Introduction

Masson pine (Pinus massoniana Lamb., MP) is not merely a major timber and wood pulp species but also an ecologically important one in montane forest ecosystems. This species grows extensively throughout across central and southern China, spreading some 12° in latitude and 21° in longitude (21°41′-33°56′N, 102°10′-123°14′E) (Zhou 2000; Qin et al. 2003). It accounts for one half of the forest stock volume there (Xu and Wang 2001). Of high fiber quality, MP is widely used in construction, furniture, and pulp industries. It is indigenous, grows fast, even under conditions too harsh for other species, and is commonly planted as a pioneer species for reforestation (Zhou 2000). Even back in the early 1980s, a national improvement program was initiated with the goal of meeting logging and pulp industry demands (Ding et al. 2005; Zhou et al. 1997). However, in terms of population genetics, there have been no marker-assisted selection (MAS) studies for MP because of the lack of effective molecular markers. Furthermore, the collection and conservation of genetic germplasm resources are important issues for MP long-term breeding. The informative DNA markers are powerful high-resolution tools for evaluation of genetic variation at the molecular level.

Microsatellites, or simple sequence repeats (SSRs) have distinguishing features such as reproducibility, abundant polymorphism, codominant inheritance, and even distribution along chromosomes. These make them a predominant tool for population genetics, molecular ecology, and MAS studies (Li et al. 2002; Selkoe and Toonen 2006; Oliveira et al. 2006). Conventional methods for developing SSR markers in nonmodel organisms involve the construction of small–insert genomic libraries and subsequent enrichment of clones containing tandem repeat motifs (Powell et al. 1996; Kalia et al. 2011). Those methods are laborious and difficult to use. To date, only 20 MP genome SSR markers have been published (Guan et al. 2011; Hung et al. 2012), in part, probably because pine species have extremely large genomes, and thus less effort has been put into microsatellite identification. An alternative method of SSR mining is to transfer existing markers from related species, though some publications have reported available transferability with low SSR polymorphism in conifer species (Bérubé et al. 2003; Echt et al. 1999; Gonzalez-Martinez et al. 2004; Fu and Shi 2005; Mariette et al. 2001). Ai et al. (2006) tested 116 SSRs from other Pinus species (P. taeda and P. strobus), and only 6.90% of them were polymorphic in MP. EST-based SSRs have relatively high cross-transferability owning to the conserved nature of protein-coding regions (Lesser et al. 2012; Chagné et al. 2004; Decroocq et al. 2003; Liewlaksaneeyanawin et al. 2004). Another advantage is the genic-SSRs probably affecting gene functions and regulation (Kalia et al. 2011). Nevertheless, a main potential issue is that EST–SSRs have generally lower polymorphism levels than those of genomic SSRs. Many studies have reported a higher polymorphism and resolution in genome-derived SSRs than in EST-derived ones (Song et al. 2012; Pandey et al. 2012; Yang et al. 2005; Chabane et al. 2005; Wen et al. 2010; Ueno et al. 2012; Chagné et al. 2004). Therefore, the development of species-specific genomic SSRs should be useful for fingerprinting and parentage analysis in closely related lines. Besides, previous reports have demonstrated a higher GSS–SSR density than EST-derived ones in loblolly pine (P. taeda, LP) (Echt et al. 2011). If SSR markers have similar properties in MP, mining SSRs from GSSs rather than ESTs may be more practicable, especially if working on a limited budget. Fortunately, the advent, extensive utilization, and declining cost of high-throughput sequencing technologies have allowed for fast advances in the sequencing of nonmodel organisms, in particular, for organisms with large genomes such as conifers (Ritland 2012; Magbanua et al. 2011; Fernández-Pozo et al. 2011; Kovach et al. 2010; Keeling et al. 2011; Parchman et al. 2010).

In this study, we investigated and analyzed the MP microsatellites DNA obtained from 454 GS-FLX shotgun sequencing, and compared them with the GSS–SSRs from three nonconifer species, to explore the feasibility of GSS–SSR development. We then designed a set of PCR primer pairs based on the GSSs obtained to evaluate MP–SSR marker amplification and polymorphism.

Materials and methods

DNA extraction, sequencing, and assembly

Fresh MP needles were collected from a superior tree (No. 125) in a seed orchard located in Fujian province, one of the main MP production areas in China. The Cetyl–trimethyl ammonium bromide (CTAB) method (Porebski et al. 1997) with minor modifications was used for total genomic DNA extraction. Five micrograms of genomic DNA were cut into 1000–1400 bp fragments with a Sonic Dismembrator 550 (Fisher Scientific, US) and purified with Agencourt AMPure beads (Agencourt Bioscience, US). The purified DNA was then prepared as a single strand library using a GS DNA Library Preparation Kit (Roche Applied Science, US), and bound on beads using a GS emPCR Kit (Roche Applied Science, US), and eventually sequenced on a Roche 454 GS-FLX platform. A quarter run was performed, and a total of 218,871 reads with an average length of 580 bp were identified. After adaptor removal and assemblage by Newbler version 2.3 (Roche Applied Science, USA) with default parameters (minimum overlap length = 40, minimum overlap identity = 90%, contig length threshold = 100 bp), we got 21,122 contigs and 106,428 (48.6% of total reads) singletons.

SSR mining and analysis

Only nonredundant singletons (Table 1) were utilized for SSR mining and analysis owing to the fact that the extremely large conifer species genome (C. 20–40 Gb) (Kovach et al. 2010) may result in unreliable assembly. A total 5,393 GSS sequences and 101 LP BAC clone sequences were downloaded from http://www.ncbi.nlm.nih.gov/dbGSS and http://www.ncbi.nlm.nih.gov/nuccore, respectively to compare SSR frequency and characteristics with those of the MP genome shotgun sequences. To further understand the GSS–SSR characteristics, the GSSs of the three nonconifer species, namely Populus trichocarpa (9.5 Mb), Vitis vinifera (136.7 Mb), and Oryza sativa (250.6 Mb) were also downloaded from National Center for Biotechnology Information (NCBI) to investigate their SSR densities and proportions. The LP BACs were also used as a reference and compared to LP and MP GSSs because of their sequence length.

Table 1 Number and total size (bp) of sequences examined, average seq. length (bp), average G + C content, total number of SSRs, No. of SSR-containing seq., frequency of SSRs, and No. of c-SSRs in five species
Full size table

The MISA program (http://pgrc.ipk-gatersleben.de/misa/misa.html) was used for SSR searches. The definition of unit size/minimum number of repeats were set as 2/9, 3/6, 4/5, 5/5, and 6/4, and 0 bp was set as the interruption of two adjacent SSRs. All recovered microsatellites were classified by type (di-, tri-, tetra-, etc.) and motif (AT/TA, AG/CT, CG/GC, etc.) (Bérubé et al. 2007). Chi-square (χ 2) tests were used to compare observed SSR frequencies on various unit types between the GSSs of species according to Echt et al. (2011).

SSR marker evaluation

Two wild populations and clonal seed orchard population were employed for marker evaluation. The wild populations were Shanghang (SH) and Wuping (WP) located at 25°18′N, 116°52′E, in an elevation of 970–1280 m; and 25°09′N, 115°55′E, in an elevation of 352–472 m, the orchard was Wuyi clonal seed orchard (WCS) located at 24°58′N, 117°27′E, in an elevation 434 m. DNA was isolated from the needles of 143 samples via the method mentioned above (Porebski et al. 1997) and quantified by NanoDrop 2000 (Thermo Scientific, US).

The 10 μl PCR reaction mixture containing 50 ng DNA, 10 mM Tris–HCl, 2.5 mM MgCl2, 0.25 mM dNTP, 0.3 μM primer, and 0.5 U Taq polymerase (Takara, Dalian, China) was run on a GeneAmp® PCR system 9700 (Applied Biosystems, US). The PCR program was as follows: 5 min denaturation at 94 °C, followed by 30 cycles of 30 s at 94 °C, 30 s at 50/51 °C (see Table 1), and 30 s at 72 °C, then a final extension at 72 °C for 5 min. PCR products were run on an 8 % denaturing polyacrylamide gel and the bands in each lane were carefully recorded.

Data analysis was performed on Popgene version 1.32 (Yeh et al. 1999), including the number of alleles per locus (Na), observed heterozygosity (Ho) and expected heterozygosity (He), probability of deviation from Hardy–Weinberg equilibrium (HWE), and calculation of linkage disequilibrium (LD) between all loci. The polymorphic information content (PIC) were calculated by PowerMarker version 3.2.5 (Liu and Muse 2005), and Micro-Checker version 2.2.3 (van Oosterhout et al. 2004) was used to estimate the frequencies of null alleles for each locus.

Results

Microsatellite identification and analysis

In total, 106,428 unique MP GSSs with an average length of 565 bp were analyzed after discarding the redundant sequences. Of these, 1,699 SSRs conforming to the definitions (i.e., unit/minimum number of repeats 2/9, 3/6, 4/5, 5/5, 6/4) were recovered from 1,530 sequences (1.4 % of the total sequences). Compound SSRs accounted for 3.2 % and the average SSR density was 28 SSR/Mb (Table 1). Among all of the MP GSS unit types, dimers were dominant (59.74 %), in which the AT/TA motif occurred at the highest frequency (61.67 %) (Fig. 1), followed by AG/CT (19.90 %), and AC/GT (11.01 %). AAT/ATT was the most abundant trimer (25.66 %) motif (58.49 %), and AAG/CTT was second at 14.91 %. In the tetramers (7.65 %), AAAT/ATTT (52.31 %) was the most common, and AAAAT/ATTTTT (18.82 %) was the most common hexamer (5.00 %). However, of the pentamers (1.94 %), AGCCT/AGGCT (18.18 %) and AAAAC/GTTTT (18.18 %) were the most common motifs; AAAAT/ATTTT (12.12 %) and ACTCT/AGAGT (12.12 %) were the second most.

Fig. 1
figure 1

Motif proportions in each P. massoniana unit type. Each motif in corresponding unit type showed in the form of percentage. The ‘others’ denotes types other than specified

Full size image

Of the 5,393 LP GSSs (covering 2.5 Mb), a total of 122 SSRs were recovered from 110 sequences (2.0 %) with a density of 48 SSRs/Mb, which also contained 3.2 % compound SSRs. Of the 101 BACs (covering 11.8 Mb), 99 possessed microsatellites, and there were 758 SSRs (containing 5.8 % compound forms) in total with an average density of 64 SSRs/Mb (Table 1). Notably, MP GSSs exhibited a significantly lower SSR frequency than those of both LP GSSs (χ 2 = 64.79, p < 0.0001) and BACs (χ 2 = 828.80, p < 0.0001). MP also had the lowest average GSS–SSR density when compared to O. sativa (χ 2 = 256.89, p < 0.0001), V. vinifera (χ 2 = 1,461.21, p < 0.0001), and P. trichocarpa (χ 2 = 918.53, p < 0.0001). Likewise, in the LP genome, dimeric SSRs were the most frequent GSSs (69.67 %) and BACs (69.79 %), followed in order by tri-, tetra-, hexa-, and pentanucleotides (Fig. 2). The most common motifs of all five unit types in BACs were AT/TA (77.50 %), AAT/ATT (61.82 %), AAAT/ATTT (56.16 %), AAAAT/ATTTT (38.46 %), and AAAAAT/ATTTTT (27.27 %). Similar results were found in GSSs, except for the pentanucleotides in which only one class of motif (AAGAT/ATCTT) was identified. Interestingly, chi-square tests revealed that the dimer proportions between LP GSSs and BACs (χ 2 = 0.0007, p = 0.98) were closer than those of MP and LP GSSs (χ 2 = 4.69, p = 0.03). Besides, variation tendency of each unit type against repeats (especially dimers) differed slightly between MP and LP (more of a fluctuating decline with the increasing number of repeats in LP BACs and GSSs than in MP GSSs) (Fig. 2), although significant Pearson's correlations were observed not only between LP BACs and GSSs (r = 0.67, p = 0.0005) but also between LP and MP GSSs (r = 0.63. p = 0.0014).

Fig. 2
figure 2

Percentages of various unit types with different numbers of repeats in masson pine (MP), loblolly pine (LP) genome survey sequences, and loblolly pine BAC (LPB) clone sequences. The di-, tri-, tetra-, penta-, and hexa- denote the corresponding nucleotides unit types of microsatellites

Full size image

In comparing the MP GSS–SSR properties with those of the nonconifer species, similar trends in the proportions of unit types were found in P. trichocarpa and V. vinifera. Dimers were the dominant type and accounted for 59.44 % in P. trichocarpa and 55.75 % in V. vinifera, followed in order by tri-, tetra-, hexa-, and pentamers. In O. sativa, triplet SSRs (53.61 %) replaced dinucleotides (32.86 %) as the most common type. In terms of the motifs of the five unit types, similar to MP and LP, AT/TA, AAT/ATT, and AAAT/ATTT were the predominant motifs in P. trichocarpa and V. vinifera. But in O. sativa, apart from AT/TA taking first place with a percentage of 44.74 %, AAT/ATT and AAAT/ATTT were not the most abundant motifs.

Microsatellite evaluation

In total, 240 primer pairs (containing 104 di-, 91 tri-, 20 tetra-, 4 penta-, 15 hexamers, and 6 compound SSRs) were designed using Primer3 version 4.0.0 (Koressaar and Remm 2007) with a preferred amplicon size of 100–400 bp, GC content of 30–70 %, T m value of 50–60 °C, and primer lengths of 18–25 bp. Prescreening all primer pairs with one individual randomly chosen from a wild population (SH), we successfully amplified 121 SSRs (50 %, containing 47 di-, 51 tri-, 9 tetra-, 1 penta-, 11 hexamers, and 2 compound forms) with the expected sizes, and subsequently checked in six samples. Twenty (16.5 %) SSRs (Table 2) with obvious and repeatable polymorphic bands were evaluated in the SH, WP, and WCS populations. Of these, dimers were the most abundant (9 of 20), followed by six hexamers, four trimers, and one compound trimer (Table 2). No tetrameric or pentameric loci displayed polymorphism in the candidates. An online BLASTn search showed that all of these polymorphic loci were novel and no significantly similar nucleotide sequences were found on NCBI.

Table 2 Forward (F) and reverse (R) primer sequence, repeat motif, expected amplicon size, optimized annealing temperature (Ta), and GenBank accession No. of 20 Pinus massoniana microsatellite loci
Full size table

The polymorphic levels of each primer pair in three populations are listed in Table 3. The mean Na ranged from 3.3 to 3.5, Ho/He ranged from 0.472/0.533 to 0.487/0.540, and PIC ranged from 0.527 to 0.534. 8 of the 20 loci deviated from HWE expectations in two wild populations, and three of them (P. Ma22, P. Ma96 and P. Ma116) showed evidence for null alleles in the form of observed heterozygosity deficiency (Table 3). Though a significant LD was detected in a few pairs of loci (e.g., P. Ma22 vs 95 (p = 0.038) in SH; P. Ma24 vs 27 (p = 0.017) in WP) in one or two populations, no locus pair synchronously revealed significant LD in all of the three populations, implying that all 20 loci are not closely linked to each other. The Ho/He/PIC values in the clonal seed orchard (WCS) were similar to the two wild populations, indicating that genetic diversity has been maintained in the seed orchard.

Table 3 Number of alleles (Na), observed (Ho) and expected (He) heterozygosity, polymorphic information content (PIC), and estimated null allele frequency for three populations of masson pine
Full size table

Discussion

SSRs have emerged as an important marker system in many areas of molecular genetics because they exhibit a high degree of allelic diversity. The identification and development of SSR markers represent significant challenges for nonmodel or extremely large genome size organisms. We examined the characteristics of microsatellites contained in 2–6 bp unit types of MP based on next-generation genome sequencing technology. Our work revealed an extremely low GSS–SSR frequency in MP when compared with LP and the three nonconifer species. We inferred that the MP genome has a naturally low SSR density, though the actual mechanism causing this phenomenon remains unclear. The low GSS–SSR density observed in MP compared with the other species is likely due to evolutionary divergence among the species rather than due to sequencing technique or sampling error biases. Lower GSS–SSR density in MP and LP compared with V. vinifera and P. trichocarpa partially agrees with previous reports that the occurrence of SSRs in conifers is low (Bérubé et al. 2007; Echt et al. 2011; Ueno et al. 2012), that means that using high-throughput sequencing technology in large-scale SSR mining of MP and other conifer species is a preferable or even necessary approach. It should be pointed out that, in LP, the higher SSR density in BACs than that in GSSs may imply that the frequency of GSS–SSR underrepresented the true density of genome wide SSR, because the longer and relatively complete BAC sequences may better represent the true characteristics of the LP genome.

The AT motif occurred at a higher proportion than any other dinucleotide motifs in all the analyzed species. Furthermore, the frequency of the A/T-rich motif in tri-, tetra-, penta-, and hexa- unit also took the highest or relatively higher rank in all the species except for O. sativa (see Table 1). This possibly demonstrates that the AT/TA motif does not only universally appear in the pine genome (Victoria et al. 2011), but also in other tree species (Ueno et al. 2012; Pootakham et al. 2012; Tuskan et al. 2004). The GSS–SSR characteristics among these species were partly compatible with the phylogenetic history of these species. The genome of P. trichocarpa and V. vinifera (xylophyta) may undergo similar environmental selection pressures to MP and LP, while O. sativa, a herbaceous crop, experiences different selection forces (Li et al. 2000).

Among the five categories of hexa- (73 %) and trinucleotides (56 %) seem to have higher amplification capabilities, while of penta- (25 %), tetra- (45 %), and dinucleotides (45 %) were comparatively low. To test this statistically, we classified all five unit types into two groups and then performed a χ 2-test because fewer tetra- and penta-nucleotides were observed. Group I was trimer plus hexamer, and group II was the sum of the other three types. The test result (χ 2 = 4.52, p = 0.034) supported our suggestion. Considering the fact that the amplification success rate of tri- and hexamer SSRs being higher than others implies that these types of SSRs may locate in conserved or coding regions, we carried out an online BLASTn and no significant similar sequence was found. Besides, we divided these 20 novel polymorphic SSRs into two groups, group I was dimer and group II was the sum of tri- and hexamer, and then performed t tests between two groups on Na, Ho, He and PIC values from three examined populations. The result showed no significant difference between two SSR groups on Na (p = 0.181), Ho (p = 0.981), He (p = 0.138), and PIC (p = 0.139). This interesting result probably indicates that we should pay more attention to hexa- and trinucleotides SSRs in subsequent MP genomic microsatellites development, though it could not be rejected that the tri- and hexamer SSRs may reside in coding regions.

Microsatellite markers generally have a high degree of cross-transferability and are an important marker source for genetically related species. Previous studies reported 54 % to 94.2 % EST–SSR transfer rates in a few Pinus species (Lesser et al. 2012; Chagné et al. 2004). However, the high marker transferability usually accompanies low polymorphism. Although a large percentage of EST–SSR markers were able to amplify in related species, some of them usually have very rare or even monomorphic alleles (Liewlaksaneeyanawin et al. 2004; Lesser et al. 2012; Echt et al. 2011). Instead, the species-specific genomic SSR markers generally have higher polymorphic levels than EST-based ones (Chagné et al. 2004; Liewlaksaneeyanawin et al. 2004), which is useful for fingerprinting and pedigree reconstruction of closely related individuals. According to the criteria of Botstein et al. (1980), more than half of the 20 novel polymorphic SSRs demonstrated highly informative scores (PIC > 0.50), and the others were reasonably informative scores (0.50 > PIC > 0.25) except for the SSR of P. Ma77 (PIC < 0.25). This means that these markers can be considered as an important marker resource in MP population genetic studies. Of these, 4–6 loci exhibited null alleles (estimated null allele frequencies range from 0.114 to 0.350) in examined populations and especially for P. Ma96 and P. Ma116, null alleles were detected in all three populations, implying that caution should be exercised when using these loci in population genetics.

Genetic improvement of MP has so far mainly involved traditional selection and crossing based on phenotypes (Ding et al. 2005; Zhou et al. 1997). The SSR markers we developed and others (Guan et al. 2011; Hung et al. 2012) can be used in fingerprinting of superior clones, paternity analysis and pedigree reconstruction of breeding populations. Thus, we expect to enhance the accuracy and progess of the MP breeding program significantly. Furthermore, our subsequent development of large-scale GSS- and EST-based SSR markers development should greatly facilitate the QTL mapping, MAS, and comparative genomics studies.