Introduction

The gibberellin (GA) phytohormones function as important endogenous growth regulators in plants. The bioactive GAs (e.g., GA1, GA3, and GA4) modulate diverse physiological and developmental processes including stem elongation, leaf expansion, seed germination, pollen maturation, and the induction of flowers and fruits (Olszewski et al. 2002; Sun and Gubler 2004). In some species, bioactive GAs also play a major role in stimulating early stages of xylem differentiation during wood formation and ultimately in determining wood quality (Israelsson et al. 2005). The accumulating evidence that GAs stimulate these processes indicates a close link between GA metabolism and GA response pathways to maintain GA homeostasis in plants. The precise mechanisms of GA perception and signal transduction have remained longstanding questions that are central to the regulation of gene expression and plant phenotype. Extensive studies showed that the components of GA signaling pathways include the transcriptional regulators DELLA proteins and the soluble GA receptor encoded by GIBBERELLIN INSENSITIVE DWARF 1 (GID1). DELLAs act as repressors, often negatively regulating GA signaling, and are degraded in response to GA treatments during growth (Peng et al. 1997; Ikeda et al. 2001; Chandler et al. 2002). GID1 is essential for GA perception and was first isolated in rice (Oryza sativa, OsGID1) based on the GA insensitive dwarf 1 (gid1) mutant (Ueguchi-Tanaka et al. 2005). Subsequently, three homologous GID1s (AtGID1a, AtGID1b, and AtGID1c) were cloned in Arabidopsis thaliana. Each AtGID1 could complement the rice gid1 mutant, restoring the wild phenotype, showing that GID1s can respond to GA and function as GA receptors (Nakajima et al. 2006; Griffiths et al. 2006).

Based on the identification of GID1s in rice and Arabidopsis, research in the past decade has identified GID1 genes in fern (Hirano et al. 2007), cotton (Aleman et al. 2008), barley (Chandler et al. 2008), aspen (Mauriat and Moritz 2009), and wheat (Li et al. 2013). Previous studies have focused mainly on the identification and characterization of GID1 loci in various species (Ueguchi-Tanaka et al. 2005; Nakajima et al. 2006; Chandler et al. 2008) including the functional expression (Iuchi et al. 2007), the crystal structure of GID1 bound with GA or a ternary complex of GA-GID1-DELLA (Murase et al. 2008; Shimada et al. 2008), and the molecular mechanisms of GA perception and signaling (Ueguchi-Tanaka et al. 2007a, b; Hirano et al. 2008; Sun 2011). However, many questions remain, such as the allelic variation of GID1 associated with complex quantitative traits, especially the specific genetic variants in GID1 related to growth and wood quality traits.

Linkage disequilibrium (LD)-based association studies (also known as LD mapping) can effectively bridge our understanding of complex quantitative traits and the underlying genetic variation in specific candidate genes in a natural or breeding population (Zhang and Zhang 2005; Eckert et al. 2009). For example, Wegrzyn et al. (2010) found 27 significant single marker associations across 40 candidate genes with three composite traits (lignin content, syringyl-to-guaiacyl ratio, and C6 sugars) in Populus nigra. In addition, a recent study identified eight significant single nucleotide polymorphisms (SNPs) from two S-adenosyl-l-homocysteine hydrolase (SAHH) genes associated with growth or wood traits in a natural population of P. tomentosa (Du et al. 2014a). These studies suggested that association mapping can effectively identify alleles associated with complex quantitative traits in tree breeding programs. Therefore, based on the SNP and haplotype association tests, the identification of new GID1 genes and dissection of their allelic variation will provide essential information for improving growth and wood formation.

In this study, we identified PtoGID1a and PtoGID1b, which encode GID1 proteins, and characterized these genes using LD-based association studies in P. tomentosa. Quantitative real-time PCR (qRT-PCR) analysis showed that PtoGID1a and PtoGID1b were expressed in all tissues and organs, in accordance with GA being involved in all developmental stages of the plants, and may be inducible by GA treatment in trees. Association genetics identified six common SNPs and six haplotypes in PtoGID1a and PtoGID1b that were associated with growth and wood properties in 426 unrelated individuals from a natural population of P. tomentosa. To extend these observations, we investigated the genetic variance in PtoGID1a and PtoGID1b, including additive and dominant effects. This comprehensive study of PtoGID1a and PtoGID1b polymorphisms can provide important resources and information for marker-assisted selection and breeding in forest trees.

Materials and methods

Population materials

The association population consisted of 426 unrelated individuals growing in Guan Xian County, Shandong Province (36° 23′ N, 115° 47′ E), China, where root segments of 1047 native individuals collected throughout the entire natural distribution range of P. tomentosa (covering an area of 1 million km2) were used to establish a clonal arboretum using a randomized complete block design with three replications in 1982 (Zhang et al. 2010). The total distribution of P. tomentosa can be divided into three climatic regions, southern, northwestern, and northeastern, on the basis of principal component analysis and ISODATA fuzzy clustering of 16 meteorological factors (Zhu 1992; Huang 1992). In the present study, these 426 uncorrelated individuals, representing almost the entire geographic distribution of P. tomentosa (148 from the southern region, 126 from the northwestern region, and 152 from the northeastern region), were used for association analysis. In addition, a panel of 45 individuals was randomly selected from the association population (15 individuals from each region) and directly sequenced to test for SNPs within PtoGID1a and PtoGID1b.

Phenotypic data

Ten phenotypic traits were measured for at least three ramets per genotype in all individuals of the association population. Of which, three growth characteristics including tree height, diameter at breast height (DBH), and stem volume were directly measured during field surveys in 2011 using standard procedures described in Zhang et al. (2006). For seven wood property measurement, each wood sample contains bark and pith (15 cm in length and 10 mm in diameter) were cored from the 426 trees at breast height (1.3 m above ground level) using instrument borers. Then, these wood cores were pulverized to wood meal, in which the contents of holocellulose, α-cellulose, hemicelluloses, and lignin were consulted to test using near-infrared reflectance spectroscopy (NIRS) according to Schimleck et al. (2004) based on training sets and calibration models derived from wet chemistry analysis technique described by Porth et al. (2013) and Tian et al. (2012). Fiber length and width were determined by the Color CCTV Camera (Panasonic SDII) after macerated in Franklin solution at 70 °C for 48 h and washed with deionized water until the wood materials had been neutralized. An X-ray powder diffractometer (Philips, Eindhoven, the Netherlands) was used to measure the microfibril angle, and the X-ray diffraction profile was integrated at Chi between −180° and +180°. The distribution of values for each trait was approximately consistent with a normal distribution (Tian et al. 2012; Du et al. 2013, 2014b).

Isolation of PtoGID1a and PtoGID1b cDNAs

Total RNA was extracted from the leaves of 1-year-old P. tomentosa using the Plant Qiagen RNeasy kit and then reverse transcribed into cDNA with the Superscript First-Strand Synthesis system according to the manufacturer’s instructions (Life Technologies, Carlsbad, CA, USA). The constructed cDNA library was composed of 5.0 × 106 pfu with insert size of 1.0–4.0 kb. Subsequently, random end sequencing of 5000 cDNA clones and comparison with all available A. thaliana GID1 sequences and P. trichocarpa GID1 sequences in NCBI (http://www.ncbi.nlm.nih.gov) identified two clones that were highly similar to AtGID1s and PtGID1s. The BLAST program was used to analyze these expressed sequence tag (EST) sequences. Finally, two cDNA clones encoding GID1 members were detected and isolated from P. tomentosa by amplification using gene-specific primers and named PtoGID1a and PtoGID1b, respectively, according to their order of discovery.

DNA extraction and full-length genomic DNA identification

Using the DNeasy Plant Mini kit (Qiagen China, Shanghai), total genomic DNA was extracted from young leaves of each P. tomentosa individual following the manufacturer’s protocol. For sequencing the genomic DNA of PtoGID1s, a pair of specific primer (Table S1) was designed for amplification of PtoGID1a and PtoGID1b, respectively, based on their cDNA sequences and the available P. trichocarpa GID1 sequences (Potri.013G028700 and Potri.002G213100, respectively) in NCBI, which contain the promoter regions and 3′ UTR-downstream regions. PCR amplification was performed according to the procedure described by Tian et al. (2014a), but the annealing temperature was replaced by 68 °C. Next, PCR products were resolved by agarose gel electrophoresis, excised, purified, ligated, transformed, and isolated. Ultimately, the BigDye Terminator Cycle Sequencing Kit, version 3.1 (Applied Biosystems, Beijing, China), run on a Li-Cor 4300 genetic analyzer (Li-Cor Biosciences, Lincoln, NE), was used to directly sequence both strands of the plasmid DNA (three positive clones for each individual). For PtoGID1a and PtoGID1b, 5326 and 4912 bp of genomic DNA sequence were obtained, both including 2000 bp of the promoter region and 500 bp of 3′ UTR-downstream sequence. The 90 genomic clone sequences for PtoGID1a and PtoGID1b were deposited in GenBank under the accession numbers KU161176–KU161220 and KU161221–KU161265, respectively.

Gene structure and phylogenetic analysis

The gene structure diagrams of PtoGID1a and PtoGID1b were produced with Gene Structure Display Server (GSDS 2.0) (http://gsds.cbi.pku.edu.cn) after independently submitting their coding and genomic sequences. To investigate the phylogenetic relationship of PtoGID1a and PtoGID1b to the GID1 proteins from other species, 37 amino acid sequences of GID1 from dicotyledons, monocotyledons, gymnospermous plants, as well as lower plants such as pteridophyte and bryophyte plants (Table S2) were obtained from NCBI (http://www.ncbi.nlm.nih.gov). Multiple sequence alignment, phylogenetic, and molecular evolutionary analyses were conducted using MEGA version 6.06, and the neighbor-joining method was used to build an unrooted phylogenetic tree (Tamura et al. 2007). Statistical confidence of tree nodes was based on 1000 bootstrap replicates.

Tissue-specific expression analysis

Plant materials and GA3 treatment

The vegetatively propagated 1-year-old P. tomentosa “LM50” were grown under 16-h light and 8-h dark in a greenhouse at Beijing Forestry University, Beijing, China (40° 0′ N, 116° 20′ E). For GA treatment, 100 μM GA3 (Sigma-Aldrich, St. Louis, MO, USA) was sprayed on seeding leaves until drops of liquid dripped down. The control plants were treated with water in the same manner. Mature leaves from the same position of control plants (at 0 h) and GA-treated plants were collected at 0, 3, 12, and 24 h after treatment, and three biological replicates were sampled per time points. Nine tissues (root, apex, bark, phloem, cambium, young leaf, mature leaf, developing xylem, and mature xylem) and four tissues (mature leaf, cambium, developing, and mature xylem) of three ramets were collected at 0 and 12 h, respectively, were used to examine the expression of PtoGID1a and PtoGID1b in different tissues and organs of P. tomentosa. All the harvested tissues were immediately frozen in liquid nitrogen after collection and stored at −80 °C until use.

RNA extraction

Total RNA was extracted using the Plant Qiagen RNeasy kit (Qiagen China, Shanghai) according to the manufacturer’s instructions. During the RNA purification, additional on-column DNase digestions were performed three times using the RNase-Free DNase Set (Qiagen). Then, RNA was quantified based on absorption at 260 nm and reverse transcribed into cDNA using the Superscript First-Strand Synthesis system and the supplied polythymine primers (Invitrogen). All cDNA samples were used for tissue-specific expression analysis of PtoGID1a and PtoGID1b with qRT-PCR.

Quantitative real-time PCR

With the PtoGID1a- and PtoGID1b-specific and internal control (Actin) primer pairs (Table S3) designed with Primer Express 3.0 software (Applied Biosystems), qRT-PCR was performed on a 7500 Fast Real-Time PCR System (ABI). Two microliters of cDNA for all tissues was amplified in a reaction containing 10 μl of Quanti-Tect SYBR Green PCR reagent (Qiagen), 0.4 μl of ROX Reference Dye II solution (Invitrogen), 0.4 μl of 10 nM forward and reverse primers, and 6.8 μl of dH2O. The PCR program included an initial denaturation at 95 °C for 30 s, 40 cycles of 5 s at 95 °C, 30 s at 60 °C, and a final melt curve of 72–95 °C. All reactions were carried out in triplicate for technical and biological repetitions with three plants. The specificity of the amplified fragments was checked through the melting curve, and the generated data was analyzed using the Opticon Monitor Analysis software 3.1 tools (Bio-Rad, USA) and standardized to the level of the Actin gene.

SNP identification and genotyping

The full-length genomic DNAs of PtoGID1a and PtoGID1b, both including 2000 bp of the promoter region and 500 bp of the 3′ UTR-downstream region, were resequenced and analyzed in 45 unrelated individuals from the association population to identify SNPs without consideration of insertions/deletions (InDels). To identify putative SNP variants, plasmid DNA was sequenced on both strands with conserved primers using the BigDye Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems, Beijing, China) and a Li-Cor 4300 genetic analyzer (Li-Cor Biosciences, Lincoln, NE). Then, BioEdit and DNA sequence polymorphism (DnaSP) software version 5.10 (Librado and Rozas 2009) were used for multiple-sequence alignment and polymorphism analysis. Ultimately, 159 common SNPs (minor allele frequencies ≥0.10, 78 SNPs for PtoGID1a, and 81 SNPs for PtoGID1b) were genotyped by single nucleotide primer extension using a Beckman Coulter sequencing system for all DNA samples.

Nucleotide diversity and linkage disequilibrium

DnaSP version 5.10 (Librado and Rozas 2009) was used to calculate summary statistics for nucleotide diversity and divergence. Nucleotide diversity was estimated using both the average number of pairwise differences per site between sequences, π (Nei 1987), and the average number of polymorphisms segregating per site, θ w (Watterson 1975). In addition, the diversity statistics were also calculated separately for noncoding, synonymous, and nonsynonymous sites in PtoGID1a and PtoGID1b. To estimate whether natural selection (purifying selection or positive selection) was involved in the evolution of the two genes during speciation, we did d N/d S analysis on the DNA sequences of PtoGID1a and PtoGID1b.

To assess the pattern of LD in the sequenced candidate gene regions, the decay of LD within a specific physical distance (base pairs) between SNP sites within PtoGID1a and PtoGID1b was estimated by nonlinear regression analysis (Hill and Weir 1988) using DnaSP version 5.10 with sequence data. LD was measured as the squared correlation of allele frequencies (r 2), which is affected by both recombination and differences in allele frequencies between sites (Hill and Robertson 1968). The significance of the r value for each pair of common SNPs was calculated using the same software DnaSP version 5.10 (Librado and Rozas 2009).

Association analysis

Single SNP-based association

All single SNP-based association tests between 159 SNP markers of PtoGID1a and PtoGID1b and 10 traits in the natural population were conducted via three statistical models: (1) a general linear model (GLM) only take population structure into account (Q model), (2) a mixed linear model (MLM1) only inferred relative kinship as a random covariate (K model), and (3) a unified mixed linear model (MLM2) uses both Q and K values (Q + K model) (Yu et al. 2006). These analyses with above three models were implemented in TASSEL version 2.1 (Yu et al. 2006; Bradbury et al. 2007). As described previously, the values of membership probability (Q) and pairwise kinship (K) were used to evaluate the effects of population structure and relatedness among individuals for marker-trait associations. The Q matrix was established based on the significant subpopulations (K = 11) by 20 genomic SSR markers (Du et al. 2012), as assessed according to the statistical model described by Evanno et al. (2005), using STRUCTURE version 2.3 (Pritchard et al. 2000). The relative kinship matrix (K) was obtained based on the same 20 SSR loci using the method proposed by Ritland (1996), which is built into the program SPAGeDi version 1.3 (Hardy and Vekemans 2002). The K matrix was calculated as described by Yu et al. (2006), and all negative values between individuals were set to 0. To correct for multiple testing, the false discovery rate (FDR) method was additionally applied to identify significant SNPs using QVALUE in R (Storey and Tibshirani 2003; Storey et al. 2004).

Haplotype-based association

All high-LD blocks of haplotypes (r 2 ≥ 0.75, P ≤ 0.001) were determined for each gene using the software Haploview version 4.1 (Barrett et al. 2012), and the haplotype frequencies were estimated based on the genotypic data for the 426 individuals. On the basis of the information from those significant SNPs in the haplotype blocks, haplotype-based association tests with growth and wood quality traits were performed using haplotype trend regression software (HTR) (Zaykin et al. 2002) with a three-marker sliding window. The significances of the haplotype associations were conducted based on 104 permutation tests. Singleton alleles and haplotypes with a frequency <5 % were discarded when constructing the haplotypes.

Modes of gene action

The modes of gene action were quantified using the ratio of dominant (d) to additive (a) effects (|d/a|) estimated from least square means for each genotypic class. When the ratio of dominance to additive effects was in the range 0.5 < |d/a| ≤ 1.25, gene action was defined as partial or complete dominance. Additive effects were defined as value of |d/a| ≤ 0.5; under- or over-dominance was defined as values of |d/a| > 1.25. Details of the algorithm and formulas for calculating gene action were previously described (Eckert et al. 2009; Wegrzyn et al. 2010).

Results

Isolation and sequence analysis of PtoGID1a and PtoGID1b from P. tomentosa

Two full-length cDNAs of PtoGID1a and PtoGID1b were isolated from a cDNA library prepared from the leaves of P. tomentosa by reverse transcription PCR. The cDNAs of PtoGID1a and PtoGID1b are 2153 and 1852 bp in length, respectively, including 333 and 261 bp of the 5′ untranslated region (UTR), 785 and 556 bp of 3′ UTR, and have open reading frames of equal length 1035 bp. Alignment of the full-length cDNA sequence with the genomic sequence showed that PtoGID1a and PtoGID1b each have one intron (673 and 560 bp, respectively), located between the two exons (Fig. 1a). The deduced protein sequences of PtoGID1a and PtoGID1b revealed these genes each encode 344 amino acid polypeptides with estimated molecular weights of 38.54 and 39.12 kDa and theoretical pI values of 6.15 and 6.22, respectively.

Fig. 1
figure 1

Overview of PtoGID1a and PtoGID1b. a Gene structure of PtoGID1a and PtoGID1b in P. tomentosa. Exons and upstream/downstream regions, which included untranslated regions (UTRs), are shown as black and gray boxes, respectively, and the gray lines between boxes indicate introns. Positions of the translational start (ATG) and termination codons (TAG/TAA) of the predicted coding sequences are shown. b Neighbor-joining phylogenetic tree of PtoGID1a, PtoGID1b, and GID1 proteins from other plants. Detailed information on 37 protein sequences is presented in Additional file Table S2

Full size image

We compared the amino acid sequences of PtoGID1a and PtoGID1b with those of known AtGID1s (AtGID1a, AtGID1b, AtGID1ac) (Fig. Fig. S1). The similarity between the two PtoGID1 sequences was 79.65 %, and both had 70.35–84.59 and 73.26–79.65 % similarities with the AtGID1s from A. thaliana, respectively. PtoGID1a and PtoGID1b share sequence similarity with the consensus sequence of the hormone-sensitive lipases (HSL) family, including the conserved motifs HGG and GXSXG. Moreover, the PtoGID1a and PtoGID1b amino acid sequences also retain the first two conserved residues, serine (S) and aspartic acid (D), which form the HSL catalytic triad. The third histidine (H) residue was replaced with isoleucine (I) at the corresponding positions in both PtoGID1a and PtoGID1b compared with OsGID1 and the AtGID1s (Fig. Fig. S1). In addition, there were 13 functional domains (TWVLIS, LDR, FFHGGSF, HS, IYD, YRR, DGW, GDSSGGNI, GNI, MF, LDGKYF, WYW, and GFY) important for GA binding activity or GID1-DELLA protein interaction (Ueguchi-Tanaka et al. 2007a, b) from the N-terminal to C-terminal, and these domains also occurred in PtoGID1a and PtoGID1b (Fig. Fig. S1), except for the LDR domain replaced by LER in PtoGID1b. It is therefore proposed that the PtoGID1a and PtoGID1b genes encode fully functional GA receptors.

Phylogenetic analysis of GID1 proteins in plant

To investigate the evolutionary relationship of PtoGID1a and PtoGID1b with the GID1 proteins from other plant species, an unrooted neighbor-joining (NJ) tree was constructed using 37 complete amino acid sequences (Table S2). Alignment of all predicted proteins showed that these GID1 proteins can be divided into five distinct groups, and the sequences in each group follow the same classification as the species (Fig. 1b). This suggests that GID1 differentiation events occurred before the divergences of gymnosperm–angiosperm, moss–fern, and monocot–dicot plants. Further subdivisions of the eudicot groups revealed that GID1a and GID1c could be classified into the same groups, except for PhGID1c and MdGID1c, while GID1b was located in an independent subgroup, indicating that these individual GID1s may have distinct biological functions (Nakajima et al. 2006; Griffiths et al. 2006). The phylogenetic tree indicated that PtoGID1a is an ortholog of AtGID1a/c and that PtoGID1b should be classified with AtGID1b and other GID1b genes.

Ubiquitous expression of PtoGID1a and PtoGID1b in P. tomentosa

To examine the expression of PtoGID1a and PtoGID1b in different tissues and organs of P. tomentosa, we compared the steady-state mRNA levels of the two genes by qRT-PCR analysis with gene-specific primers and Actin as the internal control (Fig. 2a, b). Expression of the two genes was detected in all the tissues and organs at varied levels before or after GA3 treatments, consistent with that GA is active throughout the plants and during many developmental processes. The transcripts of both PtoGID1a and PtoGID1b were most abundant in the leaf and showed similar overall profiles. The differences were that PtoGID1a transcript levels were highest in mature leaf tissue (4.24 ± 0.1695, in arbitrary units, relative to the Actin control), followed by young leaf tissue (2.35 ± 0.2919), but PtoGID1b mRNA transcript levels were highest in young leaf tissue (2.74 ± 0.4208), followed by mature leaf tissue (0.35 ± 0.0506) at 0 h. By contrast, in the secondary tissues of bark, phloem, and xylem, the expression levels of PtoGID1a and PtoGID1b showed the lowest mRNA abundance. However, in comparison to untreated tissues, transcript abundances of these two genes were increased in mature xylem (0.184 ± 0.0035 vs. 0.027 ± 0.001 and 0.061 ± 0.0041 vs. 0.002 ± 0.0006) and developing xylem (0.017 ± 0.0009 vs. 0.003 ± 0.0003 and 0.058 ± 0.0030 vs. 0.0039 ± 0.0004) at the time points of 12 h after GA3 treatment (Fig. 2b). Additionally, the transcript levels of PtoGID1a and PtoGID1b were also increased in cambium (0.886 ± 0.0391 vs. 0.115 ± 0.0529 and 0.0495 ± 0.0022 vs. 0.027 ± 0.0065) relative to the original, although moderate expressions of the two genes were found in root, apex, and cambium tissue before GA3 treatment.

Fig. 2
figure 2

Relative transcript levels of PtoGID1a and PtoGID1b. a Relative expression levels of PtoGID1a and PtoGID1b in various tissues and organs from P. tomentosa at 0 h. b Relative expression levels of PtoGID1a and PtoGID1b in various tissues at the time points of 12 h after GA3 treatment. c Relative expression levels of PtoGID1a and PtoGID1b before and after GA treatment at different time points in leaves tissue. All expression levels were normalized to the mean value of Actin. The error bars represent ±SD

Full size image

We further explored whether the expression of PtoGID1a and PtoGID1b responds to GA by measure their transcript levels at different time points after GA3 treatment (Fig. 2c). PtoGID1a expression first increased and then decreased following GA3 treatment, peaking at 12 h (25.8 ± 2.114 times higher than the Actin control), suggesting that GA3 upregulates PtoGID1a. Conversely, PtoGID1b expression decreased following GA3 treatment up to 12 h but then increased gradually to a peak at 24 h (1.1 ± 0.222 times higher than the Actin control) after treatment.

Nucleotide diversity and linkage disequilibrium of PtoGID1a and PtoGID1b

To identify nucleotide polymorphisms, we sequenced 5326 and 4912 bp genomic regions of PtoGID1a and PtoGID1b, respectively, from 45 unrelated individuals of P. tomentosa. All of the 90 sequences from PtoGID1a and PtoGID1b contained the 5′ UTR, introns, 3′ UTR, 1035-bp exons, 2000-bp promoter regions, and 500-bp 3′ UTR-downstream regions. Across the samples, a total of 251 SNPs (123 from PtoGID1a and 128 from PtoGID1b) were detected in these two genes at a frequency of approximately one SNP every 43 and 38 bp, respectively. Forty-three and eight of these SNPs occurred in coding regions of PtoGID1a and PtoGID1b, respectively, of which 33 SNPs were categorized as silent mutations and 18 were missense sites (Table 1). Altogether, 78 and 81 (63.41 and 63.28 %) were considered common SNPs (minor allelic frequency, MAF ≥ 0.10) in PtoGID1a and PtoGID1b, respectively. In general, both of these genes showed high nucleotide diversity (Table 1). The average levels of nucleotide diversity (π Τ) for the different regions varied significantly ranging from 0.00336 (3′ UTR) to 0.01766 (exon 2) in PtoGID1a and from 0.00174 (intron) to 0.01365 (promoter) in PtoGID1b (Table 1). Similarly, estimates of the population mutation parameter (θ w) varied between 0.00239 (intron) and 0.01012 (exon 2) in PtoGID1a and 0.00123 (intron) and 0.00924 (exon 2) in PtoGID1b. In coding regions of PtoGID1a and PtoGID1b, the values of nonsynonymous nucleotide diversity (π nonsyn = 0.00847 and 0.00021, respectively) were markedly lower than the synonymous nucleotide diversity (π syn = 0.04375 and 0.01429, respectively) (Table 1). The π nonsyn/π syn ratios for the exons were 0.1936 and 0.0147, respectively, reflecting the strong action of purifying selection involved in nonsynonymous sites of exon regions during speciation.

Table 1 Nucleotide polymorphisms in PtoGID1a and PtoGID1b locus
Full size table

The pattern of the squared allelic correlation coefficient (r 2) with base pairs within PtoGID1a and PtoGID1b (including the promoter and the 3′ UTR-downstream regions) was used to assess the LD level. The nonlinear regression model showed that LD decayed quite rapidly with r 2 dropping to 0.1 within ~350 and ~1000 bp in PtoGID1a and PtoGID1b, respectively (Fig. 3), indicating that LD did not extend over the entire gene region. We then genotyped 78 and 81 common SNPs from 426 individuals, and 13 and four high-LD distinct haplotype blocks across the sequenced regions differed in PtoGID1a and PtoGID1b, respectively. There was a higher level of LD within each block (r 2 > 0.8), whereas LD levels were low between one block to the other haplotype block (r 2 < 0.3) (Fig. S2).

Fig. 3
figure 3

Decay of linkage disequilibrium within PtoGID1a and PtoGID1b. Pairwise correlations between SNPs were plotted against the physical distance between the SNPs in base pairs. The curve describes the nonlinear regression of the least squares fit r 2 (Er2) onto the physical distance in base pairs

Full size image

Single SNP-trait and haplotype-based associations

To explore the effects of SNPs in PtoGID1a and PtoGID1b, we conducted 1590 association tests (159 SNPs × 10 traits) using three statistical models (Q, K, and Q + K) for comparing (Table S4). For PtoGID1a, the numbers of significant markers (P < 0.05) were 33 using K model and 25 using Q model but fell to 14 with Q + K model. In PtoGID1b, we found 34 SNPs were associated with tree traits at P < 0.05 by the K model, 29 by the Q model, and only 13 by the Q + K model. After correction for multiple testing with a FDR test (Q < 0.05), the numbers of significant associations of SNPs with growth and wood property traits were reduced to four and two in PtoGID1a and PtoGID1b, respectively (Table 2). These six significant associations represented six unique sites from the promoter, 5′ UTR, and exon of PtoGID1a and PtoGID1b. These SNPs were significantly associated with five phenotypic traits: tree height, stem volume, fiber width, α-cellulose contents, and microfiber angle (Table 2). Of these markers, SNP52 and SNP55 from exon 2 of PtoGID1a were synonymous substitutions, and all associated markers from PtoGID1b occurred in noncoding regions. Altogether, the percentage of phenotypic variance (R 2) explained by each SNP ranged from 0.99 to 10.28 %, in accordance with polygenic quantitative models of wood traits.

Table 2 Significant SNP markers associated with growth and wood property traits and list of marker effects for significant marker-trait pairs
Full size table

To quantify the modes of gene action, we calculated the ratio of dominant (d) to additive (a) effects (|d/a|). Two of the six marker-trait pairs showed additive effects (|d/a| ≤ 0.5), one showed a partially to fully dominant effect (0.5 < |d/a| ≤ 1.25), and the remaining associations were split between over- or under-dominance (|d/a| > 1.25) (Table 2). For example, the two genotypes of PtoGID1a-SNP16 and PtoGID1b-SNP3 showed difference in tree height (15.70 ± 1.162 m for AA, 15.00 ± 0.296 m for GA, and 11.00 ± 0.335 m for GG) and stem volume (0.65 ± 0.049 m3 for AA, 0.53 ± 0.100 m3 for AC, and 0.35 ± 0.085 m3 for CC), consistent with additive effects. The heterozygotes of PtoGID1a-SNP52 and PtoGID1b-SNP32 had higher stem volume (0.68 ± 0.070 m3 for AG vs. 0.59 ± 0.056 m3 for AA, 0.49 ± 0.012 m3 for GG) and wider fibers (22.97 ± 0.474 mm for GT vs. 22.43 ± 0.269 mm for GG, 22.65 ± 0.733 mm for TT), respectively, than either homozygote class. The average values of microfiber angle across genotypic classes for PtoGID1a-SNP12 were suggestive of partial dominance (18.21 ± 1.512° for AA, 17.65 ± 0.464° for AG, and 17.33 ± 1.553° for GG), with the A allele being partially dominant (Fig. 4).

Fig. 4
figure 4

Genotypic effects of the significant single nucleotide polymorphisms (SNPs) in PtoGID1a and PtoGID1b on different phenotypic traits in an association population. The x axes represent the genotypic classes. The left y axes represent the phenotypic data of the association population. Upper and lower edges of each box represent the first and third quartiles of the distribution, and the central lines indicate the average value. Whiskers in the box plots represent the maxima and minima

Full size image

Moreover, we performed haplotype-based association tests to identify significant haplotypes with the same 10 growth and wood quality traits. Three obvious LD regions including eight common haplotypes were observed only in PtoGID1a with significant effect on these traits (P < 0.05, frequency > 1 %) (Table 3). Among these, two haplotypes from SNP1–4 were simultaneously associated with DBH and stem volume and three haplotypes from SNP61–64, which spans the 3′ UTR region, were associated with α-cellulose contents. SNP49–51 in exon 2 showed close association with fiber width. The proportion of phenotypic variation explained by each haplotype ranged from 3.10 to 5.39 % (Table 3). To additionally dissect the allelic variation in the haplotype identified in PtoGID1a, we also tested associations using haplotype trend regression (HTR) in PtoGID1b (Table S5). One haplotype (G-T-G) located in the promoter region extending from SNP35 to SNP37 was concurrently associated with fiber width, fiber length, and microfiber angle at a frequency of <1 %.

Table 3 PtoGID1a haplotypes significantly associated with growth and wood property traits
Full size table

Discussion

Identification and characterization of PtoGID1a and PtoGID1b from P. tomentosa

The GID1-mediated GA perception system involves transcriptional regulation of GA-responsive genes in plant growth and development. In the present study, two GID1s encoded by PtoGID1a and PtoGID1b were determined to not only contain two conserved motifs (HGG and GXSXG) within the HSL family and retain the first two residues (S and D) of the catalytic triad at the corresponding position, but they also preserve all 13 important functional domains, in accordance with the expected conserved features of GID1 identified in other plants. Ala-scanning mutagenesis of OsGID1 showed that these conserved motifs and functional domains are crucial for the interaction of GID1 with GAs or DELLA proteins (Ueguchi-Tanaka et al. 2007a, b).

The known eudicot GID1 genes fall into two distinct groups: GID1a/c and GID1b (Hirano et al. 2007; Voegele et al. 2011), which is in accordance with our findings that the two PtoGID1 genes were classified into the “GID1a/c” and “GID1b” subgroups of the phylogenetic tree (Fig. 1b). This supports the view that the different GID1 receptors seem to have distinct functions during plant development. The distinction between PtoGID1a and PtoGID1b is also supported by the tissue-specific expression patterns. In our study, although PtoGID1a and PtoGID1b were expressed in most of the tissues, their expression levels were different. For example, PtoGID1a was predominantly expressed in mature and young leaves, as was PtoGID1b. However, the PtoGID1a expression level in mature leaf was 12 times higher than that of PtoGID1b (Fig. 2a). Moreover, the hormonal regulation of expression of PtoGID1a and PtoGID1b at the same time points after GA3 treatment also differed (Fig. 2c). The idea that the individual GID1 genes have different expression profiles and display partial redundancy has been postulated identified in A. thaliana, in which AtGID1a and AtGID1c transcripts were present at high and low levels in most of the analyzed tissues, while the level of the AtGID1b transcript was moderate (Griffiths et al. 2006). In our study, we found that the deduced amino acid sequences of the PtoGID1 genes were similar. The transcriptional expression of PtoGID1a and PtoGID1b differed temporally and spatially and exhibited differential expression in various tissues and in response to GA3 treatment. Therefore, by a combination of structural analysis, functional expression, and phylogenetic analysis of PtoGID1a, PtoGID1b, and GID1s in other plants, we infer that these two PtoGID1 members as GA receptors are orthologs of OsGID1 and AtGID1s.

Analysis of statistical models for association

The method of association mapping has the potential to pinpoint genetic polymorphism underlying important traits (e.g., growth and wood properties) in highly heterozygous undomesticated forest trees. However, false discoveries (also known as false positive rates) are a major concern and can be partially attributed to spurious associations caused by population structure and unequal relatedness among individuals (Yu et al. 2006). Therefore, an appropriate statistical model is necessary to account for this point. Initially, the GLM model with Q matrix has attracted considerable attention as a powerful and effective approach for reducing the risk of false positive rates arising from population stratification. However, Q values alone are not adequate because the Q matrix only provides a rough dissection of population differentiation (Yang et al. 2010). Similarly, the simple model with kinship matrix alone would not be sufficient to correct genotypes among relatives in a population and thus can lead to spurious association in the similar way (Bourgain et al. 2003; Yu et al. 2006). To reduce the false positive rates on marker-trait associations while maintaining statistical power, a unified mixed-model approach with correction for population stratification and relatedness have been suggested. In the Q + K model, population structure is fit as a fixed effect, whereas kinship among individuals is incorporated as the variance–covariance of the random effect for the individuals (Pritchard et al. 2000; Zhao et al. 2007).

In the present study, we compared the classical GLM (Q) and MLM (K, Q + K) models on testing associations of two genes for growth and wood property traits carried out in a natural population of P. tomentosa. The numbers of detected significant markers from PtoGID1a and PtoGID1b and the explained phenotypic variance were varied with different models (Table S4). Of which, the Q + K model was even more conservative and detected fewer significant markers (14 vs. 33 or 25, and 13 vs. 34 or 29 in PtoGID1a and PtoGID1b, respectively) for growth and wood property traits than the simple Q- and K-based models. Also, almost all of those marker-trait association tests which are detected by the Q + K model appeared in other two models (Table S4). This result indicated that the Q + K model is not prone to an enhanced false positive or false negative due to the model could sufficiently control both population structure and kinship among individuals in our data set. Therefore, we ultimately decided to use the MLM (Q + K) to test our phenotype–genotype associations. Similarly, a combined population structure and kinship approach has been successful in interpreting the association results. For example, Ehrenreich et al. (2009) used the MLM to conduct candidate gene association mapping of flowering time in Arabidopsis. Additionally, Zhao et al. (2007) also demonstrated that the MLM model was the most appropriate method for associations between flowering-related phenotypes and genotypes at candidate loci in Arabidopsis. Taken together, choosing of the proper statistical models can help to reduce the number of false positive rates and elevated the accuracy of associations.

Detection of significant single SNP and haplotypes associated with growth and wood properties

Populus is a widely used model organism for studies of tree molecular biology and biotechnology, having a rapid growth rate, modest genome size, and extensive polymorphic genetic markers. Poplar not only provides a complete genomics “toolkit” for the comparison of a long-lived perennial to short-lived model plants (e.g., Arabidopsis), but it also offers abundant genetic and adaptive variation for exploring the genetic basis of shoot growth, wood formation, and other important research topics (Bradshaw et al. 2000; Taylor 2002; Brunner et al. 2004). Given this, the identification of genes and allelic variants associated with growth and wood quality is important for practical application in tree breeding programs. The detection of specific marker-trait relationships using association genetics is a widely employed approach, and these markers can be used for marker-assisted selection (MAS) breeding. In recent years, diverse groups of candidate genes affecting chemical wood properties have been identified in related Populus species using SNP-based associations (Tian et al. 2012; Du et al. 2014a).

In our study, we combined single marker association and haplotype-based tests jointly to dissect functional allelic polymorphisms within PtoGID1a and PtoGID1b for complex growth and wood property traits underlying LD-based association in P. tomentosa. Because of the rapid decay of LD in P. tomentosa (Fig. 3), once a marker-trait association has been discovered and validated, it is likely that the marker is located in close proximity to the causal polymorphisms or even is the functional variant itself (Neale and Kremer 2011). This rapid decay of LD could prevent researchers from focusing on linked but not relevant markers and missing potentially significant markers or functional variants. Therefore, LD-based methods will be a powerful approach for delineating complex functional genes. In all, we identified six single marker associations and six haplotypes within PtoGID1a and PtoGID1b that were significantly associated with wood and growth traits (Tables 2 and 3), which demonstrated that PtoGID1a and PtoGID1b may be involved in stem growth and wood formation. All of the significant SNPs identified by single marker association were synonymous mutations or in noncoding regions. Of these, two synonymous substitutions (SNP52 and SNP55) in exon 2 of PtoGID1a were strongly associated with stem volume and α-cellulose content, and the mode of gene action appeared to be over- or under-dominance, except for PtoGID1a-SNP55. Although the synonymous mutations did not cause an amino acid change, these silent SNPs can still affect the levels of gene expression and alter codon usage, thus affecting gene function via other mechanisms (Kimchi-Sarfati et al. 2007). A similar phenomenon has been discovered in many previous studies. For instance, a synonymous exonic SNP (SNP10) of PtGA20ox was associated with fiber width (Tian et al. 2012) and the same type of SNP (SNP45) in PtoCesA4 of P. tomentosa was significantly associated with holocellulose content (Du et al. 2013).

Many functional analyses of SNPs have emphasized coding regions and splice sites in candidate genes related to phenotypic traits. However, the vast majority (over 90 %) of trait-associated variants have been found in noncoding regulatory regions where they perturb binding sites of transcription factors, local chromatin structure, or co-factor recruitment, ultimately influencing important biological regulation (Thumma et al. 2009; Beaulieu et al. 2011; Maurano et al. 2012). This is especially true for promoters, which are usually assumed to be the key regulatory regions that control gene transcription (Roa-Rodrigues 2003). In the present study, we detected three significant markers located in the promoter region and one located in the 5′ UTR. For example, PtoGID1a-SNP12 and PtoGID1a-SNP16 in the promoter and 5′ UTR region, respectively, were associated with microfiber angle and tree height, while PtoGID1b-SNP3 and PtoGID1b-SNP32 both resided in the promoter region and were associated with stem volume and fiber width (Table 2). These findings were consistent with previous studies that SNP loci in noncoding regions are significantly associated with wood traits. Southerton et al. (2010) detected a strong association between a SNP (HB16 located in the 5′ UTR) and microfiber angle in Eucalyptus nitens. Similarly, PtoSAHHB-1065, in the first intron of SAHHB associated with α-cellulose content, was identified in both discovery and validation populations of P. tomentosa (Du et al. 2014a). Marker-trait associations have also been reported for markers in promoter regions. Du et al. (2013) detected a significant marker 273 bp (T/A) upstream of the transcriptional start site of CesA4 in P. tomentosa associated with α-cellulose content. These association results support the prevalent view that more SNPs occur in the promoter region or noncoding region than in the coding region of individual genes (Su et al. 2011).

Generally, the power of a single marker association test is often limited because LD information contained in the flanking markers is ignored, while haplotypes are intuitively regarded as more informative blocks since they preserve the joint LD structure among multiple adjacent markers (Akey et al. 2001). Recently, many association studies have stressed the necessity for a haplotype approach instead of merely relying on individual SNPs. Given the putative power of haplotype to assess the causal SNP loci in complex traits, we also conducted multi-locus haplotype LD tests in this mapping study. Through the comparison of single marker and haplotype-based associations (Tables 2 and 3), we found that the individual SNP loci within these haplotype blocks (SNP1–4, SNP49–51, and SNP61–64) have no significant correlation with growth and wood qualities. The marker trait-associated results of both methods demonstrated that the haplotype-based analysis may not be more powerful than single marker-based analysis in this low LD tree species, which is in accordance with a previous study on UDP-glucuronate decarboxylase (a key enzyme in synthesis of UDP-xylose) during cell wall biosynthesis in P. tomentosa by Du et al. (2013). However, contradictory studies suggest that haplotype-based tests can detect previously unknown quantitative traits and are well suited to unraveling the genetic basis of complex traits. Wegrzyn et al. (2010) examined individual genes and SNPs for complex traits controlling lignocellulosic biosynthesis in P. trichocarpa and found that the use of haplotypes can significantly increase the ability and robustness of mapping traits of interest. Therefore, we suggest that the best way is to combine single marker-based and haplotype-based methods, which will provide more insights than any one method into the functional allelic variance underlying quantitative traits in an association population.

Additionally, most of the associations identified in our study explained a small proportion of the phenotypic variance (Tables 3 and S4) and the individual SNP loci within haplotype blocks have no significant correlation with traits, which suggested that each SNP of the haplotype blocks may has a minor effect, and the collective contribution of several minor sites can significantly affect phenotypic variance. These phenomena attributed to individual SNP alleles are consistent with a polygenic quantitative model of wood traits (Beaulieu et al. 2011), in accordance with earlier studies in a same nature population of P. tomentosa. For example, Du et al. (2013) and Tian et al. (2014a) identified cellulose synthase gene PtoCesA4 and PtoCesA7 allelic variants, respectively, which associated with growth and wood quality traits under the polygenic model, explaining 1.6–5.6 and 3.62–10.59 % of the phenotypic variability. Similarly, several significant associations were found and the phenotypic variances explained by a single SNP were also having a small proportion in UDP-glucose dehydrogenase (PtUGDH) (Tian et al. 2014b) and PtGA20Ox genes (Tian et al. 2012) in P. tomentosa. These significant associations were found with both tree growth (e.g., tree height, stem volume) and wood properties trait (e.g., fiber length, fiber width, microfibril angle, holocellulose, lignin, and α-cellulose contents), which we are usually interested in selecting for breeding. Therefore, we used a candidate gene-based association mapping approach to identify genes and allele variants controlling phenotypic variability in a nature population, which could provide more useful information and resources for forest tree breeding program.

Conclusions

In conclusion, we cloned two full-length cDNAs from P. tomentosa; these are similar to OsGID1 and AtGID1s. The structural and phylogenetic analysis proved that these two clones, PtoGID1a and PtoGID1b, encode the likely GA receptors. Tissue-specific expression profiles of PtoGID1a and PtoGID1b revealed similar patterns, with either high or low expression levels in all tissues and organs, consistent with the participation of GA in growth processes throughout the life cycle of the plant. Subsequently, single marker and haplotype-based association analysis identified six single SNP loci and six haplotype blocks that were significantly associated with growth and wood properties. These findings implied that PtoGID1a and PtoGID1b may play an important role in tree growth and wood formation by mediating GA signaling. With the rapid development of next-generation sequencing technology and more sophisticated statistical software, there will be more opportunities to identify additional GID1 genes in P. tomentosa because the genome of Populus is relatively larger than those of other plants (e.g., A. thaliana, which has three GID1-type GA receptors). The present study identified functional allelic variants of PtoGID1a and PtoGID1b with growth and wood quality traits, given new insights into genes affecting wood quality and quantity. From an applied perspective, the SNPs revealed in our study have potential applications in marker-assisted breeding.