Among the citrus species, several pummelos (Citrus grandis L. Osbeck) and mandarin-like varieties are self-incompatible [1]. Clementine mandarin (Citrus clementina Hort. ex Tan.), derived from an uncontrolled cross between a sweet orange and a mandarin, is probably the most widespread citrus species showing self-incompatibility (SI). It is characterized by gametophytic SI, with the pollen tubes stopping their growth in the upper or middle style [2, 3]. Moreover, in this species SI is coupled with a variable degree of parthenocarpy. SI and parthenocarpy in citrus are very important traits for fruit production because they result in seedless fruits, which have a higher value in the markets compared to the seeded ones. Therefore, understanding the molecular basis of SI would be important to plan marker-assisted breeding to obtain new seedless genotypes.

Despite the importance of this trait, the genetic basis are still poorly understood and the key genes of SI have not been identified yet. The study of populations segregating for SI might be definitely a powerful strategy to give new insights into its genetic basis. However, the difficulties to obtain and characterize appropriate populations, with a variable degree of parthenocarpy and female/male sterility observed in the progeny might limit this approach. Such kind of strategy for the identification of the S-locus was carried out analyzing several crosses among different citrus cultivar and accessions with Got-3 isozyme, which is thought to be linked with the S-locus [4], providing only a rough estimation of their possible S-genotype.

In recent times, different research groups attempted to better understand SI and pollen-pistil interaction in several citrus genotypes, mainly trying to characterize putative homologs of key genes and proteins of already characterized SI systems. Gentile and colleagues [5] reported the involvement of Ca²⁺-dependent transglutaminase (TGase) in the self-incompatible response in pummelo, as already reported for Rosaceae [6]. Regarding the S-locus genes, a S-like RNase has been isolated from 'Zigui shatian' pummelo [7], however the authors suggested that this gene might play an important role during ovary senescence rather than in the incompatibility response. Another S-like RNase has been isolated from a mandarin variety and was partially characterized [8], but it's not still clear whether this gene is the key determinant for the self-incompatible response.

To overcome these limits, the transcriptome analysis of natural mutants displaying contrasting compatibility behaviour might be more effective to better understanding the molecular basis controlling the progamic phase in citrus. Over the last decade, the genome and/or transcriptome analysis of natural or induced citrus mutants have been a powerful strategy to study the molecular basis of agronomically important traits, such as ripening period, fruit pigmentation, seedlessness and other traits related to quality [9–12]. Concerning SI, a few citrus natural mutants displaying different sexual behaviour with respect to their original varieties have been identified and characterized [13–16]. Differences between the mutants and the original cultivars were related to differences in pollen [13], style [15] or ovary [14] functionality. In some cases, different behaviours during the progamic phase were associated to abnormal embryo development [13, 16]. Therefore, it seems clear the mechanisms preventing fertilization are different in these genotypes, so it's reasonable to hypothesize that the mutations affected different genes or pathways implicated in reproduction.

Recently we chose two clementine clones with contrasting behaviour relating to self-pollen recognition ('Comune', self-incompatible; and 'Monreal', self-compatible mutation of 'Comune' [17]) as a model to identify candidate genes implicated in pollen-pistil interaction. Histological assays and analysis of pollen tube kinetics were performed to study the pollen tube behaviour in the two genotypes, and to assess whether the breakdown of SI in 'Monreal' was caused by changes in pistil or pollen functionality. The analysis demonstrated that the 'Monreal' mutation affected pistil functions, since pollen tubes of the two varieties grew equally in the pistils of self-compatible mutant, while 'Comune' rejected the pollen of both varieties, recognizing the pollen of 'Monreal' as self-pollen [15]. A first transcriptome comparison was conducted analyzing whole styles with stigmas, leading to the identification of a first set of genes differentially expressed in non-pollinated flowers and during pollen-tube elongation in self-pollination condition [15] including stress related genes, and transcripts related to Ca²⁺ and hormone signalling. Surprisingly, a relatively high number of gene tags of different classes of retrotransposons were isolated, indicating their possible activation in response to pollination. However, the cDNA-AFLP analysis covered only part of the transcriptome and, due to the presence of pollen tubes growing along the pistil, the isolated gene tags were presumably not pistil-specific.

Here we describe a complementary approach to identify another set of candidate genes in 'Comune' and 'Monreal', to provide a better view of the molecular aspects related to citrus reproduction. We used laser capture microdissection (LCM), which has been efficiently used to investigate several aspects of plant reproductive biology [18–21]. In our approach, LCM was coupled to microarray analysis to identify genes specifically expressed in the stylar canal cells (SCC). We demonstrate that our transcriptomic survey of SCC is an efficient strategy to discover candidate genes involved in the SI response in clementine.

In this study, LCM was used to spefically isolate SCC of two clementine genotypes differing for the SI response. 'Comune' is a widespread self-incompatible variety, while 'Monreal' is a self-compatible variety originated from a spontaneous bud mutation [17]. The microdissection of the SCC allowed to perform a highly specific study of the transcriptome of the cells implicated in the interaction between pistil and pollen tubes, with the main aim of identifying candidate genes involved in self-pollen rejection. The results of microarray analysis suggested that the differential regulation of few specific transcripts might have lead to the breakdown of SI in 'Monreal'. Based on functional information retrieved from the databases, these genes do not show any clear functional enrichment, and many of them have no annotation, making it difficult to compare our results with SI systems characterized on model plants. On the other hand, our experiment provided a new set of transcripts that are very likely to play a key role during the progamic phase in citrus. Searches in the databases revealed that most of these genes are not flower specific, suggesting that probably the mutation leading to the breakdown of SI did not affected the S-locus determinants. However, it is known that non S-locus genes are implicated in SI [30]. Most of the preferentially expressed genes were low-represented in EST databases and showed a weak expression in whole styles with stigmas before the SI reaction occurred. This indicates the usefulness of the LCM in the identification of highly specific and/or low expressed genes.

Flowers for LCM were sampled one day after pollination, when pollen was just germinating. Self-pollinations were performed to assess whether the presence of pollen in the stigma might have elicited a differential response of the SCC in self-compatible vs. self-incompatible conditions. So, while pollen tubes where still in the stigma, we targeted the rectangular SCC located in the upper/middle style, which is the site where incompatibility occurs in 'Comune', without contamination of pollen tubes. Although we did not analyze mRNAs of microdissected SCC of non-pollinated flowers, qRT-PCR demonstrated that slight differences in gene expression were also present in whole styles with stigmas before pollination, indicating that, at least in the cases of cit.7568, cit.11563, cit.5456 and cit.5776, the different mRNA levels in the SCC between 'Comune' and 'Monreal' were not induced by the pollen germination in the stigma.

Among the identified genes, our analysis focused on four unigenes over-represented in 'Comune' (cit.7568, cit.11563, cit.5456 and cit.5776). A time course analysis coupled to histological observation of pollen tube elongation was performed to correlate pollen tube behaviour with the difference in gene expression between 'Comune' and 'Monreal'. Until the 4th DAP we could not observe any clear differences in the mRNA level of the 4 analyzed genes, and no clear differences were evident in pollen tube behaviour between the two varieties. Form the 5^th DAP we observed an impressive up-regulation in the self-incompatible combination. Selected candidate genes showed a clear differential expression (even > 100 fold change) during pollen - style interaction (Figure 7), in agreement with histological observation (pollen tube growth in 'Monreal' and no growth in 'Comune'). The peak of up-regulation was evidenced in concomitance with the pollen tube arrest, suggesting that all the analyzed genes have a key role in the stop of pollen tube elongation. The up-regulation was clearly induced by pollen tubes in the styles, as confirmed with the comparison of the expression levels of non-pollinated and self-pollinated flowers at T6 (Figure 7). The candidate genes were weakly expressed in pollen tubes and no differences in the mRNA levels were observed between 'Comune' and 'Monreal' (Figure 8). As a result, it is unlikely that the transcriptome of pollen tubes influenced the mRNA levels detected in the time course analysis. In fact, our expression data demonstrate that the drastic changes in mRNA levels occurred in the stylar tissues as a response to self-incompatible pollen tubes. Also, the drastic up-regulation of the four genes should not represent a downstream response to SI, since these genes are already differentially regulated in SCC before SI occurs.

Bioinformatic analysis provided useful information to hypothesize the role of the four unigenes. cit.7568 probe set matched to a putative F-box protein gene sharing homology to an Arabidopsis F-box (At5g04010) annotated as non-specific [28]. This protein belongs to the C2 group of the F-box superfamily, of which SLY1 belongs. However, the similarity with SLY1 regards the F-box domain, while the C-terminus, responsible for targeting the substrate for ubiquitination, does not share any homology with SLY1.

F-box protein superfamily is one of the largest in plants [31, 32]. These proteins are key regulators of proteolysis, conferring specificity to the SCF (Skp1, Cullin, F-box) E3 ubiquitin ligase complex, which is responsible for recognizing the substrate for ubiquitin-mediated protein degradation. Despite the fact that it is not possible to correlate the role of the novel clementine F-box gene with any characterized orthologs, database searches and transcriptional data revealed that the clementine F-box gene was not previously identified in clementine cDNA libraries and evidenced highly specific expression patterns, since it was almost 250 fold up-regulated in SCC (Figure 2) and about 50 to 300 fold up-regulated in whole styles with stigmas (Figure 7A) in concomitance with the arrest of pollen tube elongation. This drastic up-regulation, and the absence of identical ESTs in databases support the high specificity of this gene and point out to its possible involvement in highly specific proteolytic events occurring in the style during the self-incompatible response. Studies on the families where SI system have been already characterized (namely Brassicaceae, Solanaceae and Papaveraceae) showed that proteolysis has a key role during self-pollen rejection. Usually proteolytic events follow the initial SI signal perception leading to the eventual death of the male gametophyte [33]. Functional analysis will be necessary to shed light on the role of the clementine F-box during pollen-pistil interaction.

Another striking finding was the identification of the three putative Asp-rich protein encoding genes, cit.11563, cit.5456 and cit.5776, up regulated in 'Comune'. In addition to the high similarity in their primary structure, they co-localize in the clementine genome (Figure 5). The clustered genes displayed similar expression patters (Figure 7B, C, D) which might indicate that they are commonly regulated as already reported for other clusters [34, 35]. Although their function has not been investigated, it is known that the expression of the Arabidopsis homologous At1g47400 is sharply up regulated in plants exposed to 3-(30,40-dihydroxyphenyl)-L-alanine (L-DOPA), a phytotoxic allelochemical [36]. In other studies, this gene turns out to be strongly induced by iron deficiency [37] and by treatment with the polycyclic aromatic hydrocarbon phenanthrene [38]. These results support the hypothesis that the genes encoding the Asp-rich proteins might be triggered in response to different types of stresses. Since other information is still lacking, we attempt to assign a putative function to the Citrus Asp-rich proteins and also propose a model which overall might explain the regulation of SI in the 'Comune' variety. Because of their richness of aspartic acid residues, the Asp-rich proteins are supposed to act as novel Ca²⁺ "entrapping" proteins. Although they do not show sequence similarity with other well known Ca²⁺ interacting proteins, such as calsequestrin, calreticulin and calmodulin, the Asp-rich proteins share with those proteins the aspartate residue abundance which has been related with the protein ability of Ca²⁺-binding [39]. Therefore, assuming that Ca²⁺ levels play a decisive role during pollen-pistil interaction, as already reported for several plant species [40], the up-regulation of the Asp-rich encoding genes observed in 'Comune' might enhance the amount of proteins functioning as Ca²⁺-trap elements, and lead to an exceptional decrease in Ca²⁺ availability, thus contributing to switch off the signal cascade usually induced by the increase in cytosolic Ca²⁺ concentration or to the alteration of Ca²⁺ gradient needed for pollen tube elongation. Pollen tube growth could represent, among others, the downstream physiological process dramatically affected by this missing triggering event. To support this hypothesis, it is worthwhile to mention that the aforesaid At1g47400 gene is up regulated in a T-DNA insertional mutant of a P-type ATPase cation pump, the MALE GAMETOGENESIS IMPAIRED ANTHERS (MIA); the mutant shows reduced male fertility and imbalanced cation homeostasis [41]. Moreover, a Ca²⁺ pump (auto-inhibited Ca²⁺-ATPase - ACA), which presumably pumps Ca²⁺ out the cytosol [42], is differentially regulated in the two genotypes during pollen-pistil interaction showing an up regulation in the 'Comune' genotype [15]. Further mandatory work will be undertaken to validate both the supposed role and the functioning model of pollen-pistil interaction in Citrus genotypes and, if they are proved, this might represent a case in which specific regulatory mechanism involving different loci rather than the S-locus could be co-responsible of the SI determination.

The integration of the transcriptomic data with the synteny analysis denoted a specific genome region containing a cluster of genes activated during self-pollen rejection (Figure 5). Specifically, the Asp-rich protein genes are linked to a DELLA gene which was previously isolated in the self-pollinated 'Comune' styles with stigmas [15] and showed a preferential expression in the self-incompatible genotype. This raise intriguing questions about the possibility that the non-homologous genes located in this genome region might contribute to a common function related to self-pollen rejection. Examples of co-expressed and functionally related gene clusters in eukaryotes have emerged over the last decade [34, 43] and are likely to increase with the huge amount of data coming from the genome projects. The most investigated operon-like organizations in plants are secondary metabolic pathways [44], mostly implicated in plant defence response. Clustering appears to have occurred de novo through some form of convergent evolutionary process [43]. In our case, it is unlikely that the genes are clustered by chance, since this cluster is conserved in at least two other plant species, cacao and castor bean (Figure 5). The collinearity of the citrus genome segment comprising the Asp-rich protein genes and DELLA with other segments of unrelated genomes support the hypothesis that selection might have favoured the linkage of these genes. The advantage of clustering is related with the fact that tightly linked genes might be co-regulated at the levels of nuclear organization and/or chromatin [43, 45]. The co-localization of the three up-regulated Asp-rich protein genes as well as DELLA in the scaffold 9 of the clementine genome v0.9 suggests that the mutation leading to self compatibility probably affected the functionality of tightly linked genes. Further transcriptional analyses will be carried out on the other genes surrounding the Asp-rich protein genes to study their possible role during different stages of the pollen-pistil interaction.

LCM coupled to microarray analysis was definitely a powerful tool to identify candidate genes involved in self pollen rejection which have not been previously associated to SI. Data at the transcriptome level strongly suggested that a restricted number of differentially regulated transcripts are associated with self-pollen recognition. Although functional information is missing, we hypothesize that proteolysis and Ca²⁺ homeostasis might be crucial for SI response in clementine, reflecting to some extent molecular events occurring in other SI systems. However, further work at the protein level will be necessary for understanding the role of the candidate genes during self-pollen recognition. Also, the unigenes represented in the GeneChip do not cover the whole transcriptome, consequently other strategies such as RNA-seq might help to improve the transcriptome coverage to identify additional genes implicated in self-pollen recognition.