EST sequences of J.curcas were downloaded from NCBI’s dbEST database (http://ncbi.nlm.nih.gov/) [36] which contains sequences generated from different tissue specific cDNA libraries of embryo, root, leaf and seed. These sequences were arranged in a single FASTA file, which was used for the sequence analysis using different softwares and Analysis Tools.

To find the singletons and to assemble the contigs from the total ESTs, an online tool “EGassembler” (http://egassembler.hgc.jp/) [37] was used. The main parameter provided was ‘Overlap Identity cutoff (N > 65): 85’. From the unigenes (singletons+contigs), EST-microsatellites [EST-SSRs] were searched using “SSRlocator version 1” (http://www.ufpel.tche.br/faem/fitotecnia/fitomelhoramento/faleconosco.html) [38] .

The SSR search was carried out for repeat motifs (ranging from mono- to hexa-nucleotides). For each repeat motif the parameters were: Mononucleotide repeat-20, Dinucleotide repeat-10, Trinucleotide repeat-07, Tetranucleotide repeat-05, Pentanucleotide repeat-04, Hexanucleotide repeat-04 (the numbers indicating repeat unit i.e. minimum number of times the motif was repeated at a stretch); Space between SSRs-100, Space between imperfect SSRs [<=]-05. After obtaining the motifs, the sequence complementarity was taken into consideration and accordingly the complementary motifs like AG and CT or AC and GT or AAC and GTT motifs were grouped into a single class under mono-, di-, tri-, tetra-, penta- or hexa-nucleotides, respectively. After getting SSRs, the primers were designed from the flanking regions using the same software as for SSR search. The parameters provided in the software for primer designing are given in Table 1.

EST Sequences, which have credit in the primer designing, were searched for their gene annotations using BLASTX at The Arabidopsis Information Resource (TAIR) (http://www.arabidopsis.org/index.jsp) [39] . This data was used to get the Gene Ontology (GO) Annotations and functional categorization of ESTs using locus identifiers at Bulk Data Retrieval System of TAIR (http://www.arabidopsis.org/tools/bulk/go/index.jsp) [40] .

The size of the available EST data used in this study has been calculated in accordance with the size of the

genome of J. curcas (C = 416 Mb) reported by Carvalho and coworkers [41] . The ESTs of J. curcas generated from tissue specific cDNA library of various tissues (viz. embryo, root, leaf and seed) available in the NCBI’s public database dbEST, were downloaded and pooled. These downloaded ESTs were inclusive of the seed specific ESTs generated in our laboratory. The pooled set consisted of 13513 ESTs (~6.2 MB) in all, which comprised of 9844 ESTs of embryo, 1000 of leaf, 1304 of root and 1375 of seed library. Using the EGassembler, all the sequences were categorized into singletons and contigs. The EGassembler segregated 13513 ESTs into 6098 singletons and 7415 redundant sequences. Then it assembled the redundant sequences into 1454 contigs. Therefore, through the software, the total ESTs were categorized into contigs and singletons, which were together grouped as 7552 (~3.8 MB) Unigenes. These data showed that the 45% of the total ESTs, downloaded from the database, were singletons and the rest 55% were assembled into contigs (Figure 1). The assembling of the redundant ESTs into contigs was beneficial in reducing the errors in sequence analysis in addition to removing the redundancy so that only the unigenes were used for SSR mining and for annotation. As reported by Raji and coworkers [18] , these unigenes, when used for the mining of SSRs result in a realistic estimate of the microsatellite repeat frequency and ensures that non redundant EST-SSR markers that correspond to unique loci in the genome are obtained. Therefore, in this study the unigenes were used for SSR search. The mining of the EST-SSRs starting with downloading of all the Jatropha ESTs is outlined in Figure 1.

For searching the SSRs, the repeat motifs in the software, were selected from mono- to hexa-nucleotide as

going above this motif range, the frequency of occurrence of SSRs is drastically reduced. Thus, the SSRs were obtained in the form of repeat motifs ranging from mono- to hexa-nucleotides. Out of the 7552 unigenes searched for SSRs, 395 SSRs (Table 2) were generated from 377 unigenes. These 395 SSRs can be termed as EST-SSRs and 377 unigenes possessing SSRs can be termed as SSR-ESTs according to the convention. The 377 SSR-ESTs amounted to approximately 5% (inclusive of the mononucleotide repeat motif) of total unigenes and 2.78% of total downloaded EST data set. The various studies show a representation ranging from 2.65% - 16.82% [25] to 26.84% [42] in dicot species and 7% - 10% [43] in cereals or monocots. The workers [17] [21] [25] who have carried out similar studies are of the view that the variation in the percentage may be due to variation in sample size, search criteria, size of database, and the tools used for EST-SSR development. The percentage of SSR-ESTs in the present study could be owing to more stringent preset parameters for EST mining compared to other similar studies [21] [42] that reported a higher percentage of SSR-ESTs.

The 395 SSRs were present in 377 SSR-ESTs as 17 (4%) SSR-ESTs contained more than one SSR e.g. FM889616.1 with 3 SSRs, having motifs (GA)₃₂, (AG)₁₄, (AG)₁₄ (data not shown). The SSRs in mononucleotide class were found to be the most abundant with a frequency of 1/17.83 kb followed by dinucleotide 1/36.00 kb, trinucleotide 1/84.82 kb, tetranucleotide 1/424.11 kb, pentanucleotide 1/347.00 kb and hexanucleotide 1/381.70 kb.

The overall analysis of the distribution of the microsatellites into various classes of the repeat types (mono-, di-, tri-, tetra- penta- and hexa-nucleotides) showed that the number of the microsatellites decreased with increasing motif size (Figure 2, Table 3). It was observed that mononucleotide repeats were the most abundant (repre- senting 54% of the total microsatellites), followed by dinucleotide (27%) and trinucleotide (11%). The least frequent were tetra-, penta- and hexa-nucleotides (2% - 3%). The abundance of mononucleotides is in accordance with several previous reports [23] [25] [44] and also that these contributed to nearly half of all the SSRs, is similar to those in certain species of dicots analysed previously [25] . The dinucleotides were the second most abundant class as reported across most of the dicots investigated by Kumpatla and Mukhopadhyay [25] , suggesting an over-representation of UTRs (un-translated regions) compared with ORFs (Open Reading Frames).

The non-dominance of trinucleotides compared to other classes, by virtue of which the decreasing trend of various classes with increasing motif size, is in contrast to several earlier studies but in concurrence to that reported for several dicots [25] . These observations about the abundance and therefore, the dominance of one SSR motif category over other categories, holds significance in the chances of fixation of mutations against selection pressure [45] . The trinucleotides have more chances of getting fixed against mutation pressure due to selection against frameshift events [45] The prevalence of di- over tri-nucleotide in this study could be attributed to 1. increased stringency of preset parameters in this study compared to previous studies [21] [22] , so as not to compromise on polymorphism level and thus their utility as markers. The results were also computed with relaxed preset parameter of repeat length which gave a higher percentage of total SSRs especially trinucleotides (data not shown). But, the results reported here are those obtained with more stringent parameter of minimum repeat length 2. a bias in representation of 5’ and 3’UTRs in the EST dataset used for mining. A lowered representation of tetranucleotides, as also observed in this study, is also suggestive of under representation of 3’UTRs [25] .

In terms of SSR coverage of available Unigenes data (~3.8 MB), it was observed that a total of 11.7 kb (0.31%) region was covered by SSR motifs. Out of this, mono-represented 6.5 kb (0.17%) region, di—3.3 kb (0.08%), tri—1.1 kb (0.03%), tetra—0.18 kb (0.004%), penta—0.26 kb (0.006%) and hexa-nucleotides 0.24 kb (0.006%).

The various classes of repeat motifs, when analyzed further, showed that some motifs in each category were more abundant than others (Table 4), e.g. among the dinucleotide repeats, the AG/CT motif was the most common (33%) followed by the motifs GA/TC (31%) and, the least common was AC/GT (0.94%). The abundance of AG/CT/GA/TC motifs are in concurrence with previous studies [25] [43] [44] where ESTs were used for mining SSRs, in contrast to abundance of AT motif when genomic data was used for mining SSRs [44] [46] . Thus, abundance of the motifs is attributed to systematic bias resulting from the use of ESTs (coding sequences) instead of genomic sequences (non-coding) as a source for SSR mining [43] . The CG motif was found to be totally absent, which is in concurrence to earlier studies, where it has been observed to be either the least [43] or absent [44] . Among the trinucleotide repeats, the most common motif is AGA/TCT subclass amounting to

26.6% and rest of them ranging from 2% - 13% of the total microsatellites in this class. The CCG/CGG motif is reported to be the rarest motif in dicots [23] [25] and was observed to be absent in this study. In the tetranucleotide repeats, most of the motifs were AT rich. The most common motif was AAGA/TCTT (22%) and the rest of them were each ~11% of the total microsatellites in this class. In the pentanucleotide class of motifs the most common one was AAGAA/TTCTT and TATTT/AAATA (18% each) and others were each 9%. The hexanucleotide class TTTCTC/GAGAAA (30%) formed the most abundant subclass and the rest of them were 10% each. In general, the motifs were observed to be AT rich and less of GC rich motifs, similar to that observed for dicots [25] .

The analysis of repeat units under each motif class revealed a varying range of repeat units in each of the classes of repeat motifs. It was observed that, in dinucleotide motif, repeat units ranged from 10 - 45; in trinucleotide motif, from 7 - 13; in tetranucleotide, from 5 - 6 units; in pentanucleotide, from 4 - 6; and hexanucleotide motif was represented by a single class of 6 repeat units only. Further analysis of the number of repeat units in every class of the SSRs, especially tri-, tetra- penta- and hexa-nucleotides, showed that the number of the microsatellites decreased with increasing repeat unit length with little variation, e.g. for trinucleotide motif, SSRs with 7 repeats were represented by 42.2% while 2.2% by 13 repeat units. Amongst the pentanucleotide SSRs, the category with 4 repeat units shared as much as 45.5% of the total class in comparison to 9% for repeat unit of seven (Figure 3). Therefore, it can be said that as the class of the SSR motif size increases, like tetra-, penta- and hexa-nucleotide, higher number, rather 100% of microsatellites were found in the category of <10 repeat units (Table 2) which is similar to that observed by Varshney and co-workers [43] . These results clearly indicate the effect of increased stringency of parameters which were maintained during this study to retain the polymorphism level and utility of the SSRs as markers because the probability of polymorphism increases with increasing length of SSRs [47] -[49] and, a higher number of repeat followed by shorter stretches would be beneficial for marker development [48] . The polymorphism reported in Jatropha in earlier studies was very low, therefore, the parameters for mining the SSRs were kept more stringent in this study, which lead to lower frequency of SSRs but with a longer repeat length; as in the case of trinucleotide repeats, keeping the minimum repeat length of 7 resulted in it not being the most abundant class, as reported in other similar studies.

For the use of SSRs as markers, it is necessary to design the primers. The SSRs commonly used for marker development are those belonging to di-, tri- and tetra-nucleotides [25] . The mononucleotides are useful for population genetic analyses of chloroplast genomes [50] and can also be useful in filling gaps in linkage maps created by di-, tri-, and tetra-nucleotide repeats [25] but, at the same time they cause difficulties in accurate sizing of polymorphisms [18] . Therefore, to design the primers for potential SSR markers, the mononucleotide repeats were not included. Thus, out of 395 EST-SSRs generated from 377 SSR-ESTs, the primers were designed for only 181 SSRs.

For each of the SSRs, a pair of reverse and forward primer was designed from the flanking regions of their respective SSR-ESTs by the software. 181 SSRs generated from 172 SSR-ESTs were used for primer designing and yielded 79 SSR mediated primer pairs (data not shown). These 79 primer pairs were designed from 76 SSR-ESTs as some of these contained more than one SSR e.g. JES 56 and 57 (Supplementary Table A). Thus, 76 SSR-ESTs having credit in primer designing have been termed as ESTs-PD and were further annotated. The primers could not be designed for some of the EST-SSRs from their respective SSR-ESTs. As reported by Varshney and coworkers [42] , this could be due to any or all of the following reasons, (a) SSR-ESTs are too short, (b) EST-SSRs are too close to the cloning site of the SSR-ESTs, or (c) the flanking sequences are not unique, as was also observed for some of the SSRs in this study.

The GC level of the genome of J. curcas is typical of core dicots, therefore, it should be easy to annotate by sequence comparison with Arabidopsis [41] , hence, ESTs-PD were searched for their gene annotations using BLASTX at TAIR. The Gene Ontology (GO) Annotations and functional categorization of ESTs-PD obtained using locus identifiers are given in Supplementary Table A.

The data showed that most of the ESTs-PD are expressing functional proteins and still there are some for which the protein is not yet predicted. On the basis of the functions related to the predicted protein, the ESTs-PD were classified into three major classes viz. Cellular Component, Biological Process and Molecular Function (Figure 4). In the limits of the available data in the public database for the ESTs of J. curcas, it was found that one of the ESTs-PD (Contig1345) containing SSR (JES35) expresses gene of oil biosynthesis pathway (AT1G48750).