Yellow lupine has a great potential as a substitute for soybean in animal feed production in Europe. An ability to limit the economic drawbacks resulting from excessive flower abscission would be the most convincing argument for lupine cultivation. This, however, can only be achieved if we understand the plant’s biology, especially the molecular basis for the development and maintenance of lupine flowers, of which little is known. Therefore, we believe that the pathways controlling these processes deserve intensive research focus. The commonly used NGS technique allows for comprehensive analyses of sRNAoms, transcriptoms, and degradomes, as well as identification of sRNAs and their target genes in both model and non-model plants, for example: tomato [58], soybean [59], Arabidopsis [60, 61], oilseed rape [62], Brassica juncea [63], lily [64, 65], peanut [66], orchardgrass [67] or wheat [68, 69].
Our previous analyses of yellow lupine transcriptomes resulted in the identification of transcripts of many genes involved in flower and pod abscission, and suggested sRNA involvement in this process [10]. What is puzzling, our observations of L. lupinus floral development indicate that their fate (abscission or pod formation) is determined prior to AZ activation. Therefore, we performed comparative analyses between sRNAs from flowers developing on upper and lower parts of the raceme. Identifying the miRNAs and their target genes involved in the above-mentioned processes will further our knowledge of the biology of not only lupines, but plants in general, since the role played by sRNA in organ abscission is still obscure.
Functions of miRNA/Target Modules During Yellow Lupine Flower Development
Involvement of sRNAs in flower development has undergone intense investigation and already been explored in many plant species, such as wild rice (Oryza rufipogon) [70], hickory (Carya cathayensis Sarg) [71], yellow-horn (Xanthoceras sorbifolia) [72], spring orchid (Cymbidium ensifolium) [73], and roses (Rosa sp.) [74]. Our sRNA-seq analyses have not only shed light on the molecular mechanisms that control flower development in one more plant species and confirmed the involvement of known miRNAs, such as miR159, miR167 or miR172, in this process [75], but we have also explored the roles of sRNAs in flower abscission and identified species-specific miRNAs.
In order to identify miRNAs and their target genes involved in flower development and abscission in L. luteus we conducted comprehensive analyses of ten sRNA and ten transcriptome libraries, and performed one degradome sequencing. The libraries were created for flowers in different developmental stages located in the top and bottom whorls of the raceme, and for flower pedicels with active and inactive AZs. This approach resulted in the identification of 394 conserved miRNAs belonging to 67 families, 28 novel miRNAs and 316 phased siRNAs. Additionally, we found their target genes and, consequently, were able to identify their functions. We also described miRNAs characteristic of the subsequent stages of flower development independently from the flower’s position on the raceme, and identified miRNAs present only in either upper or lower flowers, in the particular stages of development.
Known sRNAs and Their Target Genes are Involved in Regulating Flower Development in Yellow Lupine
We explored the large set of data concerning flower development that we obtained through sRNA-Seq analysis, focusing on three aspects: (i) conserved miRNAs involved in flower morphogenesis and development, (ii) expression dynamics during flower development, (iii) miRNAs specific for particular developmental stages.
Conserved miRNA families associated with flower morphogenesis and development
Among the known and conserved miRNAs we spotted a number of miRNAs commonly associated with flower morphogenesis and development, belonging to, inter alia, the MIR156/157, MIR159, MIR165/166, MIR167 and MIR172 families (reviewed in [26]).
Within the degradome data sets, we found target genes of two out of four differentially expressed L. lupinus miRNAs belonging to the MIR156/157 family (Ll-miR124 and Ll-miR401), encoding squamosa promoter-binding-like proteins 13B (SPL13B) and SPL2, respectively. Studies have shown that miR156 is necessary for maintaining anther fertility in Arabidopsis, by orchestrating the development of primary tapetum cells and primary sporogenous cells [76]. In this plant, SPL13B expression is strictly limited by miR156 to anther tapetum in young buds, while SPL2 is weakly expressed in parietal and sporogenous cells and the surrounding cell layers in young flowers [76], where it is targeted by miR156 to regulate pollen maturation [77].
The net of factors that regulate pistil development include miR156-targeted members of the SPL gene family, such as SPL3 and the miR156-resistant SPL8 gene, with both of them redundantly controlling gynoecium patterning through interfering with auxin homeostasis and signaling. What is striking, none of miRNAs identified in yellow lupine targets SPL3, which positively regulates the MADS box genes APETALA1 (AP1), FUL and LFY, central regulators of flowering [78]. In Arabidopsis, down-regulation of miR156-targeted SPLs resulted in a shortened style and deformed gynoecium [79]. Other studies on mutants have shown, that the aforementioned SPL genes cooperate with ARF3, a master regulator of gynoecium development and morphogenesis [80], which directly regulates transcription of several genes involved, among others processes, in auxin synthesis, transport, signaling and response [81].
miR159 was shown to target the conserved GAMYB-like genes that are a part of the GA signaling pathway [82, 83]. In A. thaliana, by affecting the expression of MYB33 (MYB DOMAIN PROTEIN 33) and MYB65, miR159 influences the LFY transcription level and gibberellin-dependent flowering induction. Moreover, the same miRNA also regulates proper flower development [84–86]. miRNA159 also regulates morphogenesis of the stamen, and male fertility [85], probably by restricting MYB33 expression by miR159 to stamen tissues during flower development [84].
Two transcription factors involved in pistil and stamen development, ARF6 and ARF8, contain the target site for miR167 [87, 88]. The roles of miR167 and ARFs are conserved across different species [89]. For Arabidopsis, it has been proven that both these genes are involved in stamen filament elongation, anther dehiscence, stamen maturation and anthesis [90], and that arf6 arf8 double mutants suffer from poor pollination [91]. By inducing jasmonic acid biosynthesis, ARF6 and ARF8 initiate anther dehiscence and flower bud opening and regulate perianth growth [91–93] However, these genes may also act in a jasmonic acid-independent way, by influencing the transcriptional activity of GH3, which results in a shift in the free auxin level [94, 95]. In tomato, a reduction in the accumulation of the miR167-targeted ARF6 and ARF8 leads to the lack of trichomes on the style surface, failed pollen germination and, consequently, sterility [96].
Recent research into multiple plant species has shown that miR172 targets genes belonging to the APETALA2 (AP2, TOE1, TOE2, TOE3) family. miR172 is part of the photoperiodic flower induction pathway and is associated with the functioning of the ABCE model of floral development [97]. Overexpression of MIR172 causes formation of a phenotype characterized by the absence of perianth, transformation of sepals into pistils and early flowering [97].
Our study showed the presence of at least one member of all these families in flowers (Fig. 2, Additional File 1: Table S7), which indicated that in lupine the families were crucial in generative development, as well. MIR156 and MIR159 are the most numerous families in Lupinus luteus, which suggests they play fundamental roles in its flower development processes.
Clusters of miRNAs depending on expression dynamics
The differentially expressed miRNAs identified in yellow lupine flowers were clustered by the dynamics of their expression (Z-scaled expression, to be precise) (Fig. 4). The first of the four clusters comprised miRNAs the accumulation of which increased as the flowers developed. The second cluster grouped all miRNAs the expression of which differed only in connection with the flower’s position on the inflorescence. The third cluster comprised miRNAs the expression of which dropped as the flowers developed. The fourth cluster included miRNAs characterized by a heterogeneous trend in expression (Fig. 4).
The first cluster contained miRNAs belonging to the MIR166, MIR167, MIR319, MIR390 and MIR395 families. The first of these families includes Ll-miR177, which guides the cleavage of RADIALIS, a transcription factor from the MYB family that controls the asymmetric flower shape [98, 99], as well as Ll-miR258 and Ll-miR265, which probably target the Homeobox-leucine zipper protein ATHB-15. In Arabidopsis, both miR165 and miR166 target the same HD-ZIP III genes: ATHB15, ATHB8, REVOLUTA (REV), PHABULOSA (PHB) and PHAVOLUTA (PHV) to regulate gynoecium and microspore development [21, 100–103]. An excessive expression of miR165 in the MERISTEM ENLARGEMENT 1 (MEN1) Arabidopsis line causes a decrease in the expression of all of the aforementioned HD-ZIP III target genes and results in SAM disruption, which may take the form of an enlarged apical meristem, a shortened carpel and female sterility [104]. miR166 has also been shown to control embryonic SAM development in rice and maize [105, 106]. The MIR167 family members that accumulate in larger quantities during flower development are Ll-miR280, Ll-miR281 and Ll-miR285, targeting ARF6 and ARF8. In the thermo-sensitive wheat line, an increased tae-miR167 expression in spikelet tissue generated the phenotype of cold-induced male sterility [107]. Similarly in citrus, nineteen members of the MIR167 family, including the diverse isomiRs, have been identified, whereas most of them displayed a significant up-regulation in the anthers of male sterile mutants [42]. Ll-miR445 and Ll-miR130 are members of the MIR319 family, while their putative target genes are TCP4 and MYB33, respectively. In Arabidopsis, the miR319a/TCP4 regulatory module is necessary for petal growth and development. Moreover, the overexpression of MIR319 reduces male fertility, and this defect is hypothesized to be caused by the cross-regulation of MYB33 and MYB65 by miR319 and miR159. As the miR319 target site within the MYB33 and MYB65 transcripts exhibits a lower match with miRNA than the miR159 target site, the latter is more efficient at regulating these genes and miR319 is their secondary regulator [108]. This regulatory network is even more complex. In A. thaliana, cooperation of three miRNAs and their target genes, namely miR159/MYB, miR167/ARF6/ARF8 and miR319/TCP4, is a prerequisite for proper sepal, petal and anther development and maturation. miR159 and miR319 influence the expression of MIR167 genes, which in turn affect each other. These miRNAs orchestrate plant development by regulating the activity of the phytohormones GA, JA and auxin [109]. An increased accumulation of miR167 and miR319 in the late stages of yellow lupine flower development could also be associated with regulating the growth and development of petals and anthers, as anthers in stage 3 of their development are already mature and split. Another miRNA showing a similar expression profile is Ll-miR9/miR390-5p. It targets the TAS3 transcript, which in turn is a source of tasiR-ARF, a negative regulator of ARF2, ARF3 and ARF4 activity. This regulatory cascade plays a vivid role in plant development [110, 111]. The expression level of miR390 derived from MIR390b reflects auxin concentration in organs, while the repression of ARF2, ARF3 and ARF4 by tasiR-ARF is important for lateral organ development, including roots and leaves [11, 112–114], and flower formation [115]. Down-regulation of ARF2 using RNA interference (dsRNAi) in Arabidopsis leads to delayed flowering, rosette leaf senescence, flower abscission and the opening of siliques [116].
Ll-miR118 and Ll-miR119, which target ATP sulfurylase 1, belong to the MIR395 family. In Arabidopsis, miR395 targets two gene families, ATP sulfurylases (ATPS) and sulfate transporter 2;1 (SULTR2;1), which are elements of the sulfate metabolism pathway [117]. ATPS regulates glutathione (GSH) synthesis and is an essential enzyme in the sulfur-assimilatory pathway [118, 119]. In cotton, the miR395-APS1 module is engaged in drought and salt stress response [120]. In B. juncea [63], miR395 presumably down-regulates APS1 to cause an increase in GSH expression and suppress the oxidative stress. Sulfate is the main source of sulfur and is taken up by roots, transported throughout the plant and used for assimilation. Sulfate limitation forces a significant up-regulation of miR395 expression [121]. Consequently, during yellow lupine flower development the demand for sulfur probably increases, and the plant activates mechanisms for its efficient uptake.
Within the third cluster of miRNAs, the expression of which decreased as the flowers developed, there were homologues of miR390-3p, miR858, miR396-3p, miR168, miR408-3p and miR398 (Fig. 4). Ll-miR99, Ll-miR100 and Ll-miR102 are identical to miR390-3p (the so-called passenger strand, former star strand) and, probably similarly to miR390-5p, are able to guide TAS3 cleavage. However, their expression showed an opposite trend to that of miR390-5p. A target gene of Ll-miR115/miR858 is MYB5, engaged in regulating seed coat development and trichome branching [122], although in Arabidopsis miR858 modulates another set of MYB genes, which are flavonoid-specific and control pathogen defense [123]. Another miRNA from the third cluster is Ll-miR155/miR396-3p (passenger strand), which guides cleavage of JMJ25 demetyhylase mRNA (confirmed in degradomes), involved in preserving the active chromatin state [124]. ECERIFERUM1 (CER1), the target gene for another two homologues of miR396-3p, Ll-miR199 and Ll-miR200, is a homologue encoding an enzyme involved in alkane biosynthesis, and in cucumber is engaged both in wax synthesis and ensuring pollen viability [125]. This cluster also included a miRNA that negatively regulates elements involved in miRNA and ta-siRNA functioning, namely Ll-miR247/miR168 targeting AGO1 mRNA [126]. Another miRNA clustered here was the highly conserved Ll-miR60/miR408-3p, which guides the processing of the mRNA of the copper-binding Basic Blue protein homologue (plantacyanin, PC). In Arabidopsis, PC plays a role in fertility, exhibiting the highest expression in the inflorescence, especially in the transmitting tract. Transgenic A. thaliana plants overexpressing the gene encoding this protein suffered from severely reduced fertility due to the lack of anther dehiscence and an improper twisted path of wild-type pollen germination [127]. What is puzzling is that in yellow lupine the expression of miR408 was high in stage 1 lower flowers and stage 2 lower and upper flowers, where both the anthers and the transmitting tract were immature, only to decrease in the subsequent stages. It is possible that in lupine PC is suppressed in an alternative way, or miR408 expression actually rises in the anthers and the pistil, although in a very local manner, and this trend is unnoticeable due to the high background from other tissues where the expression profile of this miRNA may be opposite. Especially so, since it has been shown that miR408 plays a role in other Arabidopsis organs, as well. Transgenic Arabidopsis plants overexpressing MIR408 displayed altered morphology, including significantly enlarged organs, resulting in enhanced biomass and seed yield. Plant enlargement was shown to be primarily caused by cell expansion rather than cell proliferation, and in transgenic plants it was correlated with stronger accumulation of the myosin-encoding transcript and gibberellic acid (GA). In addition, this plant line exhibited a higher concentration of copper in chloroplasts and an increased level of plastocyanin (PC) expression, resulting in a significantly higher photosynthetic rate [128].
miRNAs specific for particular developmental stage
Among the miRNAs identified in yellow lupine we spotted several that seemed to be crucial in particular stages of the plant’s development (Table 4, Additional File 1: Table S9). For example, the largest quantities of miR159 (Ll-miR452 and Ll-miR454) were accumulated in stages 2 and 3 of the plant’s development. They targeted Gamma-glutamyl peptidase 5 of an undefined function in plants, and an evolutionarily conserved target for GAMYB, respectively. As already mentioned, this could be associated with miRNA family cooperating with miR167 and miR319 in regulating L. luteus anther maturation early on in flower development. The accumulation of homologues of miR5168, miR1861, miR369-5p and miR5794 increased in stage 2 upper and lower flowers, while – interestingly – in the later stages these miRNAs were only present in lower flowers. According to degradome analysis, Ll-miR251/miR5168 guides cleavage of the mRNAs of the genes encoding the Homeobox-leucine zipper protein ATHB-14 and the chaperone protein dnaJ 13. The miR5168 sequence displays a great similarity to that of miR166, thanks to which they may perhaps share the same target gene ATHB-14, the putative transcription factor engaged in the adaxial-abaxial polarity determination in the ovule primordium [129]. As confirmed by yellow lupine degradome sequencing, Ll-miR229/miR396-5p targets Growth-regulating factor 5 (GRF5) and GRF4 transcripts. In Arabidopsis, miR396 plays an essential role in leaf growth by down-regulating GRF, which in turn interacts with GRF-interacting factor 1 (GIF1) to take part in cell proliferation [130]. In this plant, GRF5 is expressed in anthers at early stages of flower development and in gynoecia throughout the whole flower development, and transcripts of GRF4 accumulate later in sepals, tapetum, and endocarpic tissues of ovary valves [131]. Additionally, miR396 impacts the carpel number by targeting GRF6 [132]. Transgenic rice with OsmiR396 overexpression and GRF6 knock-down suffers from open husks and sterile seeds [133]. GRF6 cooperates with GRF10 to transactivate the JMJC gene 706 (OsJMJ706) and CRINKLY4 RECEPTOR-LIKE KINASE (OsCR4) responding to GA, which is a prerequisite for the flower to successfully develop into a normal seed [133]. The presence of these miRNAs in yellow lupine flowers suggests that they regulate cell proliferation mainly in the pistil and the developing ovules. An increased share of miRNAs involved in cell division, namely miR396, miR319 and miR164, in NGS analyses was also observed in early grain development in wheat [69].
Involvement of New miRNAs in L. luteus Flower Development
Using ShortStack [52] software we predicted 28 candidates for new miRNAs (Table 3). Fifteen of them were new members of known families, for example MIR167 (Ll-miRn12 and Ll-miRn27), MIR172 (Ll-miR4), MIR393 (Ll-miR19) or MIR169 (Ll-miRn3, Ll-miRn11 and Ll-miRn15) (Additional File 1: Table S8). The other 13 had no homologues among known miRNAs and were recognized as lupine-specific miRNAs.
Some of the new miRNAs displayed differential expression during L. luteus flower development. Ll-miRn3, which shows similarity to pre-miR169, displayed differential expression in UF1 vs LF1 and LF2 vs LF1 library comparisons, wherein it is the most accumulated in LF1, and in flower pedicels (up-regulated in FPNAB). According to degradome data, this miRNA targets SCARECROW2 (SCR2) homologue, a putative activator of the calcium-dependent activation of RBOHF that enhances reactive oxygen species (ROS) production and may be involved in cold stress response [134]. In rice roots, the SCARECROW genes play a role in cellular asymmetric division [135]. In this plant, SCR2 expression is relatively high in flower buds and flowers, and after flowering rises in the leaves and roots [136]. In yellow lupine, this gene may be involved in intense cell divisions during early flower development and is down-regulated in the pedicels with an active AZ to stop its growth. Another frequently encountered DEM, Ll-miRn10, shows differential expression in LF2 vs LF1 and UF2 vs UF1 comparisons, being up-regulated in stage 1 flowers. Its target gene is CIPK6 that is engaged in calcium-dependent CBL kinase activation, which in turn regulates potassium channel activity [137]. Ll-miRn22, which shows sequence similarity to pre-miR1507, is up-regulated in LF3 vs LF2 and LF2 vs LF1 library comparisons, and its expression escalates with flower development in the bottom whorl. The MiR1507 family is annotated as legume-specific [138]. Through analyses of our degradomic data we did not find its target gene, and the psRNATarget hit was the putative disease resistance RPP13-like protein 1. Unfortunately, this protein has been poorly described, therefore it is difficult to determine its function in yellow lupine flowers.
Noteworthily, the target genes of Ll-miRn1 and Ll-miRn30 identified through degradome sequencing are SGS3 and DCL2, respectively, and the miRNAs are up-regulated in LF3 vs LF2 comparisons and down-regulated in UF1 vs LF1 comparisons, respectively. SGS3- and DCL2-encoded proteins are involved in sRNA biogenesis [139, 140]. Importantly, novel miRNA identified in soybean Soy_25 displays a high sequence similarity to Ll-miRn1 and also targets SGS3, which indicates that this regulatory feedback loop for sRNA biogenesis is common for Fabaceae [141]. These results indicate that L. luteus miRNAs play a regulatory role in siRNA biogenesis in early flower development.
miRNA Accumulation Varies in Lower and Upper Flowers in Different Stages of Development
One of our goals was to identify the sRNAs engaged in yellow lupine flower development, with a particular emphasis on the differences between flowers from lower and upper parts of the inflorescence, in order to gain an insight into how early the flower fate is determined.
In our study we spotted differences in miRNA accumulation patterns as early as the first stage of flower development. Lower flowers in this developmental stage exhibited higher expression of miR396-3p, miR393-3p, miR858, miR390-3p, miR167-5p, and miR319. From the second stage until the end of their development, upper flowers accumulated more miR319, miR394, miR160, and miR393 (Fig. 4, Table 5). The elevated expression levels of these miRNAs suggests a reduction in the abundance of the transcripts of their target genes encoding auxin receptors and auxin response factors. This in turn may have led to a reduction in auxin sensitivity, and through decreasing the number of transcription factors belonging to the TCP family, probably caused different cell proliferation profiles in comparisons of flowers collected from the upper and lower whorls.
miR393 participates in regulating several developmental processes in the leaves [142], and contributes to root architecture development [143], pathogen defense [144–146] and plant response to nitrogen limitation or changes in salinity [13, 143, 147, 148]. miR393 regulates the accumulation of transcripts encoding auxin receptors belonging to the TAAR family, which, in the presence of auxin, guide proteasomal degradation of AUX/IAA repressors, thus enabling transcriptional regulation by ARF. Changes in receptor abundance affect the sensitivity of the given tissue to auxin and this is how this molecule influences plant development [149].
In A. thaliana, miR160 directly controls three ARF genes, namely: ARF10, ARF16 and ARF17 [150]. It is involved in many processes in plants, e.g. flower identity specification, leaf development or fruit formation [150, 151]. In tomato, sly-miR160 is abundant in ovaries, and changes in its expression affect plant fertility [152]. Down-regulation of sly-miR160 by introducing a target mimic caused improper ovary patterning and thinning of the placenta already prior to anthesis. Changes in gynoecium shape were followed by changes in fruit morphology [152]. Moreover, the ectopic expression of SlARF10A caused by a mutation in the target site for miR160 (mSlARF10A) resulted in a similar phenotype, with an additional strictly limited number of seeds that failed to germinate [153]. In view of these facts, the higher expression of miR160 in lupine lower flowers early on in their development means that a slightly different organization of the gynoecia may be one of the crucial determinants of flower fate.
Flowers collected from the lower whorls displayed higher accumulations of miR5490, miR5794, miR1861, miR396-5p, miR395, miR166, and miR159-3p (Table 5). miR1861 and miR396 were recognized as positive cell proliferation and development regulators [154–156]. In rice, for example, miR1861 exhibited differential expression during grain filling [157], and its expression was higher in superior grains in comparison to inferior ones [154]. This is consistent with our results, and a higher occurrence of miR1861 and miR396 in lower flowers may be an indication of the plant investing more supplies in this part of the inflorescence.
sRNAs are Involved in Flower Abscission in L. luteus
Little is known about sRNA engagement in flower abscission. Research into the involvement of miRNAs in this process has been already carried out in cotton [158], tomato [152, 159], and sugarcane [160]. Our investigation provided a more comprehensive view of this process and allowed us to explore cues for abscission present long before AZ formation.
For a genome-wide investigation of miRNAs involved in the formation of the abscission layer in cotton, two sRNA libraries were constructed using the abscission zones (AZ) of cotton pedicels treated with ethephon or water. Among the 460 identified miRNAs, only gra-MIR530b and seven novel showed differential expression in abscission tissues [158], and these miRNAs have no homologues in our dataset.
In tomato, sly–miR160 regulates three auxin-mediated developmental processes, namely ovary patterning, floral organ abscission and lateral organ lamina outgrowth [152]. Down-regulation of sly-miR160 and the resulting higher expression of its target genes, transcriptional repressors of auxin response ARF10 and ARF17, also resulted in the narrowing of leaves, sepals and petals and an impeded shedding of the perianth after successful pollination [152]. This was consistent with the higher accumulation of miR160 in upper flowers designated to fall off. As this miRNA showed no differential expression in flower pedicels, it probably does not play a role in the executory module of abscission itself, but is rather part of a mechanism that determines flower fate.
Another research on tomato using sRNA and degradome sequencing libraries explored the roles of sRNAs in AZ formation in early and late stages of the process, additionally accelerated or not by ethylene or control treatment [161]. The study showed that in tomato pedicels, when compared to the early stages of abscission, in comparison to the early stage the accumulation levels of, inter alia, miR156, miR166, miR167, miR169, miR171 and miR172 rose in late stages of abscission, while the abundance of miR160, miR396 and miR477 dropped [161]. Although it is difficult to compare ethylene-treated tomato pedicel results to our data, it is worth noting that in the corresponding FPAB vs. FPNAB comparison in our study, the accumulation of miR396 was lower, and the levels of miR167 and miR166 were higher, as well (Table 6).
It has been proven for sugarcane that miR156, miR319-3p, miR396 and miR408 take part in leaf abscission [160]. In their study, both mature (5p) and passenger (3p) miRNAs from MIR167 family were up-regulated in leaf abscission sugarcane plants (LASP), comparing to leaf packaging sugarcane plants (LPSP) (which corresponds to the FPAB vs. FPNAB comparison in our study). A total of four pairs of mature and passenger miRNAs were differentially expressed in this comparison, and they were members of the MIR156, MIR167 and MIR393 families [160]. Apart from the above-mentioned sugarcane miRNAs, miR396, miR395 and miR171 were up-regulated in this library comparison, as well. Four miRNAs, namely miR164, 166, 159, 319-3, showed an opposite trend [160]. In our study, both mature and passenger members of the MIR167 family were leaders among DEMs, too, (Table 6) pointing to their crucial role in both vegetative and generative organ abscission. Significantly, this applies to evolutionarily distant taxa: both monocots and dicots.
The situation is different for miR396. In contrast to sugarcane, passenger miR396-3p in yellow lupine is mainly accumulated in the pedicels with an inactive AZ (Table 6). It is also abundant in flowers from the lower whorls (Table 5), which suggests that in yellow lupine miR396-3p and miR396-5p (present in LF3) are engaged in maintaining flowers on the plant. This is probably related to the fact that this miRNA* may have different target genes in different organisms due to the higher rates of change characteristic of passenger miRNAs [162]. Thus, it is possible that in L. luteus miR396-3p targets other genes than it does in sugarcane.
Possible miRNA-dependent Regulatory Pathways that Participate in Both Development and Abscission in Yellow Lupine Flowers
Auxin-related sRNAs Involved in Regulating Flower Development and Abscission
Recent studies have shown that sRNA activity is associated with hormonal regulation of plant development through influencing the spatio-temporal localization of the hormone response pathway [163].
The auxin signal transduction pathway mainly consists of the three elements discussed below. Auxin is perceived by members of the TAAR family, namely TIR1 and its homologues AUXIN SIGNALING F-BOX PROTEIN (AFB), F-box proteins comprised in the SCF complex (SKP1-Cullin -F-box-RBX1). Downstream these receptors there are AUX/IAA repressor proteins and ARF transcription factors [151, 164, 165]. The expression of TAAR receptors is regulated by miR393 and secondary ta-siRNA derived from their own transcripts [13]. miR167 and miR160, by affecting the ARF6, ARF8 [88] and ARF10, ARF16, ARF17 [166] transcript accumulation, respectively, are able to modulate the expression of GH3-like [167] and, by this, influence auxin-regulated plant development. It has been proven that the expression of ARF2, together with ARF3 and ARF4, is regulated by the ta-siRNA/miR390 module [168]. In the two-hit model, ta-siRNA-containing the TAS transcript is recognized by two miR390 molecules, one of which guides its cleavage, and the other, in a complex with AGO7, serves as a primer for complementary strand synthesis, with its subsequent processing ultimately resulting in ARF-targeting siRNA biogenesis [169].
In our study, among the differentially expressed sRNAa in flowers and flower pedicels, there were members of the MIR167, MIR160, MIR393 and MIR390 families, as well as phased siRNAs targeting ARF2, ARF3, and ARF4. This fact suggests a vivid role of auxin-related sRNAs in flower development and abscission in L. luteus and confirms our previously published results of transcriptome-wide analyses [10], where we observed differences in expression levels of genes encoding several elements of the auxin signal transduction pathway.
When comparing organs to be abscised with those to be maintained, Ll-miR281 belonging to the MIR167 family was down-regulated in the UF1 vs LF1 comparison, but in the FPAB vs FPNAB comparison this miRNA, together with Ll-miR276, Ll-miR280, and Ll-miR39, was up-regulated. During flower development, the accumulation of Ll-miR285 changed as the flower at the top of the inflorescence entered stage 2 (UF2 vs. UF1) and, subsequently, the expression of Ll-miR280 and Ll-miR281 shifted as it entered stage 4 (UF4 vs. UF3). The relatively high number of members of the MIR167 family showing differential expression in the studied variants indicates that miR167 is one of the key regulators of flower development and abscission in yellow lupine.
Ll-miR224, a member of the MIR393 family, which guides cleavage of TIR1 mRNA , was identified as differential and unique for the upper part of lupine inflorescence, which makes it a molecular marker of flowers fated to fall off.
Three members of the MIR160 family displayed differential expression, namely Ll-miR332 and Ll-miR329 in UF2 vs LF2, and Ll-miR332 and Ll-miR333 in UF3 vs LF3, all being up-regulated. As already mentioned, these miRNAs may promote flower abscission.
Strikingly, our sRNA-seq analyses showed differential expression of three miRNAs identical with miR390-3p deposited in miRbase for other plant species. These miRNAs (Ll-miR99, Ll-miR100 and Ll-miR102) were down-regulated in all of the following comparisons: between upper vs lower flowers (UF1 vs LF1, UF4 vs LF4), between pedicels (FPAB vs FPNAB), and between flowers in different stages of development (LF2 vs LF1, UF3 vs UF2). There was one exception, namely that of Ll-miR9, which was identical to, inter alia, aly-miR390a-5p or ath-miR390a-5p (miRBase), which was up-regulated in the LF3 vs LF2 and UF4 vs UF3 comparisons. The differential expression and functioning of passenger miRNAs has already been described. The research carried out by Xie and Zhang 2015 on cotton showed that the formation of some miRNA*s, such as miR172* and miR390*, was associated with the phases of the plant’s growth. These miRNA*s were up-regulated in seedlings, but down-regulated in other growth stages [170]. Therefore, miRNA*s can be specifically expressed in various tissues to maintain the steady state of the organism. Our degradome analysis showed for yellow lupine that Ll-miR9/miR390-5p was able to guide the cleavage of the TAS3 transcript. There is no certainty as to the status of its passenger strand, and further research is required in order to identify its accumulation and function in the organs concerned.
ARF2, ARF3 and ARF4 are possibly down-regulated in the processing that is guided by Ll-siR249 and Ll-siR308, which were shown to be differential in the LF2 vs LF1 and UF2 vs UF1 comparisons (Table 4), and which are identical to tasiR-ARFs in many plant species according to the tasiRNAdb database [171] (data not shown). These tasi-ARFs probably originate from TAS3 transcript (TRINITY_DN55534_c4_g1) containing two binding sites for miR390 (data not shown). Ll-miR9/miR390, differentially expressed in LF3 vs LF2 and UF4 vs UF3, and, surprisingly, also Ll-siR240, guide cleavage of another TAS3 mRNA (TRINITY_DN54998_c6_g5_i2) (Figure S8) which contains only one target site for miR390 (data not shown). This is the first report on TAS3 processing regulated by siRNA. The target site for Ll-siR240 is shifted by 10 nucleotides relative to the target site for Ll-miR9/miR390 (Additional File 2: Figure S8). The expression of Ll-siR249, Ll-siR308 and Ll-miR9 showed a similar profile, as it rose during flower development and was the highest in the pedicels (Fig. 7). Ll-siR240 accumulated proportionally to TAS3, which means that it was least expressed in the pedicels, while in flowers its expression increased with time (Additional File 2: Table S21). The identified target transcripts belonging to the ARF2, ARF3 and ARF4 families showed organ-specific differential expression, which indicated their strongly local regulation by siRNA (Additional File 2: Table S21). The presence of all the elements of the miR390/TAS3/tasiR-ARF module among the DE sRNAs in yellow lupine suggests that alterations in its functioning have a great impact on L. luteus flower development. The additional element in the form of siRNA that processes TAS3 mRNA seems to be a new species-specific adjuster of this regulation module.
Function of the Potential miRNA Targets
Our functional analysis of the putative targets identified for miRNAs in lupine indicated their engagement in regulating a number of metabolic – especially ‘carbohydrate metabolism’ and ‘nucleotide metabolism’ – pathways (Additional File 5: Figure S4). ‘Carbohydrate metabolism’ was also one of the most enriched KEGG pathways in our previous Lupinus luteus transcriptome analysis [10], and its activation may be an indication of the rebuilding of cell walls or changes in nutrient supply. The next most numerous group of miRNA targets was categorized into the ‘Genetic information processing’ KEGG pathways, namely, ‘spliceosome’, ‘RNA transport’ and ‘ubiquitin proteolysis’. This suggests that in yellow lupine flowers most miRNAs regulate processes related to post-transcriptional events and protein degradation. Three KEGG categories within the ‘Environmental information processing’ category are extremely important in terms of plant development, and they are ‘Signal transduction pathways’ comprising the MAPK cascade, ‘phosphatidylinositol’ and ‘plant hormone’ signaling pathways (Fig. 9, Additional File 6: S5, Additional File 7: S6, Additional File 8: S7). The MAPK pathway is involved in regulating several processes, such as biotic and abiotic stress response (reviewed in [172, 173]), and associated with the functioning of hormones such as ethylene [174] and abscisic acid, engaged in organ abscission and other processes (reviewed in [175, 176]). The MAPK cascade is also an element of the positive feedback loop amplifying the abscission signal [177]. Auxin is not the only hormone whose signal transduction pathway is regulated by sRNAs in yellow lupine. Our KEGG enrichment analyses of the identified target genes for lupine miRNAs indicated that the signal transduction pathways of gibberellin, cytokinin, the already mentioned ethylene and ABA were potentially modulated by miRNAs in yellow lupine, as well (Fig. 9). These data show how important sRNAs are for growth and development regulation.
We also conducted a GO enrichment analysis (Fig. 8, Additional File 2: Table S13). Within the ‘Cellular component’, 83.3 % of the target genes were categorized as ‘cell’, with only 4.7 % categorized as ‘extracellular region’ and 3.9 % as ‘cell junction’. This means that miRNAs impacted processes mainly occurring within the cell. As for the ‘Molecular function’, most of the targets fell into the ‘binding’ (67.5 %) and ‘catalytic activity’ (52.2 %) categories. Thus, in yellow lupine flowers, miRNAs were found to most frequently influence molecular interactions and biochemical reaction rates. Within the ‘Biological Process’ group, most of the targets were categorized as ‘cellular’ (72.1 %) and ‘metabolic process’ (63.6 %). What is most interesting is that quite a considerable number of the target genes fell within the ‘response to stimulus’ and ‘signaling’ categories, which means that miRNAs modulated the way the plant adapted to environmental stimuli. Surprisingly few genes were categorized into the ‘growth’ and ‘cell proliferation’ categories, which indicated that these processes were probably regulated by other factors.
An in-depth analysis of KEGG and GO terms concerning plant hormones (Fig. 8b) showed that most of the miRNAs identified in yellow lupine modulated more than one hormone signaling pathway. For example, in a GO analysis performed exclusively for the targets exhibiting the ‘Flower development’ term (Fig. 8c), Ll-miR181 belonging to the MIR166 family modulated processes associated with four hormones, namely auxin, gibberellin, jasmonic acid and salicylic acid, by targeting not only transcription factor AS1, a central cell division regulator [178], but also Cullin-3A, an element of the ubiquitination complex [179]. Another two members of this family, Ll-miR173 and Ll-miR177, targeted the same gene, 26S proteasome non-ATPase regulatory subunit 8 homolog A (RPN12A), involved in the ATP-dependent degradation of ubiquitinated proteins during auxin and cytokinin response [180]. Similarly, a KEGG analysis for the MIR166 family showed that it was involved in the auxin, cytokinin, and brassinosteroid signal transduction pathways (Fig. 9).
Our GO analysis additionally showed for yellow lupine flowers that miRNAs were responsible for guiding the processing of target genes simultaneously involved in multiple processes associated with flower development (Fig. 8c). For example, AP2 is involved in the specification of floral organ identit [181], and ovule [182] and seed development [183, 184], and in our study it was targeted by ten lupine miRNAs: 8 belonging to the MIR172 family (Ll-miR25, Ll-miR26, Ll-miR27, Ll-miR28, Ll-miR29, Ll-miR30, Ll-miR140, Ll-miR171), one to the MIR166 family (Ll-miR249), and Ll-miR24. On the other hand, seven of these miRNAs additionally targeted a negative flower development regulator, chromodomain-containing protein (LHP1), which is a structural component of heterochromatin involved in repressing floral homeotic genes and FLT that regulates flowering time [185, 186]. This highly degenerated and ambiguous model of gene regulation by lupine miRNAs shows that in this plant the adjustment of key biological processes related to fertility is dependent on a complex network of interconnected factors.