Email updates

Keep up to date with the latest news and content from BMC Plant Biology and BioMed Central.

Open Access Research article

Integrating genome annotation and QTL position to identify candidate genes for productivity, architecture and water-use efficiency in Populus spp

Romain Monclus2, Jean-Charles Leplé1, Catherine Bastien1, Pierre-François Bert16, Marc Villar1, Nicolas Marron45, Franck Brignolas23 and Véronique Jorge1*

Author affiliations

1 INRA, UR0588 Amélioration Génétique et Physiologie Forestières (AGPF), F-45075, Orléans, France

2 UFR-Faculté des Sciences, UPRES EA 1207 Laboratoire de Biologie des Ligneux et des Grandes Cultures (LBLGC), Université d'Orléans, F-45067, Orléans, France

3 INRA, USC1328 Arbres et Réponses aux Contraintes Hydriques et Environnementales (ARCHE), F-45067, Orléans, France

4 INRA, UMR1137 Écologie et Écophysiologie Forestières (EEF), F-54280, Champenoux, France

5 Université de Lorraine, UMR 1137, Ecologie et Ecophysiologie Forestières (EEF), Faculté des Sciences, F-54500, Vandœuvre-lès-Nancy, France

6 Present address: INRA, UMR1287 Ecophysiologie et Génomique Fonctionnelle de la Vigne, F-33882, Villenave d'Ornon, France

For all author emails, please log on.

Citation and License

BMC Plant Biology 2012, 12:173  doi:10.1186/1471-2229-12-173


The electronic version of this article is the complete one and can be found online at: http://www.biomedcentral.com/1471-2229/12/173


Received:26 March 2012
Accepted:22 September 2012
Published:26 September 2012

© 2012 Monclus et al.; licensee BioMed Central Ltd.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Background

Hybrid poplars species are candidates for biomass production but breeding efforts are needed to combine productivity and water use efficiency in improved cultivars. The understanding of the genetic architecture of growth in poplar by a Quantitative Trait Loci (QTL) approach can help us to elucidate the molecular basis of such integrative traits but identifying candidate genes underlying these QTLs remains difficult. Nevertheless, the increase of genomic information together with the accessibility to a reference genome sequence (Populus trichocarpa Nisqually-1) allow to bridge QTL information on genetic maps and physical location of candidate genes on the genome. The objective of the study is to identify QTLs controlling productivity, architecture and leaf traits in a P. deltoides x P. trichocarpa F1 progeny and to identify candidate genes underlying QTLs based on the anchoring of genetic maps on the genome and the gene ontology information linked to genome annotation. The strategy to explore genome annotation was to use Gene Ontology enrichment tools to test if some functional categories are statistically over-represented in QTL regions.

Results

Four leaf traits and 7 growth traits were measured on 330 F1 P. deltoides x P. trichocarpa progeny. A total of 77 QTLs controlling 11 traits were identified explaining from 1.8 to 17.2% of the variation of traits. For 58 QTLs, confidence intervals could be projected on the genome. An extended functional annotation was built based on data retrieved from the plant genome database Phytozome and from an inference of function using homology between Populus and the model plant Arabidopsis. Genes located within QTL confidence intervals were retrieved and enrichments in gene ontology (GO) terms were determined using different methods. Significant enrichments were found for all traits. Particularly relevant biological processes GO terms were identified for QTLs controlling number of sylleptic branches: intervals were enriched in GO terms of biological process like ‘ripening’ and ‘adventitious roots development’.

Conclusion

Beyond the simple identification of QTLs, this study is the first to use a global approach of GO terms enrichment analysis to fully explore gene function under QTLs confidence intervals in plants. This global approach may lead to identification of new candidate genes for traits of interest.

Background

Dissection of genetic architecture of complex trait such as growth and yield in plants has been achieved by Quantitative Trait Loci (QTL) detection approaches. Dissecting loci to find the causative genes can be considered as the Holy Grail for geneticists. The ultimate road to find the genes, positional cloning, has been achieved in main crop plants [1] but it is slow and labour intensive especially because large segregating populations have to be developed. The advent of plant whole genome sequences has opened the possibility of anchoring genetic maps and positioning QTL on a physical map. Nevertheless, QTL intervals correspond to several hundreds of genes [1]. Despite of some successful positional cloning stories in plants, there is room for complementary approaches, like association mapping (reviewed in [2]) or integrative –omics strategies [3], proposed to narrow intervals and length of candidate gene lists to be studied further [1].

The identification of Quantitative Trait Loci (QTL) has already been reported for poplar species including several yield and growth components such as stem dimension (height, circumference) and architecture [4-16]. These studies together identified more than 600 QTLs explaining up to 73% of the trait variation, with confidence intervals ranging from 2.1 to 261.2 cM. These global summary statistics confirm the common idea that growth traits in Populus are controlled by many loci. However, ranges of variation explained and size of confidence intervals highly depend on the mapping population used and the environment(s) in which traits are measured.

Today, the anchoring of genetic maps and QTLs on Populus trichocarpa genome sequence [17] allows identifying large genome regions containing several hundred of genes. In order to reduce this number to a reasonable number of candidate genes, one strategy consists in selecting candidate genes based on functional knowledge (for ex. known biosynthetic pathways; transcriptomic data; annotation inferred from homologous genes in other species) and check if the candidates co-locate with QTLs [11,18,19]. Another strategy, applied in Populus spp., consists in combining QTL position, structural information and transcriptomic experiments to refine a gene list for functional characterization [11,20]. The present study focused on traits that are of notable importance for poplars biomass production systems, where juvenile growth, architecture and water use efficiency (WUE) are the main criteria for breeding programs. Among published QTL studies carried out on these traits, few have anchored QTLs on the poplar genome to identify and analyse the underlying large candidate gene lists [9,11]. Today, the biological interpretation of these gene lists is made possible by the availability of biological knowledge accumulated in public databases (e.g. Gene Ontology) and bioinformatic high-throughput enrichment tools [21]. In this study, Ontologizer [22] was used in order to analyse gene sets included in QTL confidence intervals and we tested if QTL regions are statistically enriched in some functional categories compared to the entire genome.

The objectives of this study were: (1) to identify QTLs controlling productivity, architecture and leaf traits in hybrid poplars; (2) to identify candidate genes under QTL intervals using enrichment tools and Gene Ontology.

Results

Trait variation, distribution and relationship between traits

Frequency distributions were not significantly departed from normal distribution (data not shown; Additional file 1). Genotype effect was significant for all traits (Data not shown). The P. deltoides parent showed higher overall growth in height and circumference than the P. trichocarpa parent (Table 1). The progeny showed generally a higher growth than both parents. A proportion of transgressive segregation (considered as heterosis by other authors) was calculated and it was high for all traits related to growth which was not the case for leaf traits. The coefficients of genetic variation (CVG) ranged from 0.7 to 20.3%. Values of heritability at genotype level were moderate to high (H2Genotype: 0.32 to 0.72) and H2Individual was low to moderate, ranging from 0.09 to 0.31.

Additional file 1. Distributions and relationships between all traits measured.

Format: PDF Size: 819KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

Table 1. Parental means, progeny mean,genotypic range, coefficient ofgenetic variation and heritabilitiesfor each trait measured

All leaf traits were significantly correlated, either negatively or positively and the highest correlation coefficients were observed between SLA and Δ (0.37), and between SLA and NM (0.36; Additional file 1). All productivity traits were significantly correlated except syllep1 with deltaH. The strongest correlations were observed between height and circumference within and between years, and between annual growth (deltaH, deltaC) and the second year measurements (height2, circum2). Carbon isotope discrimination is negatively correlated with all productivity traits. Number of sylleptic branches was correlated more tightly with circumference than with height. Correlations between leaf and productivity traits of the first year were higher than with second year productivity traits.

Genetic maps and QTL analysis

Among the 110 new SSR tested, only 6 were distorted and, as they were not linked between them, they were discarded for the linkage analysis. Genetic maps used for the detection of QTLs and their projection on the genome are summarized in Additional file 2. Briefly, genetic maps cover 3 126 cM and 3 222 cM for the P. deltoides and P. trichocarpa maps respectively, with a mean distance between markers of 18.75 cM. The mean numbers of marker per linkage groups were 8.0 and 6.1 for P. deltoides and P. trichocarpa respectively. A total of 67 and 81 markers anchored to the genome sequence for P. deltoides and P. trichocarpa respectively. A mean of 3.5 and 4.2 markers were anchored per linkage groups. All of them, with the exception of 2 linkage groups of the P. trichocarpa genetic map, were assigned to a chromosome (Figure 1). Estimated genome coverage was 78% and 66% for P. deltoides and P. trichocarpa map respectively.

Additional file 2. Summary statistics of the framework genetic maps and status of alignment on the genome.

Format: PDF Size: 86KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

thumbnailFigure 1. Framework linkage maps andQTLs from the segregationanalysis of the P. deltoides x P. trichocarpa pedigree aligned on P. trichocarpa Nisqually-1 sequence. Genome version assembly was Phytozome annotation v2.2. From the left to the right, P. deltoides female 73028–62 genetic map (in white), P. trichocarpa Nisqually-1 physical map with position of genome anchored markers (in black), and P. trichocarpa 101–74 male genetic map (in white). Scaffolds were numbered according to the v2.2 genome version. Additionally, extra scaffold (>19) containing markers mapped on the genetic maps were also shown. The length of bars is proportional to the map distance in cM or to sequence length in bp. Marker names are explained in Material and Methods. Markers in bolds are anchoring markers. Markers in bold and underlined are QTL flanking markers used for the projection of QTL confidence intervals on the physical map. QTLs were represented by vertical lines with horizontal small lines indicating start and stop of the confidence intervals and position of the LOD peak. Trait names were explained in Material and Methods.

The results of QTL detection are listed in Additional file 3 and details on genetic maps and QTL positions are graphically presented in Figure 1. A total of 77 QTL were detected explaining between 1.8% and 17.2% of the trait variation (at chromosome level P-value threshold of 0.05). The maximum was reached for NM on linkage group II on the P. deltoides map. An average of 3.3 and 3.7 QTLs were detected per trait respectively on the P. deltoides and P. trichocarpa maps, and total explained variance varied from 4.9% to 34.7%, maximum being reached by NM. Thirty five QTLs were detected on the P. deltoides map and 42 on the P. trichocarpa map.

Additional file 3. Positions of QTLs controlling 11 growth and leaf traits on genetic maps and on theP. trichocarpagenome Nisqually-1 v2.2.

Format: PDF Size: 327KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

All QTLs were not distributed evenly on the genome (P-value of chi squared test: 7 10-9), and one linkage group (V) came out to be a hot spot with 10 QTLs. Among the 27 QTLs controlling height and circumference measured the first and the second year, only few cases of co-location were identified (LG IV, LG V and LG X).

QTL projection on the genome

Among the 126 markers having sequence information (primer and gene sequences), only 10 did not show matching or showed inconsistent matching on the genome sequence (data not shown). Genetic and physical positions for 116 markers were used to calculate and compare within chromosome pair wise genetic and physical distances (Figure 2; Additional file 4). The global ratio between physical distance (bp) and genetic distance (cM) was 95,184 bp/cM for the P. deltoides map and 77,803 bp/cM for the P. trichocarpa map. This ratio varied between linkage groups, from 42,413 to 132,309 bp/cM for the P. deltoides linkage groups, and from 9,487 to 230,033 cM for the P. trichocarpa linkage groups.

Additional file 4. Position of genome anchoring markers on genetic maps and on genome sequence of P. trichocarpa Nisqually-1 v2.2.

Format: PDF Size: 367KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

thumbnailFigure 2. Global relationship between physicaldistance (in bp) andgenetic distance (in cM). The relationship is shown for both parental maps, P. deltoides map (circle), P. trichocarpa (triangles). Each point represents a physical and a genetic interval between 2 markers within a chromosome/linkage group.

Among the 77 detected QTLs, only 19 could not be projected on the genome (Additional file 3). For 18 QTLs, the corresponding LG carried only one marker that could be anchored on the genome and this configuration did not allow to correctly orientate the QTL on the genome. For one QTL (LG A), there was no anchoring marker. Mean size of projected QTL interval was 11,099,561 bp for P. deltoides and 8,096,353 bp for P. trichocarpa. Mean number of genes included in all QTLs confidence intervals for a particular trait was 4,216 and varied between 2,445 (Syllep1) and 5,937 (Height1; Table 2).

Table 2. Global analysis of geneannotation and gene ontology(GO) among gene setby traits

GO terms enrichment in QTL confidence intervals

Projecting the QTLs into physical genomic regions provided the opportunity to search for possible enrichment in gene functions that could be related to the traits under study. A rational and without a priori strategy to do this is to analyse the Gene Ontology (GO) annotation associated with the gene sets. The available Gene ontology (GO) annotation was relatively limited for Populus trichocarpa as compared to the model plant Arabidopsis thaliana. Indeed, searching for GO association in Phytozome plant genome database, 18,542 gene models have been found associated with at least one GO term (data not shown). In order to increase the number of annotated gene models, a Populus GO annotation has been inferred based on protein-protein similarity between Arabidopsis and poplar. At least one GO term annotation could be retrieved for 35,467 genes among the 40,668 poplar gene models identified in Phytozome, which represent 87.21% of genes annotated (see Material and Methods). Main results obtained from the gene ontology analysis are presented in Table 2. For each QTL intervals, the percentage of annotated genes did not differ significantly from the 87.21% determined above. Consequently, enrichment analysis was not biased by a distortion in the percentage of annotated genes within the QTLs. At a P-value threshold of 1%, the number of over-represented GO terms varied between 34 and 124 depending on the considered trait. The percentage of significantly over-represented (enriched) GO terms on total number of GO terms for a particular trait varied between 0.78% and 4.14%. There was no significant correlation between the cumulated size of QTL confidence intervals on the genome for a particular trait and the number of significant GO terms in the corresponding intervals (P-value = 0.21). Remarkably, a significant higher percentage of over-represented GO terms was observed for the gene set included in confidence intervals of all QTLs controlling the number of sylleptic branches (Syllep1). Interestingly, this trait showed the highest heritability (H2Individual and H2Genotype) and coefficient of genetic variation (CVG). Thus, we decided to explore further gene ontology enrichment within QTLs controlling number of sylleptic branches (Syllep1). Complete lists of GO terms significantly enriched are presented in Additional file 5.

Additional file 5. Results of GO term enrichment analysis for all traits.

Format: XLSX Size: 81KB Download fileOpen Data

Functional analysis of genes in QTL confidence intervals controlling number of sylleptic branches

An additional enrichment method called MGSA for model-gene based analysis, developed recently and integrated to the Ontologizer webtool, has been tested. This method is presented as faster and more accurate in identifying less redundant GO terms than previous methods such as the Topology-weighted methods [23]. Figure 3A presents a detailed view of the enrichment analysis using the Topology-Weighted (TW) method. Only one run is needed with this method that provides a P-value associated with each GO term. As already mentioned earlier by [23], the TW method may give some redundant results as observed for example in lines 3 to 8 for the GO term related to DNA binding. Results were not so redundant with MGSA (Figure 3B). However, the main disadvantage of MGSA is the lack of consistency between replicated analyses. This could be due to the Bayesian approach and must be taken in consideration. Several runs of MGSA were performed in order to be able to have a ranking of the most pertinent GO terms and to compare this ranking between TW and MGSA (Figure 3, Table 3). Significant variability between the 20 runs of MGSA was observed leading to the conclusion that the MGSA method did not gave reliable results. However, when comparing the first 10 terms identified using both enrichment methods, common GO term were found corresponding to the biological processes of adventitious root development (Table 3). The second most significant GO term referred to the process of shade avoidance, which could make sense regarding branch formation. If the 210 DNA binding and transcription factors terms identified were removed considering them as generic and thus not so biologically informative, the third enrichment class corresponded to genes potentially involved in the ripening process, an important process in fruit development but somewhat intriguing in shoot development. These three sets were analysed further.

thumbnailFigure 3. Comparison of enrichment analysesof GO terms usingTW or MGAS methods. Application to genes within QTLs for the number of sylleptic branches. (A) Ranked list of the 25 overrepresented terms using a Topology-Weighted (TW). (B) Ranked list of the top 25 terms identified by 20 runs of MGSA. Error bars (95 % confidence intervals) obtained with the 20 runs of MGSA. In case of MGSA each of the 25 terms was identified with a marginal value >0.5 in at least one of the 20 runs. GO:xxxxxxx: gene ontology accession; Green label and “P” prefix to GO accession refer to the ontology domain “biological process”; Yellow label and “F” : “molecular function”; magenta label and “C”: “cellular component”.

Table 3. The first ten rankedGO terms represented withinthe QTLs controlling numberof sylleptic branches

Eight genes corresponding to adventitious root development were found: POPTR_0002s02690, POPTR_0002s02700, POPTR_0002s02730, POPTR_0002s02740, POPTR_0002s02770, POPTR_0002s02790, POPTR_0002s02800, POPTR_0002s09050 (Additional file 6). All these genes belonged to the chromosome 2. In the text below, all gene names refer to Arabidopsis naming nomenclature. Annotation was analysed looking to the peptide homologs on Phytozome (See Additional file 6). Seven genes encode cytochrome P450, family 83, subfamily B and their protein homologs in Arabidopsis correspond to CYP83B1 / SUR2. The percentage of similarity at the amino acid level was between 60% and 65%. The last gene, POPTR_0002s09050, encodes an ARF17 homolog. These 8 genes were also gathered under the enriched GO term related to shade avoidance identified as the second most significant term. In addition, two other genes were within this group: POPTR_0002s06090 and POPTR_0005s22330. They were localized on chromosome 2 and chromosome 5 and they encode PAR2 (AT3G58850) homologs.

Additional file 6. Detailed annotation of genes with GO terms overrepresented for QTLs controlling number of sylleptic branches (Syllep1).

Format: PDF Size: 75KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

The third selected enrichment gathered five genes related to the biological process of ripening as inferred from ontology annotation in Arabidopsis (POPTR_0002s10240. POPTR_0002s10250. POPTR_0002s13340. POPTR_0005s17480. POPTR_0007s14390; Additional file 6). The official GO annotation of the two Arabidopsis homologous transcripts identified here (AT5G65380.1 and AT1G47530.1) was inferred from sequence similarity with a ripening regulated protein DDTFR18 identified in Solanum lycopersicum (GenBank accession AAG49032.1). The five poplar homologues genes originated from three different chromosomes. They encoded proteins belonging to the multidrug and toxic compound extrusion (MATE) protein family, included in the large superfamily of multidrug transporters [24].

Discussion

Are trait variation and correlations among traits consistent with previous studies?

The baseline for all genetic improvement is the availability of information on genetic variation for the traits of interest. Efforts have been made to describe growth and WUE variations and relationships among traits in different unrelated poplar cultivars and hybrid families under different water regimes [25-32]. High levels of genotypic variability for productivity and for Δ have been found under non limited water supply or moderate water deficit. In the present study, progeny mean values for stem height at the end of the first and second growing seasons were in agreement with those previously reported for other P. deltoides x P. trichocarpa F1 large family on the same trial [5,33]. Nevertheless, the range of values was lower. In a subset of 33 genotypes of the same family measured in an adjacent trial [29], height and circumference means were not comparable because of the trial management (pruned each year), but for leaf traits, mean values were similar, showing that the trial management did not influenced leaf trait variability. On the contrary, for a different P. deltoides x P. trichocarpa F1 [34] measured on the same trial, parental and progeny means for leaf traits (SLA, NM, CM), were higher than those reported in the present study.

A proportion of transgressive segregation was calculated to make comparisons with previous studies where it was considered as heterosis [5]. This proportion was lower for Circum2, deltaC and deltaH. These differences could be explained by a difference in circumference for the P. trichocarpa parent. The P. trichocarpa 101–74 parent of the present study had a lower juvenile growth than the P. trichocarpa ‘V24’ parent from [5]. For Populus spp., parental specific effect can explain a significant part of the trait variation in progenies (Bastien et al., in prep.). For leaf traits, low or no significant levels of transgressive segregation were observed as previously reported [5].

Analysis of the relationship among traits measured the first year shows that productive clones (high circumference) were characterized by low SLA (dense/ thick leaves) and low carbon isotope discrimination (high water-use efficiency). Water use efficient clones (low Δ) were characterized by low SLA (dense/thick leaves) and low leaf nitrogen content. Very low or no correlation was observed between leaf nitrogen content and growth traits, which is contradictory with previous studies on the subset of the same progeny or other progeny [29,35]. The strength of the links between productivity and Δ differ according to poplar species: a positive relationship was found for Populus davidiana[25], whereas negative correlations were detected for P. trichocarpa × P. deltoides and no correlation was observed in P. deltoides × P. nigra[26-28,36]. Occurrence of genetic variability and no systematic trade-off between these traits suggest that it should be possible to select genotypes combining large levels of productivity and WUE. The different trial managements and genetic or species backgrounds could explain these discrepancies.

Could QTLs be identified for growth and leaf traits?

Given the range of variation, the significant genotype effect and levels of heritability for all traits measured, significant results from the QTL detection were expected. Average QTL number for each trait was very similar to those reported previously [5] but less than average QTL number reported elsewhere [8,9]. This could be explained by a larger trait variation existing in a F2 progeny in the latter studies. A hot spot of QTL co-location on LG V was related to growth traits and SLA. These results were in agreement with the negative correlation observed between both types of traits. Additionally, the opposite effects of QTLs for both types of traits were also in agreement with the sign of trait correlation (LG Vb from P. deltoides map; Additional file 3). This linkage group in this particular pedigree was also involved in bud phenology (explaining up to 25% of the budset variation, [18]). Growing season length explained a large part of productivity in poplar [34] and could explain the colocation of growth and budset QTLs. On the other hand, for highly correlated traits, like circumference and height measured in the first and second growing seasons, the observed number of co-locations was lower than expected. This could be partly explained by the cut off (P-value) for declaring a QTL, which eliminate QTLs that were just below the threshold (data not shown). Additionally, the high correlation might be driven by plant intrinsic mechanistic correlations between traits.

All QTLs controlling Δ were on the P. trichocarpa parental map. Despite of a significant negative correlation between circumference measured the first year (Circum1) and carbon isotope discrimination (Δ), no QTLs co-location (overlapping confidence intervals) has been detected for these two traits. The map coverage was not exhaustive and could explain the absence of co-location. Nevertheless, availability of unlinked QTLs for different trait of interest has a particular advantage for breeding purposes: it opens here the possibility to select clones combining both high productivity and high water efficiency.

For comparison with published QTLs in other mapping pedigrees, we focused on studies with the same Populus species and with genetic maps carrying a significant number of genome anchoring markers [5,8,9]. In many cases, co-locations occurred on the same chromosomes only when pairs of studies were compared (Additional file 7). In only one case (LG X), a co-location occurred in the four studies compared ( [5,8,9] and the present study; Additional file 8). There were very few common anchoring markers between all genetic maps, which impeded a meta-analysis of QTLs, but anchoring on genome allowed aligning all the maps. The genome interval to consider included actually almost the whole chromosome, and need to be narrowed for further analysis and interpretation.

Additional file 7. Comparison of QTLs detected in the present study and in Dillen et al. (2009) and Rae et al. (2008, 2009) works.

Format: PDF Size: 199KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

Additional file 8. Comparison between linkage maps and QTLs on Linkage Group X from the present study and previous studies.

Format: PDF Size: 278KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

Could candidate genes for leaf and growth traits be identified through GO term enrichment analysis within QTL confidence intervals?

It was the first time GO term enrichment analysis is used in plants to identify putative candidate genes for QTLs. Generally, a priori identified candidate genes were searched within QTL intervals: their presence therefore “validate” putative important gene function related to the trait (see for example [37-39]). However, a recent analysis in animals provided evidence of over represented GO terms in QTL regions of the bovine genome [40]. Moreover, the authors showed that enrichment classifications are consistent with the trait category controlled by the QTLs. In the present study, for all traits, significant GO terms enrichment were found in QTL confidence intervals. For QTLs controlling sylleptic branches, 8 genes linked to the GO: 0048830 defining the biological process of adventitious root formation were found. These genes not only belong to the chromosome 2 but they are also clustered together on the chromosome. So they may have been identified by the enrichment methods because they are duplicated in tandem and not because they could be involved in the trait variation. This remark is also true for all associations found. However, the significant association due to a physical linkage between genes seems less plausible when the significant GO terms are associated with genes from different chromosomes. Five genes identified correspond to the enrichment in GO term related to ripening. These 5 genes are located on 3 different chromosomes. In that case, the association due to tandem duplication biases of this enrichment seems unlikely.

What is the meaning of finding enrichment in biological processes such as adventitious root formation and ripening when studying sylleptic branches formation?

In perennial dicotyledonous species such as poplar, sylleptic branches are formed either on the elongating stem during the growing season or at the time of regrowth, from arrested axillary buds formed the year before. In the present study, the number of sylleptic branches was measured at the end of the first growing season and do not proceed from axillary bud break. Therefore, sylleptic branches formation measured here might have some common features with the mechanism of shoot branching well studied in model plant such as Arabidopsis. The two class of genes related to adventitious root formation were SUR2 and ARF17 homologs. These GO classification make sense because SUR2 defines the first step in making indolic glucosinolates from indole-3-acetaldoxime (IAOx). Loss-of-function sur2 mutants block the production of glucosinolates from IAOx, leading to an increased IAOx flux for IAA biosynthesis. In that way, SUR2 participates in auxin homeostasis [41,42]. Likewise, ARF17 (AT1G77850) belongs to the AUXIN RESPONSE FACTOR (ARF) gene family products which together with the AUXIN / INDOLE-3-ACETIC ACID (Aux/IAA) proteins regulate auxin-mediated transcriptional activation / repression [43]. ARF17 was predicted to control adventitious rooting by modulating indole-3-acetic acid (IAA) homoeostasis [44]. Several hormones such as auxin, cytokinin, and the newly identified strigolactones are known to act directly or indirectly on bud outgrowth and to control shoot branching [45-48]. Consequently, the identified poplar genes could be related to auxin homeostasis regulation in shoot branching rather than adventitious root formation; and this makes sense with respect to the trait under study. These 8 genes were also gathered under the enriched GO term related to shade avoidance identified using the Topology-Weighted (Table 3; Additional file 6). In addition, two other genes are within this group: POPTR_0002s06090 and POPTR_0005s22330. They are localized on chromosome 2 and chromosome 5 and they encode PAR2 (AT3G58850) homologs. PAR2 functions as transcriptional repressor of auxin-responsive genes SAUR15 (AT4G38850) and SAUR68 (AT1G29510). It is obvious that the similarity was very low and thus the function largely undetermined. However, their co-localization with the SUR2 and ARF17 homologs was an argument to explore further their function in auxin response signalling in poplar.

Are the MATE-type transporters involved in auxin/strigolactones signalling?

The second most significant enrichment gathered five genes related to a GO term defining ‘ripening’. These five genes encode putative MATE efflux family proteins. MATE proteins are believed to function as proton-dependent efflux transporters. These classes of proteins were identified through their trans-membrane protein domain and were grouped in pfam database [49] under the pfam family accession PF01554. In Phytozome database, 73 poplar proteins carrying this protein domain were found. Similarly, 56 putative MATE proteins in Arabidopsis were identified. This is consistent with previous studies indicating that MATE proteins have exceedingly large numbers of homologs in plants per species, in that aspect contrasting to what is observed in bacteria and animals [50,51]. But little is known about their function in plants. One can hypothesize that the genes identified here could have a function in shoot branching through hormones regulation. In mammals, MATE-type transporters recognize hydrophobic hormones and transmitters, such as testosterone and corticosterone [50]. In Arabidopsis, TT12 was a MATE efflux protein proposed to transport glycosylated flavan-3-ols in vivo[52]. Therefore, plant MATE efflux proteins represent interesting short/long-distance transporter candidates for highly hydrophobic metabolites. In the process of shoot branching, the importance of auxin was already discussed. Also is to consider the newly identified plant hormone strigolactones derived from carotenoids [45,47]. Several genes involved in their biosynthesis have been identified but knowledge of both the biosynthesis pathway and the transport of strigolactones and their intermediates is still incomplete [53]. How some highly hydrophobic intermediates such as carotenoid-derived compound can be transported within the cytoplasm of the cell or on more long distance? The MATE family proteins identified are good candidates for such a function and need now to be considered further.

Conclusion

Beyond the detection of QTLs for growth and leaf traits, this study explored further the genomic resources as the genome sequence and its annotation to identify candidate genes for the traits under study. For sylleptic branches, candidate genes identified by enrichment tools analysis of genes inside QTL confidence intervals were promising. Nevertheless, the improvement of poplar genome annotation and the reduction of QTL confidence intervals might help to refine this strategy of candidate gene discovery.

Methods

Plant material

Poplar material consisted of a cloned 336 F1 progeny from an interspecific cross between the female Populus deltoides (Bartr. Ex Marsh.) ‘73028-62’ from Illinois and the male P. trichocarpa (Torr. and Gray) ’101-74’ from Washington State. Several pollinations were conducted in 1990, 1995 and 1996 by INRA (Orléans, France) to produce this F1 progeny. Cuttings were produced from stoolbeds of the same age.

Experimental design and trait measurement

The field trial was established in April 2003 from 25 cm- homogenous hardwood cuttings planted at an initial spacing of 0.75 m x 2 m, accommodating a plant density of 6670 trees per ha. The experimental trial, located in Central France (Ardon, 47°49’41''N, 1°54’39''E, 110 m), consisted in 6 randomized complete blocks where each F1 genotype and each parent was represented by one replicate. To reduce the border effects, a double border row was planted around the plantation. The plantation management included irrigation and the use of insecticides and fungicides as needed throughout two growing seasons.

As productivity traits, circumference and stem height were measured at the end of the first (winter 2003–2004) and second (winter 2004–2005) growing seasons as described in [54]; measured variables were named Circum1, Circum2 and Height1 and Heigth2 respectively. Total number of sylleptic branches (Syllep1) was counted at the end of the first growing season. Growth increment in height and circumference during the second growing season were calculated as: deltaH = Height2–Height1 and deltaC = Circum2–Circum1.

For leaf traits, on September 2nd, 2003, one fully illuminated mature leaf was collected on each tree; this leaf presented the largest width along the main stem (Foliar Index between 15 and 17; see [27] for details). Six calibrated discs of lamina (2 cm2) were cut from this leaf; leaf discs were then dried at 50 °C during 48 °C and weighed, and specific leaf area (SLA, cm2 g−1) was computed. Leaf discs were ground to fine powder for analysis of leaf carbon isotope composition (δ13C), carbon (CM) and nitrogen (NM) contents. All analyses were performed at the technical platform of functional ecology at the INRA-Nancy. One-milligram subsamples of ground material were enclosed in tin capsules and combusted. The CO2 produced by combustion was purified and its 13CO2/12CO2 ratio was analysed by a continuous flux isotope ratio mass spectrometer (IRMS Delta S, Finnigan MAT, Bremen, Germany) with a precision over measurements of ± 0.14 ‰. The discrimination between atmospheric CO2air assumed to be close to −8 ‰) and plant material (δplant) was calculated as Δ = (δair - δplant) / (1 + (δplant / 1000)) according to [55]. CM and NM were expressed on a dry weight basis (mg.g-1 DW).

Statistical analysis

For each variable, assumptions on residual distributions of the linear models were checked with the Shapiro–Wilk statistic. All statistical tests were considered significant at P ≤ 0.05. For adjustment of individual data to block effect, the following model was used: Yij = μ + Bi + εij where μ is the general mean and Bi is the effect of block i considered as fixed. Bi was calculated as the difference between the general mean of the whole family and the mean of block i (Bi = YYi).

To evaluate the genetic variation, the following model was used: Y’ij = μ + Gj + εij, where Gj is the effect of genotype j considered as random. Genetic and residual variance components (σG2 and σ2ε) were calculated by equating observed mean squares to expected mean squares and solving the resulting equations according to the Henderson III procedure [56,57]. The coefficient of genetic variation (CVG) was estimated as σG /MeanFamily. Broad-sense heritabilities were estimated on a genotypic basis, H2Genotype = σG2/(σG2 + (σε2/nj)), where nj is the average number of replicates per genotype, and on an individual basis, H2Individual = σG2/(σG2 + σε2) [58]. H2Genotype is reported to give information about the precision of estimated genetic values when genotypic means were used as phenotypic predictors. H2Individual can be considered as a reference value, calculated for one individual, and more easily comparable to literature values. The standard errors of broad-sense heritability were calculated as described previously [59].

Proportion of transgressive segregation was expressed as the percentage of superiority of hybrids over the mean of the two parents: (MeanFamily-MeanParents)/MeanParents) x 100 [60].

Genotyping and map construction

Genotyping was conducted using RFLP, STS, RAPD, and microsatellite markers in a first subset of 90 genotypes. Microsatellite and AFLP genotyping was then extended to 253 supplementary genotypes. Marker data were taken from Jorge et al. [61] and 110 new SSR markers were tested. RFLP and STS genotyping was conducted as described previously [62]. The letter “P” followed by a 3- to 4-digit number (e.g. P1273) refers to the RFLP markers. RAPD genotyping was performed according to [63]. RAPD markers were called using the Operon kit primer name followed by the molecular weight of the polymorphic band (e.g. M02-1150). AFLP genotyping was performed as described in [61]. AFLP markers were named after the code of the EcoRI/MseI combination from the kit followed by the band ranking on the gel (e.g. E5M5-7).

The SSR primers came from two different sources: (1) the International Populus Genome Consortium, SSR named "PMGC" (http://www.ornl.gov/sci/ipgc/ssr_resource.htm webcite), "ORPM" [64] and "WPMS" [65] and (2) SSR named "ai", "bi" and "bu" developed from public Populus spp. EST databases (see Additional file 9 for details).

Additional file 9. Details of EST – SSR used in the genetic map.

Format: PDF Size: 264KB Download file

This file can be viewed with: Adobe Acrobat ReaderOpen Data

Different labelling techniques were successively used for SSR genotyping: primers were labelled with γ-33P]ATP, with forward fluorochrome-labelled primer or the M13 tailing strategy [66]. These first two were described elsewhere [61]. The third one follow such procedure: PCR reactions were performed in a total volume of 10 μl containing 10 mM Tris–HCl, pH 8.3; 50 mM KCl; 1.5 mM MgCl2; 200 μM each of dATP, dCTP, dGTP and dTTP; 0.2 unit Taq Polymerase (Invitrogen); 5 pmoles reverse primer; 0.5 pmoles M13-tailed forward primer; 5 pmoles M13-labelled primer; and 40 ng DNA. PCR were conducted with 42 cycles of a 30-s denaturation at 94°C, 30-s annealing at 55°C or 52°C and 30-s extension at 72°C. Each forward primer was 5’-tailed with the M13 forward consensus sequence. The M13-tailed forward primers were then used in combination with a standard M13 primer labelled with fluorescent dye (6FAM, HEX or NED) at its 5’-end. The amplicons for each SSR marker were separately produced, diluted and pooled post-PCR by three-color multiplexes (6FAM, HEX, and NED) for polymorphism screening. Microsatellite polymorphisms were visualized using an ABI PRISM 3100 genetic analyzer (PE Applied Biosystems, Foster City, Calif.). SSR allele lengths were recorded by GeneScan and Genotyper or Genemapper softwares.

Markers significantly deviating from Mendelian segregation ratios [1:1] (i.e. P>χ2 < 1%) were eliminated from the linkage analysis. Computations were made with Mapmaker version 3.0b [67] and were based on the pseudo-testcross strategy that led to the construction of two parental maps [68]. Steps for the framework map construction were the same than [61]. Briefly, LOD threshold of 3.0 and a recombination fraction of 0.3 were used for grouping markers. Kosambi map function was used to calculate genetic distances. The best order within each linkage group was found using the following succession of commands: ‘order’, ‘compare’, ‘try’ and ‘ripple’ (likelihood difference between the first and second best order = 2.0). Genome coverage of framework genetic maps were calculated using Hulbert et al. [69] method modified by Chakravarti et al. ( [70]; method 3).

QTL analysis

QTLs were determined using MultiQTL 2.5 (http://www.multiqtl.com/ webcite; MultiQTL Ltd, Institute of Evolution, Haifa University, Haifa, Israel). The option ‘marker restoration’ was used to reduce the effect of missing information. The Kosambi mapping function was chosen for recalculation of maps on genotypic data. Single trait analysis was performed using a combination of interval mapping approach and multiple interval mapping. The entire genome was first scanned using the one QTL model and then using the two-linked QTL model. Permutation tests (1000 runs), comparing hypotheses H1 (there is one QTL in the chromosome), and H0 (no QTL in the chromosome) were run to obtain chromosome-wise statistical significance. In a second step, the genome was scanned for QTLs assuming a two-linked QTL model. For chromosomes for which a single QTL was already detected, permutation tests (1000 runs) were run to compare the hypotheses H2 (two-linked QTLs in the chromosome) versus H1. Subsequently, when p(H2vsH1) < 0.05, permutations were run to compare H2 vs. H0. A two-linked QTL model was only accepted, if one of the two confidence intervals were coinciding with the peaks of the single QTL model. To speed up calculation time, permutations for the two-linked QTL models were only conducted with 1000 runs when the p-value was <0.1 after 100 runs. In a last step, multiple interval mapping was performed including all the significant IM QTLs (single and 2 linked QTLs), and threshold was 0.05. For the remaining significant QTLs, permutations were run per chromosome, using thresholds (0.05) per chromosome. We further computed an adjusted Type I error rate at the chromosome level (αchr) using an Type I error rate at whole genome level of 0.05 and the formula: αchr = 1 – {1 – [1 – (1 – αg)1/M}m , where M is the total number of markers used for the QTL detection on each map and m the number of markers in the linkage group carrying the QTL [71]. Bootstrap analysis was performed to estimate the 95% confidence intervals. The total variance explained by all the QTLs for a same trait was estimated by the multiQTL model MIM for each parental map separately. Genetic maps were drawn using Mapchart [72].

Map and QTL projection on the genome

SSR markers and gene markers position on the P. trichocarpa Nisqually-1 genome (http://www.phytozome.net/poplar.php webcite; version 2 of the Populus genome assembly) were determined using blast algorithms and primers and probe sequences. An “averaged” ratio of base pair per cM (R) was calculated for each linkage group (LG) and each parental map using linear regression of physical distance (y) on genetic distance (x) with y = 0 when x = 0. This R ratio and positions of the nearest flanking markers were used to calculate physical start (yQ1) and physical stop (yQ2) positions of QTL confidence intervals following the formulas:

<a onClick="popup('http://www.biomedcentral.com/1471-2229/12/173/mathml/M1','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2229/12/173/mathml/M1">View MathML</a>

<a onClick="popup('http://www.biomedcentral.com/1471-2229/12/173/mathml/M2','MathML',630,470);return false;" target="_blank" href="http://www.biomedcentral.com/1471-2229/12/173/mathml/M2">View MathML</a>

where xQ1 and xQ2 are the start and stop genome position of the QTL confidence interval, xM1 and xM2 are genetic map position of the flanking markers. When only two genome anchoring markers were available for a LG, the ratio R was calculated as the base pair distance between the two markers divided by the genetic distance between these two markers. The R ratio for a given LG was used to calculate the physical length of a QTL confidence interval located in another LG only when the former LG belongs to the same chromosome and carries at least one genome anchoring marker. When conflicts existed in the marker order between genetic map and genome sequence, the nearest markers which exhibit an order consistent with the physical order were used. Gene models contained in the physical QTL intervals were retrieved using Biomart tool from Phytozome 8.0 database (annotation version 2.2). When one of the limits (start or stop) of the QTL confidence interval converted in bp fell into a gene model, the gene was included in the QTL interval.

Gene ontology analysis

Poplar gene models and Arabidopsis BLAST results against the TAIR10 version were obtained from Phytozome v8.0 in the version 2.2 of the annotation published in January 16th, 2012. Gene Ontology (GO) annotations for the 40,668 poplar gene loci were obtained by mapping the poplar gene loci to the Arabidopsis best-hit loci. GO annotations were available for 35,467 poplar gene loci and were mapped to 15,283 Arabidopsis loci; 13 and 43 were respectively mitochondrion- and chloroplast-encoded genes and were removed, the reminders are chromosome-encoded.

All enrichment analyses have been realized using the Topology-Weighted (TW, [73]) and model-based gene set analysis (MGSA, [23]) and implemented into Ontologizer web tool [22,74]. The TW method improves the classical enrichment analysis of GO terms by integrating GO graph topology on a global scale and giving to genes annotated with a GO term a weight based on the scores of neighbouring GO terms. On the other hand, MGSA analyses all categories at once by embedding them in a Bayesian network in which gene response is modelled as a function of the activation of biological categories. Probabilistic inference is used to identify the active categories.

Competing interests

The author(s) declare that they have no competing interests.

Authors’ contributions

RM collected the field data and interpreted the phenotypic data. JCL performed and interpreted enrichment analysis and helped to draft manuscript. CB analysed quantitative data and helped to draft the manuscript. PFB performed the genotyping of SSR markers and contributed to the map construction. MV selected the P. deltoides and P. trichocarpa parents and realised the original cross, added RAPD markers in the genetic map and collected the field data. NM collected field data and interpreted phenotypic data. FB supervised the PhD of RM, coordinated the ecophysiological part of the experiment, contributed to the acquisition, the analysis and the interpretation of the phenotypic data, and helped to draft the manuscript. VJ performed and interpreted the QTL mapping, made projection on the genome, collected the genes within QTL intervals, coordinated and wrote the draft manuscript.

Acknowledgments

R. Monclus was supported by a Ph.D. grant from the Conseil Régional, Région Centre, France and from the Institut National de la Recherche Agronomique. R. Monclus died during a car crash in September 2006. This research was supported by the European Commission through the Directorate General Research within the Fifth Framework Programme for Research—Quality of Life and Management of the Living Resources Programme, contract N° QLK5-CT-2002-00953 (POPYOMICS; http://www.soton.ac.uk/~popyomic). The authors thank Jean Gauvin and Patrick Poursat (UE0995 GBFor, INRA Orléans, France) for their help with the management of field experiments and with the data collection, and Patricia Faivre-Rampant, Marie-Christine Goué, Marie-Claude Lesage-Descauses, Bénédicte Le Guerroué and Vanina Guérin for their contribution to the genetic map marker enrichment. Isotopic and elemental analyses were performed by technical platform of functional ecology (OC 081) at INRA Nancy – UMR1137 EEF. Authors thank also the two reviewers and the editor for their useful comments on the manuscript.

References

  1. Price AH: Believe it or not, QTLs are accurate!

    Trends Plant Sci 2006, 11(5):213-216. PubMed Abstract | Publisher Full Text OpenURL

  2. Morrell PL, Buckler ES, Ross-Ibarra J: Crop genomics: advances and applications.

    Nat Rev Genet 2012, 13(2):85-96. OpenURL

  3. Boerjan W, Vuylsteke M: Integrative genetical genomics in Arabidopsis.

    Nat Genet 2009, 41(2):144-145. PubMed Abstract | Publisher Full Text OpenURL

  4. Bradshaw HDJ, Stettler RF: Molecular genetics of growth and development in Populus. IV. Mapping QTLs with large effects on growth, form, and phenology traits in a forest tree.

    Genetics 1995, 139:963-973. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  5. Dillen SY, Storme V, Marron N, Bastien C, Neyrinck S, Steenackers M, Ceulemans R, Boerjan W: Genomic regions involved in productivity of two interspecific poplar families in Europe. 1. Stem height, circumference and volume.

    Tree Genetics and Genomes 2009, 5(2):375-376. Publisher Full Text OpenURL

  6. Marron N, Ricciotti L, Bastien C, Beritognolo I, Gaudet M, Paolucci I, Fabbrini F, Salani F, Dillen SY, Ceulemans R, et al.: Plasticity of growth and biomass production of an intraspecific Populus alba family grown at three sites across Europe during three growing seasons.

    Canadian Journal of Forest Research 2010, 40(10):1887-1903. Publisher Full Text OpenURL

  7. Novaes E, Osorio L, Drost DR, Miles BL, Boaventura-Novaes CRD, Benedict C, Dervinis C, Yu Q, Sykes R, Davis M, et al.: Quantitative genetic analysis of biomass and wood chemistry of Populus under different nitrogen levels.

    New Phytol 2009, 182(4):878-890. PubMed Abstract | Publisher Full Text OpenURL

  8. Rae AM, Pinel MPC, Bastien C, Sabatti M, Street NR, Tucker J, Dixon C, Marron N, Dillen SY, Taylor G: QTL for yield in bioenergy Populus: identifying GxE interactions from growth at three contrasting sites.

    Tree Genetics & Genomes 2008, 4(1):97-112. PubMed Abstract | Publisher Full Text OpenURL

  9. Rae AM, Street NR, Robinson KM, Harris N, Taylor G: Five QTL hotspots for yield in short rotation coppice bioenergy poplar: the Poplar Biomass Loci.

    BMC Plant Biol 2009., 9(23) OpenURL

  10. Rae AM, Tricker PJ, Bunn SM, Taylor G: Adaptation of tree growth to elevated CO2: quantitative trait loci for biomass in Populus.

    New Phytol 2007, 175(1):59-69. PubMed Abstract | Publisher Full Text OpenURL

  11. Street NR, Skogstrom O, Sjodin A, Tucker J, Rodriguez-Acosta M, Nilsson P, Jansson S, Taylor G: The genetics and genomics of the drought response in Populus.

    Plant J 2006, 48(3):321-341. PubMed Abstract | Publisher Full Text OpenURL

  12. Street NR, Tallis MJ, Tucker J, Brosche M, Kangasjarvi J, Broadmeadow M, Taylor G: The physiological, transcriptional and genetic responses of an ozone- sensitive and an ozone tolerant poplar and selected extremes of their F2 progeny.

    Environ Pollut 2011, 159(1):45-54. PubMed Abstract | Publisher Full Text OpenURL

  13. Wu RL: Genetic mapping of QTLs affecting tree growth and architecture in Populus: implication for ideotype breeding.

    Theor Appl Genet 1998, 96(3–4):447-457. OpenURL

  14. Wu R, Bradshaw HD, Stettler RF: Developmental quantitative genetics of growth in Populus.

    Theor Appl Genet 1998, 97(7):1110-1119. Publisher Full Text OpenURL

  15. Wullschleger S, Yin TM, DiFazio SP, Tschaplinski TJ, Gunter LE, Davis MF, Tuskan GA: Phenotypic variation in growth and biomass distribution for two advanced-generation pedigrees of hybrid poplar.

    Canadian Journal of Forest Research 2005, 35(8):1779-1789. Publisher Full Text OpenURL

  16. Zhang D, Zhang Z, Yang K: QTL analysis of growth and wood chemical content traits in an interspecific backcross family of white poplar (Populus tomentosa x P. bolleana) x P. tomentosa.

    Canadian Journal of Forest Research 2006, 36(8):2015-2023. Publisher Full Text OpenURL

  17. Tuskan GA, DiFazio S, Jansson S, Bohlmann J, Grigoriev I, Hellsten U, Putnam N, Ralph S, Rombauts S, Salamov A, et al.: The genome of black cottonwood, Populus trichocarpa (Torr. & Gray).

    Science 2006, 31(5793):1596-1604. OpenURL

  18. Rohde A, Storme V, Jorge V, Gaudet M, Vitacolonna N, Fabbrini F, Ruttink T, Zaina G, Marron N, Dillen S, et al.: Bud set in poplar - genetic dissection of a complex trait in natural and hybrid populations.

    New Phytol 2011, 189(1):106-121. PubMed Abstract | Publisher Full Text OpenURL

  19. Labbé J, Jorge V, Kohler A, Vion P, Marçais B, Bastien C, Tuskan G, Martin F, Le Tacon F: Identification of quantitative trait loci affecting ectomycorrhizal symbiosis in an interspecific poplar F1 cross and differential expression of genes in ectomycorrhizas of the two parents: Populus deltoides and Populus trichocarpa.

    Tree Genetics & Genomes 2011, 7(3):617-627. PubMed Abstract | Publisher Full Text OpenURL

  20. Ranjan P, Yin T, Zhang X, Kalluri U, Yang X, Jawdy S, Tuskan G: Bioinformatics-Based Identification of Candidate Genes from QTLs Associated with Cell Wall Traits in Populus.

    BioEnergy Research 2010, 3(2):172-182. Publisher Full Text OpenURL

  21. Huang DW, Sherman BT, Lempicki RA: Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists.

    Nucleic Acids Res 2009, 37(1):1-13. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  22. Bauer S, Grossmann S, Vingron M, Robinson PN: Ontologizer 2.0 - a multifunctional tool for GO term enrichment analysis and data exploration.

    Bioinformatics 2008, 24(14):1650-1651. PubMed Abstract | Publisher Full Text OpenURL

  23. Bauer S, Gagneur J, Robinson PN: GOing Bayesian: model-based gene set analysis of genome-scale data.

    Nucleic Acids Res 2010, 38(11):3523-3532. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  24. Brown MH, Paulsen IT, Skurray RA: The multidrug efflux protein NorM is a prototype of a new family of transporters.

    Mol Microbiol 1999, 31(1):394-395. PubMed Abstract | Publisher Full Text OpenURL

  25. Zhang XL, Zang RG, Li CY: Population differences in physiological and morphological adaptations of Populus davidiana seedlings in response to progressive drought stress.

    Plant Sci 2004, 166(3):791-797. Publisher Full Text OpenURL

  26. Marron N, Villar M, Dreyer E, Delay D, Boudouresque E, Petit J-M, Delmotte FM, Guehl J-M, Brignolas F: Diversity of leaf traits related to productivity in 31 Populus deltoides × Populus nigra clones.

    Tree Physiol 2005, 25(4):425-435. PubMed Abstract | Publisher Full Text OpenURL

  27. Monclus R, Dreyer E, Delmotte FM, Villar M, Delay D, Boudouresque E, Petit JM, Marron N, Brechet C, Brignolas F: Productivity, leaf traits and carbon isotope discrimination in 29 Populus deltoides x P. nigra clones.

    New Phytol 2005, 167(1):53-62. PubMed Abstract | Publisher Full Text OpenURL

  28. Monclus R, Dreyer E, Villar M, Delmotte FM, Delay D, Petit J-M, Barbaroux C, Thiec D, Brechet C, Brignolas F: Impact of drought on productivity and water use efficiency in 29 genotypes of Populus deltoides x Populus nigra.

    New Phytol 2006, 169(4):765-777. PubMed Abstract | Publisher Full Text OpenURL

  29. Monclus R, Villar M, Barbaroux C, Bastien C, Fichot R, Delmotte FM, Delay D, Petit JM, Brechet C, Dreyer E, et al.: Productivity, water-use efficiency and tolerance to moderate water deficit correlate in 33 poplar genotypes from a Populus deltoides x Populus trichocarpa F1 progeny.

    Tree Physiol 2009, 29(11):1329-1339. PubMed Abstract | Publisher Full Text OpenURL

  30. Dillen SY, Marron N, Koch B, Ceulemans R: Genetic variation of stomatal traits and carbon isotope discrimination in two hybrid poplar families (Populus deltoides 'S9–2' x P. nigra 'Ghoy' and P. deltoides 'S9–2' x P. trichocarpa 'V24').

    Ann Bot 2008, 102(3):399-407. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  31. Dillen SY, Monclus R, Barbaroux C, Bastien C, Ceulemans R, Dreyer E, Villar M, Brignolas F, Marron N: Is the ranking of poplar genotypes for leaf carbon isotope discrimination stable across sites and years in two different full-sib families?

    Ann Sci 2011, 68(7):1265-1275. Publisher Full Text OpenURL

  32. Chamaillard S, Fichot R, Vincent-Barbaroux C, Bastien C, Depierreux C, Dreyer E, Villar M, Brignolas F: Variations in bulk leaf carbon isotope discrimination, growth and related leaf traits among three Populus nigra L. populations.

    Tree Physiol 2011, 31(10):1076-1087. PubMed Abstract | Publisher Full Text OpenURL

  33. Marron N, Bastien C, Sabatti M, Taylor G, Ceulemans R: Plasticity of growth and sylleptic branchiness in two poplar families grown at three sites across Europe.

    Tree Physiol 2006, 26(7):935-946. PubMed Abstract | Publisher Full Text OpenURL

  34. Marron N, Dillen SY, Ceulemans R: Evaluation of leaf traits for indirect selection of high yielding poplar hybrids.

    Environmental and Experimental Botany 2007, 61(2):103-116. Publisher Full Text OpenURL

  35. Marron N, Ceulemans R: Genetic variation of leaf traits related to productivity in a Populus deltoides x Populus nigra family.

    Canadian Journal of Forest Research 2006, 36(2):390-400. Publisher Full Text OpenURL

  36. Rae AM, Robinson KM, Street NR, Taylor G: Morphological and physiological traits influencing biomass productivity in short-rotation coppice poplar.

    Canadian Journal of Forest Research 2004, 34(7):1488-1498. Publisher Full Text OpenURL

  37. Borevitz JO, Chory J: Genomics tools for QTL analysis and gene discovery.

    Curr Opin Plant Biol 2004, 7(2):132-136. PubMed Abstract | Publisher Full Text OpenURL

  38. Brotman Y, Riewe D, Lisec J, Meyer RC, Willmitzer L, Altmann T: Identification of enzymatic and regulatory genes of plant metabolism through QTL analysis in Arabidopsis.

    J Plant Physiol 2011, 168(12):1387-1394. PubMed Abstract | Publisher Full Text OpenURL

  39. Marguerit E, Brendel O, Lebon E, Van Leeuwen C, Ollat N: Rootstock control of scion transpiration and its acclimation to water deficit are controlled by different genes.

    New Phytol 2012, 194(2):416-429. PubMed Abstract | Publisher Full Text OpenURL

  40. Salih H, Adelson D: QTL global meta-analysis: are trait determining genes clustered?

    BMC Genomics 2009, 10(1):184. PubMed Abstract | BioMed Central Full Text | PubMed Central Full Text OpenURL

  41. Zhao YD: Auxin Biosynthesis and Its Role in Plant Development.

    In Annual Review of Plant Biology, Vol 61 Edited by Merchant S, Briggs WR, Ort D. 2010, 49-64. OpenURL

  42. Barlier I, Kowalczyk M, Marchant A, Ljung K, Bhalerao R, Bennett M, Sandberg G, Bellini C: The SUR2 gene of Arabidopsis thaliana encodes the cytochrome P450 CYP83B1, a modulator of auxin homeostasis.

    Proc Natl Acad Sci USA 2000, 97(26):14819-14824. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  43. Okushima Y, Overvoorde PJ, Arima K, Alonso JM, Chan A, Chang C, Ecker JR, Hughes B, Lui A, Nguyen D, et al.: Functional genomic analysis of the AUXIN RESPONSE FACTOR gene family members in Arabidopsis thaliana: Unique and overlapping functions of ARF7 and ARF19.

    Plant Cell 2005, 17(2):444-463. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  44. Sorin C, Bussell JD, Camus I, Ljung K, Kowalczyk M, Geiss G, McKhann H, Garcion C, Vaucheret H, Sandberg G, et al.: Auxin and light control of adventitious rooting in Arabidopsis require ARGONAUTE1.

    Plant Cell 2005, 17(5):1343-1359. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  45. Gomez-Roldan V, Fermas S, Brewer PB, Puech-Pages V, Dun EA, Pillot JP, Letisse F, Matusova R, Danoun S, Portais JC, et al.: Strigolactone inhibition of shoot branching.

    Nature 2008, 455(7210):189-U122. PubMed Abstract | Publisher Full Text OpenURL

  46. Ongaro V, Leyser O: Hormonal control of shoot branching.

    J Exp Bot 2008, 59(1):67-74. PubMed Abstract | Publisher Full Text OpenURL

  47. Umehara M, Hanada A, Yoshida S, Akiyama K, Arite T, Takeda-Kamiya N, Magome H, Kamiya Y, Shirasu K, Yoneyama K, et al.: Inhibition of shoot branching by new terpenoid plant hormones.

    Nature 2008, 455(7210):195-U129. PubMed Abstract | Publisher Full Text OpenURL

  48. Beveridge CA, Kyozuka J: New genes in the strigolactone-related shoot branching pathway.

    Curr Opin Plant Biol 2010, 13(1):34-39. PubMed Abstract | Publisher Full Text OpenURL

  49. Sonnhammer ELL, Eddy SR, Durbin R: Pfam: A comprehensive database of protein domain families based on seed alignments.

    Proteins-Structure Function and Genetics 1997, 28(3):405-420. Publisher Full Text OpenURL

  50. Omote H, Hiasa M, Matsumoto T, Otsuka M, Moriyama Y: The MATE proteins as fundamental transporters of metabolic and xenobiotic organic cations.

    Trends Pharmacol Sci 2006, 27(11):587-593. PubMed Abstract | Publisher Full Text OpenURL

  51. Kuroda T, Tsuchiya T: Multidrug efflux transporters in the MATE family.

    Biochimica Et Biophysica Acta-Proteins and Proteomics 2009, 1794(5):763-768. Publisher Full Text OpenURL

  52. Marinova K, Pourcel L, Weder B, Schwarz M, Barron D, Routaboul JM, Debeaujon I, Klein M: The Arabidopsis MATE transporter TT12 acts as a vacuolar flavonoid/H + −antiporter active in proanthocyanidin-accumulating cells of the seed coat.

    Plant Cell 2007, 19(6):2023-2038. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  53. Goulet C, Klee HJ: Climbing the Branches of the Strigolactones Pathway One Discovery at a Time.

    Plant Physiol 2010, 154(2):493-496. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  54. Dillen SY, Marron N, Bastien C, Ricciotti L, Salani F, Sabatti M, Pinel MPC, Rae AM, Taylor G, Ceulemans R: Effects of environment and progeny on biomass estimations of five hybrid poplar families grown at three contrasting sites across Europe.

    Forest Ecol Manag 2007, 252(1–3):12-23. OpenURL

  55. Farquhar GD, Richards RA: Isotopic composition of plant carbon correlates with water-use efficiency of wheat genotypes.

    Australian Journal of Plant Physiology 1984, 11(6):539-552. Publisher Full Text OpenURL

  56. Henderson CR: Estimation of variance and co-variance components.

    Biometrics 1953, 9:226-252. Publisher Full Text OpenURL

  57. Searle SR, Casella G, McCulloch CE: Variance components. New York: Wiley; 1992. OpenURL

  58. Nyquist WE: Estimation of heritability and prediction of selection response in plant populations.

    Critical Reviews in Plant Sciences 1991, 10(3):235-322. Publisher Full Text OpenURL

  59. Singh M, Ceccarelli S, Hamblin J: Estimation of Heritability from Varietal Trials Data.

    Theor Appl Genet 1993, 86(4):437-441. Publisher Full Text OpenURL

  60. Li B, Wu R: Heterosis and genotype × environment interactions of juvenile aspens in two contrasting sites.

    Canadian Journal of Forest Research 1997, 27:1525-1537. OpenURL

  61. Jorge V, Dowkiw A, Faivre Rampant P, Bastien C: Genetic architecture of qualitative and quantitative Melampsora larici-populina leaf rust resistance in hybrid poplar: genetic mapping and QTL detection.

    New Phytol 2005, 167:113-127. PubMed Abstract | Publisher Full Text OpenURL

  62. Bradshaw HD Jr, Stettler RF: Molecular genetics of growth and development in Populus. I. Triploidy in hybrid poplars.

    Theor Appl Genet 1993, 86(2–3):301-307. OpenURL

  63. Villar M, Lefèvre F, Bradshaw HDJ, Teissier du Cros E: Molecular genetics of rust resistance in Poplars (Melampsora larici-populina Kleb. / Populus sp.) by bulked-segregant analysis in a 2x2 factorial mating design.

    Genetics 1996, 143:531-536. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  64. Tuskan GA, Gunter LE, Yang ZMK, Yin TM, Sewell MM, DiFazio SP: Characterization of microsatellites revealed by genomic sequencing of Populus trichocarpa.

    Canadian Journal of Forest Research 2004, 34(1):85-93. Publisher Full Text OpenURL

  65. van der Schoot J, Pospiskova M, Vosman B, Smulders MJM: Development and characterization of microsatellittes markers in black poplar.

    Theoritical and Applied Genetics 2000, 101:317-322. Publisher Full Text OpenURL

  66. Schuelke M: An economic method for the fluorescent labeling of PCR fragments.

    Nat Biotechnol 2000, 18:233-234. PubMed Abstract | Publisher Full Text OpenURL

  67. Lander ES, Botstein D: Mapping mendelian factors underlying quantitative traits using RFLP linkage maps.

    Genetics 1989, 121:185-199. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  68. Grattapaglia D, Sederoff R: Genetic linkage maps of Eucalyptus grandis and Eucalyptus urophylla using a Pseudo-Testcross: Mapping strategy and RAPD markers.

    Genetics 1994, 137:1121-1137. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  69. Hulbert SH, Ilott TW, Legg EJ, Lincoln SE, Lander ES, Michelmore RW: Genetic analysis of the fungus, Bremia lactucae, using restriction fragment length polymorphisms.

    Genetics 1988, 120:947-958. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  70. Chakravarti A, Lasher LK, Reefer JE: A Maximum-Likelihood Method for Estimating Genome Length Using Genetic-Linkage Data.

    Genetics 1991, 128(1):175-182. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  71. Saintagne C, Bodenes C, Barreneche T, Pot D, Plomion C, Kremer A: Distribution of genomic regions differentiating oak species assessed by QTL detection.

    Heredity 2004, 92(1):20-30. PubMed Abstract | Publisher Full Text OpenURL

  72. Voorrips RE: MapChart: Software for the graphical presentation of linkage maps and QTLs.

    J Hered 2002, 93(1):77-78. PubMed Abstract | Publisher Full Text OpenURL

  73. Alexa A, Rahnenfuhrer J, Lengauer T: Improved scoring of functional groups from gene expression data by decorrelating GO graph structure.

    Bioinformatics 2006, 22(13):1600-1607. PubMed Abstract | Publisher Full Text OpenURL

  74. Robinson PN, Wollstein A, Bohme U, Beattie B: Ontologizing gene-expression microarray data: characterizing clusters with Gene Ontology.

    Bioinformatics 2004, 20(6):979-981. PubMed Abstract | Publisher Full Text OpenURL