in castor is a homologue of TASSELSEED2 (TS2) of maize, a short-chain alcohol dehydrogenase or hydroxysteroid dehydrogenase, which determines sexual fate of floral meristems in maize through sub-epidermal expression in the gynoecium primordium of male flowers, resulting in pistil abortion through programmed cell death 54 . Though all pistil primordia in maize express TS2 RNA, TS2-induced cell death in functional pistil primordia of primary ear florets is blocked by the silkless1 gene 69 . TS2 is expressed throughout the plant in maize, rice and sorghum and the expression of its homologues in different species vary such as in tapetal cells of male flowers in Arabidopsis thaliana and Silene latifolia and in all sampled tissues of grasses, suggesting that TS2 may not be involved in sex determination and may have a general developmental role 67,70 . SDR2a in castor, similar to TS2 in grasses is single copy, but there are many putative SDRs and secoisolariciresinol dehydrogenases in castor which belong to the same class as SDR2a. Although the stamen and pistil abortion for female and male flower development are under two different genetic pathways, considering the wide morphological diversity of sexuality and flowers in castor, SDR is unlikely the single master switch affecting transition of sexuality, suggesting the role of multiple genes in sex determination. SDR2a in castor may be one major gene player involved in sex determination in castor and with multiple roles, as indicated by the presence of CREs for light-responsiveness and inducibility to abiotic and biotic stresses as well as elicitors in its promoter region.
WUSCHEL (WUS) is a homeobox gene encoding homeodomain transcription factor required for the maintenance of stem cell homeostasis in the SAM and floral meristem (FM), expresses in a few cells in the L3 layer or stem cell-organizing center (OC) and is antagonized by CLAVATA (CLV) signaling 71,72 . WUS flowers display many fewer stamens (usually one or two) and no carpels, consistent with precocious FM termination 73 . WUS orthologue in castor is a single copy gene without any paralogues, expressed only in differentiated SAM and male buds, but absent from vegetative SAM, leaf and female buds of monoecious genotype and absent even from the differentiated SAM in pistillate genotype DPC 9, suggestive of its role in male floral development pathway. Expression of WUS in castor was similar to ROSULATA (ROA), a WUS homologue in Antirrhinum, where higher expression was noticed in inflorescence and stamens while expression was absent from stem and leaf and low in vegetative apex, petal and carpel 71 . WUS was expressed in SAM in Arabidopsis, in young leaf primordia in rice, and in reproductive meristems in maize 74 . Also, WUS expression did not vary with temperature fluctuations, which is also supported by absence of temperature-responsive CREs associated with anther or pollen-specific motifs in the 1 kb promoter region. WUS is a transcription factor which can act as a repressor as well as an activator. WUS represses certain genes in the SAM but activates them in the floral meristem (FM). WUS and AGAMOUS (AG, a class C homeotic gene in Arabidopsis) feedback loop starts with the activation of AG transcription at stage 3, and ends with the repression of WUS (by AG) in the centre of the FM, at stage 6 after carpel initiation, thus maintaining the WUS expression within the FM organizing centre 73 . Absence of WUS expression in female buds of castor indicates that floral meristem is not active after carpel primordia initiation which is also suggestive of parietal placentation in castor where placenta and ovules develop after the FM has terminated 73 . WUS binds to WUSATAg target sequence in the regulatory intron of AG, which causes repression of WUS in SAM and activation of AG in the floral meristem. Presence of WUSATAg sequence near to TSS (at 29-) and intron 4 (1582 +) of DEF and close sequence similarity of DEF with AG gene (78.15%), indicates overlap of functions DEF and AG genes of castor and that WUS may regulate DEF as well. DEF and AG of castor are closely related. Preliminary analysis of expression pattern of AG in castor showed upregulation of AG in differentiated SAM or inflorescence bud, female buds of monoecious and pistillate genotypes and male buds of staminate genotype (data not shown). DEFICIENS (DEF), a class B homeotic gene is necessary for development of petals in whorl 2 and stamens in whorl 3, along with APETALA1 (AP1) and AG respectively. Upregulation of DEF in differentiated or reproductive SAM and male flowers of castor is similar to that of expression pattern of WUS in castor, which however shows expression only in these. DEF A protein controls floral organogenesis www.nature.com/scientificreports/ in Antirrhinum majus and expression of DEF orthologues also varied in different species such as pMADS1 of petunia, a petal organ identity gene, expressed strongly in petals, but moderately in stamens, in root nodules in Medicago, in developing fruit of tomato, and in four floral whorls as well as in the leaves of orchid 75,76 . The class B gene (AP3) in a monoecious, diclinous species Vernicia fordii, was significantly increased in male flowers 50 . Comparative approaches in the MADS box gene family have provided some evidence that basic principles of the ABC model of flower development such as B-function genes are conserved in angiosperms 75 . GLOBOSA (GLO) is another class B gene and its homologue in castor (Gene ID: 255558565; XM_002520262.1) was found to be differentially expressed during raceme formation 29 . DEF and GLOBOSA (GLO) are hypothesised to have originated from a duplication event and duplication of DEF/GLO genes is an ongoing process 76 . ACC (1-amino cyclopropane-1-carboxylate) synthases are responsible for production of plant hormone ethylene. Nine ACC synthases were identified in castor. Expression pattern of two paralogues ACS and ACS-1 were similar, with higher expression in male than female buds, indicating its role in promoting maleness. Ethylene and ethylene-like substances promote maleness and can transform female flowers into male ones in monoecious plants 26 . In cucumber, ethylene promotes female flower by inhibition of stamen development, through organspecific DNA damage in the primordial anther of female cucumber flowers 77,78 . Downregulation of ethylene receptor in stamens of female flowers, carpel-dependent expression patterns of pre-miRs and M gene (CsACS2) in cucumber its ortholog in melon A gene (CmACS7) etc. were involved unisexual flower development [78][79][80] . In castor, whether organ-specific DNA damage in gynoecium of male flowers is triggered by ethylene and whether the ethylene receptors are differentially expressed or downregulated in pistil primordium of male buds is not clear. Although gene expression pattern and levels may not exactly reflect the functions of the genes, the variable expression pattern in the paralogues and orthologues of these genes in different species suggests the multiple functions of the ancestral genes and conservation of imperative functions during evolution of these genes. In situ hybridisation on SAM and floral meristem can further reveal the detailed expression of these genes. We report here 5 male-specific or male abundant genes in castor, viz., SDR2a, WUS, DEF, ACS and ACS-1, which may be involved in sex determination. A MYB-like gene, Male Specific Expression 1 (MSE1), specifically expressed in males in early anther development and tight linked with the Y chromosome, acts in sex determination in dioecious Asparagus officinalis 81 .
SDR and DEF in castor male flowers are regulated by temperature as evidenced by their expression pattern and presence and close-association of temperature-responsive elements with the anthe or pollen-specific motifs in the 1 kb promoter region.ACS genes also show similar association of CREs, indicating their temperature responsiveness. However, WUS was devoid of such temperature regulation of spatial expression. A predicted CRE in the 1 kb upstream region is potentially diagnostic of the regulatory function of that CRE 82 and hence we had chosen 1 kb region upstream of TSS for our analysis. Ten CREs viz., CACTFTPPCA1, DOFCOREZM, CAATBOX1, GT1CONSENSUS, POLLEN1LELAT52, ARRIAT, GATABOX, ROOTMOTIFTAPOX1, MYCCONSENSUSAT and EBOXBNNAPA were found to be predominant in the differentially expressed genes which were male-specific or male-abundant showing higher expression in male buds and staminate genotypes. Seven of these elements DOFCOREZM, CAATBOX1 GT1CONSENSUS, POLLEN1LELAT52, GATABOX, ROOTMOTIFTAPOX1, EBOXBNNAPA and other elements WRKY71OS, GTGANTG10 (pollen-specific), ACG TAT ERD1 and YACT were reported as 11 representative common elements in the 140 male gamete-and tapetum-expressed genes in rice with a frequency from 51.2 to 86.3% 83 . DOFCOREZM, CACTFTPPCA1 and CAATBOX1 were also the most abundant CREs in 414 Putative Promoter Regions (PPRs) of MLO (powdery Mildew Locus O) genes in plants 84 . Mere presence or predominance of certain cis-regulatory elements or motifs (CREs) in the promoter regions of genes do not necessarily imply their biological role. Some motifs such as ACGT may be frequently present in most plant gene promoters. The predicted or putative CREs could be short motifs that occur randomly throughout the genome without regulatory function and their biological significance can be determined through combinatorial or independent action of CREs by correlating their presence or absence to the expression profile of the genes 82 . The cis-regulatory codes specifying pCRE presence and absence, combinatorial relationships, location, and copy number were used to predict stress-responsive expression 85 . The cis-regulatory fingerprint helps in understanding the regulation of genes under various conditions. Presence of hormone responsive CREs in putative promoter regions of these genes are indicative of their regulation by plant hormones. 15 hormone-related genes involved in abortion of stamen during pre-meiosis of female flowers were identified in Litsea cubeba 86 .
Abnormalities in meiotic segregation of chromosomes or trisomy 12 may also result in sex phenotype alterations (male racemes with rarely occurring single terminal or few hermaphrodite flowers) in outcrossed pistillate lines. However, preliminary analysis of the mitotic chromosome number in different genotypes with distinct sex expression pattern like monoecious, ISF and pistillate, did not show any significant variation in chromosome number (2n = 20) (Parvathy et al. unpublished). Genomic instability due to outcrossing could also be mediated by repetitive DNA or transposons (> 50% of genome) or epigenetic changes associated with these repetitive sequences, which are abundant in castor 87 . www.nature.com/scientificreports/ almost all the genotypes, hence for desired sex phenotype, chemicals, hormones or other biotechnological tools can be employed when maximum number of flowers are at stage 4. The complex phenomenon of sex reversion in pistillate lines of castor is due to alteration in the developmental pathways of male and female flowers, a major mechanism governing sex expression and reversion. The alteration in sexuality is because of temperature rather than the branch order or ageing. Reversion or alteration of sexuality such as male to bisexual to female in flowers is also first time reported to change with temperature. Duration of thermal stress existing during critical stages for 2-3 weeks during inflorescence or floral meristem initiation and sudden variations (rise or drop) in temperatures alter sex expression, rather than the absolute value of T max . In addition to temperature, genetic factors also contribute to sex variations, since completely male inflorescences were observed even during normal temperatures of early growing season in the primaries of DPC9-OS2. Two genes, SDR2a and WUS (orthologs of TS2 of maize and WUS of A. thaliana respectively) were male-abundant and male-specific, the other three male-abundant genes being DEF, ACS and ACS-1. Knock out or silencing of these male organ-predominant or male-specific genes in bisexual stage of flowers by use of suitable promoters that drive expression during the inflorescence or floral bud initiation, will result in fully female flowers and there will be no revertants. Transgenics with knock out of these male-abundant or male-specific genes can be developed for desirable stable sex phenotype such as completely pistillate spikes or plants without reversion or ISF, even at summer. Also, the cis-regulatory regions or promoters of the male-specific genes can be cloned and used to drive expression of male-sterility or other genes in male flowers and used as additional tools for developing transgenics with desirable traits. The differentially expressed genes identified can be used as functional markers in marker-assisted selection or the gene sequences used in gene chips to identify male specificity. Understanding the morphogenetic and molecular regulation of floral developmental pathways will thus enable in devising effective chemical or biotechnological tools as well as strategies to regulate sex expression in castor. Inflorescence growth and architecture. The inflorescence buds at different stages after differentiation before and till complete emergence from bracts, located at the apex of different branch orders (secondary to quaternary) were tagged in field-grown plants of DCS 107, RG 156, DPC 9, DPC 21 and DPC 17-S3. The number of buds tagged for studies on inflorescence growth and stage transition are shown in Supplementary Table S8. Inflorescence growth and morphological stage transitions were monitored at regular time intervals of 3-4 days for 2-4 weeks. Inflorescence length (in centimetres, cm) was measured using a 30 cm metallic scale, the mean of observations at each stage calculated, and the growth tendency of inflorescence depicted in timeline graph. Number of days taken for transitions to each stage was recorded and the days to anthesis noted for each genotype.

Methods
For understanding the floral architecture, the inflorescences at different growth stages after complete emergence, but before anthesis were harvested from different branch orders, (secondary to quaternary) from fieldgrown plants of DCS 107, RG 156, DPC 9, DPC 17-S3 and M 574-OS1. The length of the inflorescence (in cm) was recorded and different floral whorls were removed in the order from bottom to top of inflorescence and arranged such that, bottom-most was first whorl and top-most was last whorl. The total number of floral whorls in each spike and flower buds (male and female) in each whorl with their positions were recorded and represented diagrammatically. The whorl number (counted from bottom of spike) in which female flower buds first appear and whorl number where all flower buds are female were noted for monoecious lines and whorl with male flowers in DPC 17-S3 (monoecious apical ISF) and M 574-OS1 were noted. To know whether floral whorls are added during elongation of inflorescence, pearson's correlation coefficient was calculated, for length of inflorescence and number of floral whorls, One-tailed t-test was performed (alpha = 0.05, 95% significance level) and compared with www.nature.com/scientificreports/ node, 3-5 buds per node), were collected from pot-grown or field-grown plants at ICAR-IIOR (Rajendranagar, Hyderabad, Telangana, India) over a 1-year period (2013-14) in FAA (10 ml formaldehyde, 5 ml acetic acid, 50 ml alcohol solution and 35 ml distilled water in 100 ml solution) and fixed overnight at room temperature or stored in FAA at 4 °C till use. The samples were rinsed with distilled water to remove fixative, dehydrated in graded series of ethanol [10,30,50,75,90, 100, 100% (with a drop of toluidine blue for staining the specimen)], two xylene treatments each for 1 h, infiltrated twice in paraffin wax, each for 1 h and embedded in wax. Wax blocks were prepared using hand-made paper boats and serial sections of 5 µm thickness were taken using handoperated microtome (Leica RM 2245, Lab India). The wax ribbons were gently placed on labelled, preheated slide smeared with Mayer's adhesive (1:1 v/v egg albumen: glycerol). The slides were dewaxed with xylene, placed in graded series of ethanol (100, 95, 70, 50 and 20%), stained with 0.7% toluidine blue, dehydrated in graded series of ethanol in the reverse order (20,50,70,95, 100, and 100%), air dried and mounted with D.P.X. mountant (Fischer Scientific, Mumbai, India). At least 3 biological replicates were included per node and 4-5 biological replicates were used in case of DCS 107 and DPC 9. Minimum 10 slides with 2 rows of sections (20 technical replicates) were analysed per specimen. Microtome sections were observed and photographed using a compound light microscope (H600L, Nikon, Tokyo, Japan) at 40X magnification. The images were captured in a series of frames and superimposed to reconstruct the whole section of the specimen.  www.nature.com/scientificreports/ Equal number of male and female buds were taken. The quality and quantity of isolated RNA were verified using agarose gel and Nanodrop ® spectrometer respectively, DNase I (Genetix)-treated and equal quantity of RNA (after adjusting volume according to the gel intensity or nano drop readings) was directly used in one-step RT-PCR using PrimeScript™ One-step RT-PCR kit (Takara Bio) for verification of expression of control (EF-1 and UBQ) and/or candidate genes as per manufacture's protocol. Alternatively, first strand of cDNA was synthesised using Superscript™ III (Invitrogen) as per the standard protocol and used in two-step RT-PCR to verify the equal expression of control gene in all samples. The same quantity of cDNA used to normalise control gene expression, was used for verifying expression of candidate gene in all the samples with 0.4-0.6 µM forward and reverse primers, 0.4 mM dNTP, 1 U of Taq DNA polymerase (Merck), 2 µl cDNA (1: 20 dilution) using the programme, initial denaturation at 94 °C for 3 min, 33 cycles of 94 °C for 1 min 50-60 °C based on T m of primersfor 1 min, 72 °C for 1 min and final extension at 72 °C for 10 min, in Eppendorf Mastercycler ® Nexus gradient PCR master cycler. The RT-PCR products were run on 1% agarose gels stained with ethidium bromide and the intensity of bands were quantified using ImageJ (https ://image j.nih.gov/ij/) [*Primers were synthesized for 17 candidate genes and gene expression verified in various genotypes, but based on consistency of results and reconfirmations 5 are reported].
In silico analysis of differentially expressed genes. The 1 kb sequence upstream to the putative or annotated TSS and the gene sequences of the differentially expressed genes were extracted from NCBI (ncbi.nlm.nih.gov) and used for identifying cis-regulatory elements (CREs) using New PLACE (https ://www. dna.affrc .go.jp/PLACE /?actio n=newpl ace) 89 . The CREs in the 1 kb putative promoter region of the genes were categorised into classes based on their function, their occurrence counted in the positive and negative strand and a heat map drawn using MS Excel version 2010. The gene sequences from NCBI-GenBank and protein sequences extracted from UniProtKB (https ://www.unipr ot.org) were used for nucleotide as well as protein blast, carried out using NCBI-BLAST (blast.ncbi.nlm.nih.gov) and phylogenetic trees were constructed using the BLASTp results by neighbour joining method.

Data availability
Relevant data generated or analysed during this study are included in this published article (and its Supplementary Information files). Data are however available from the first author, who is presently at ICAR-Indian Institute of Agricultural Biotechnology (ICAR-IIAB), Ranchi, India upon reasonable request and with permission of ICAR-IIOR, Hyderabad, India. www.nature.com/scientificreports/