Introduction

Soybean (Glycine max L.) is an important crop with a high protein and oil content, and is widely used for food and feed and chemical industry. In China, with the increasing demand for soybean and the shrinking of planting area, abundant resources and diversified planting systems should be maximized to develop the soybean industry and to breed high-yield varieties.

Maize-soybean relay strip intercropping remarkably increases the planting area and production of soybean and maintains the yield of maize1,2. In this planting pattern, soybean is subjected to shading at the co-growth stage. The morphological and photosynthetic characteristics of soybean leaves are considerably affected under shading. In shading, soybean leaf size inhibited by controlling cell proliferation and enlargement3, and the dry mass distribution increases in stem but decreases in leaves4. Poor light also reduces the specific leaflet weight and the photosynthetic rate5. In serious cases, photosynthetic capacity and yield decrease1.

More than 88 quantitative trait loci (QTLs) associated with soybean leaf-related traits have been reported. Of these QTLs, 64 QTLs for leaf-related traits in Soybase (https://soybase.org), including 35 QTLs for the leaflet area6,7,8,9,10,11,12, 2 QTLs for specific leaflet area11, 1 QTL for leaflet weight11, 15 QTLs for specific leaflet weight8,12, and 11 QTLs for leaflet ash13. Other scholars also detected some QTLs related leaf traits of soybean. Shi et al.14 reported 8 QTLs for leaf length and 9 QTLs for leaf width by using recombination inbred lines (RILs) derived from a cross between ‘Charleston’ and ‘Dongnong 594’. Kim et al.15 identified 7 QTLs for leaflet length and width in two F2:10 populations. The QTLs of leaf traits associated with shading were relatively few. Only few QTLs of other crops have been identified under poor light. For example, 5 QTLs are related to the increase in leaf area in cucumber16, and 6 QTLs are associated with the specific leaf weight in maize17. However, the QTL of soybean in shading has yet to be reported.

In the present study, we analyzed the leaf phenotype under monoculture (M) and relay intercropping (RI), developed a high-density soybean molecular genetic map by using a specific-locus amplified fragment sequencing (SLAF-seq) technology, and identified the QTL for leaf-related traits by utilizing RILs derived from ‘Nandou 12’ and ‘Jiuyuehuang’, and determined the main effect of QTLs and select candidate genes that may influence the leaf phenotype through Gene Ontology (GO) enrichment analysis.

Results

Phenotypic variance and correlation

‘Jiuyuehuang’ had greater compound leaf number (CLN), total leaflet area (TLA), and leaflet dry weight (LDW) in monoculture and relay intercropping than ‘Nandou 12’ (Table 1). ‘Nandou 12’ had greater specific leaflet weight (SLW) and leaflet dry weight ratio (LDWR) in relay intercropping than ‘Jiuyuehuang’, but opposite in monoculture, and specific leaflet area (SLA) of ‘Jiuyuehuang’ was similar. The phenotypic distribution and the normality of the RILs were shown in Fig. 1. The results of variance analysis indicated the significant variation in CLN, TLA, SLA, SLW, and LDW except LDWR between monoculture and relay intercropping. CLN, TLA, SLW, and LDW in relay intercropping were significantly lower than those in monoculture (P < 0.05). By contrast, SLA in relay intercropping was significantly higher than that in monoculture. These results revealed the susceptibility of leaf phenotype to shading.

Table 1 Phenotypic performance of leaf-related traits in parents and RILs under monoculture (M) and relay intercropping (RI) in 2015 and 2016.
Full size table
Figure 1
figure 1

Histograms of frequency distribution for compound leaf number (a), total leaflet area (b), specific leaflet area (c), specific leaflet weight (d), leaflet dry weight (e) and leaflet dry weight ratio (f) in the RILs population. In each panel, the upward side is the distribution of the phenotypic data in relay intercropping (RI), whereas the downward side is in the monoculture (M). The black bars represent the data in 2015, and the slash bars correspond to the data in 2016. Arrows indicate the phenotypic value ranges of ‘Nandou 12’ (N) and ‘Jiuyuehuang’ (J) in 2 years.

Full size image

Supplementary Table S1 shows the Pearson correlation coefficients. Significantly positive correlations were found between CLN and TLA, CLN and LDW, and TLA and LDW, whereas significantly negative correlations were observed in SLA and SLW in two different planting patterns in both years.

Genetic map

The whole genome resequencing of the two parents and the SLAF-seq of RILs population revealed that 221.67 Gb sequence reads were obtained. The read number was 1,018.34 Mb, and the average sequencing Q30 was 94.70%. The average GC content was 38.69%, and the sample GC distribution was normal. A total of 931,337 high-quality polymorphic single-nucleotide polymorphism (SNP) sites were detected, and 6,552 SNPs were used for high-density linkage map construction. After filtering out the markers that will affect collinearity, there were 6,366 SNPs being mapped into 20 linkage groups by using Highmap. The total genetic distance was 2,818.67 cM, and the average interval ranged from 0.27 cM to 0.70 cM by chromosome with an overall average marker distance of 0.44 cM. In Table 2, Chr03 (290.11 cM) was the largest and had the highest number of markers (773), whereas Chr13 (76.88 cM) was the shortest. Chr04 had the fewest markers (94). Chr02 had the minimal average map distance between markers (0.27 cM). Approximately 99.10% of the intervals between adjacent markers on the genetic map were less than 5 cM, and only the intervals between adjacent markers on Chr01 were all less than 5 cM. The spearman correlations of 20 linkage groups between the linkage map and the genome map were also analyzed. Each correlation coefficient of 20 linkage groups was nearly close to 1, which showed a relatively high collinearity between the linkage groups and the soybean reference genome (Fig. 2). In the 20 linkage groups, Chr09 had the highest collinearity with a correlation coefficient (0.999).

Table 2 Characteristics of the 20 chromosomes in the genetic map with 6366 SNPs.
Full size table
Figure 2
figure 2

Collinearity of the genetic map and the physical map. The corresponding relationship and the position relationship between the soybean chromosomes (Chr) and the linkage group (LG) of the genetic map are shown.

Full size image

QTL analysis

A total of 19 QTLs for six leaf-related traits were identified in nine different chromosomes (Chr03, Chr05, Chr06, Chr07, Chr08, Chr12, Chr14, Chr15, and Chr17) in 2 years (Fig. 3 and Table 3), and 12 QTLs were identified in monoculture with their phenotypic variance effect ranging from 4.00% (qLDW15-119) to 14.47% (qLDW15-98), and seven QTLs in relay intercropping accounted for 5.37% (qCLN15-23) to 13.88% (qLDWR5-19) of phenotypic variations. Furthermore, 4 QTLs associated with SLA and SLW were only detected in monoculture, whereas 15 QTLs related to CLN, LDW, and LDWR were detected in monoculture and relay intercropping. Of 19 leaf-related QTLs, sixteen QTLs showed positive additive effects from ‘Jiuyuehuang’, whereas three QTLs were negative additive effects with ‘Nandou 12’. In addition, 3 QTL hotspots flanked by Marker374018 and Marker376290 on Chr06, Marker676371 and Marker677344 on Chr14, and Marker732404 and Marker728840 on Chr15 were identified.

Figure 3
figure 3

The QTL mapping of six leaf-related traits under monoculture and relay intercropping in 2015 and 2016. The linkage groups of QTLs are listed. The position of each marker is shown on the left side of the linkage group, while QTLs are marked on the right. The black cubes are the QTLs detected under monoculture, and the slash cubes are the QTLs detected under relay intercropping. The blue letters represent the QTLs in 2015, and the red letters represent the QTLs in 2016.

Full size image
Table 3 QTLs detected in Nandou 12 × Jiuyuehuang RILs under monoculture and relay intercropping in 2015 and 2016.
Full size table

The phenotypic data on leaf traits and genotypic data on markers were analyzed by QTL.gCIMapping.GUI software. The results indicated that QTL qLDW15-98 detected in 2016 under monoculture was precisely determined at 96.00 cM on Chr15 with 14.59% of phenotypic variations, being in marker interval of Marker732404–Marker738840. In addition, there was a QTL (LOD = 3.51) associated to CLN in 2015 under monoculture closing to qCLN17-3, which was located on 9.36 cM between Marker969731–Marker969731 with 9.99% of phenotypic variations.

GO enrichment analysis based on co-located QTLs

In terms of co-located QTL, 3 QTL hotspots were investigated in detail. Our results revealed that 6 QTLs associated with four traits explained 5.95% to 14.47% of the phenotypic variations. To further identify the genes that were related to the leaves in the three regions, we retrieved gene calls and annotations by using Glyma.Wm82.a2.v1 gene model from Soybase (https://soybase.org). Using GO enrichment analysis, we found 160, 66, and 235 genes within the three clusters on Chr06, Chr14, and Chr15, respectively. Among them, 129 annotated genes were closely related to leaves and could be classified into five groups (Table 4). The first group contained 45 genes associated with phytohormone regulation, including hormones, such as auxin, gibberellin, ethylene, and abscisic acid, which could be related to leaflet number. The second group included 48 genes related to photosynthesis, including those that played a role in the process of photosynthesis, light quality, light intensity, and relevant carbohydrates, which were implicated in leaflet dry weight and area18. The third group comprised 85 genes associated with cellular processes, including cell division, cell differentiation, and cell growth, which might affect the leaflet area. The fourth group consisted of 11 genes associated with morphogenesis, including photomorphogenesis, cell morphogenesis, and leaf morphogenesis, which could be related to leaflet shape and area. The fifth group was composed of 32 genes associated with leaf development and leaf growth-related factors, such as water, nitrogen, and phosphorus, which directly influenced leaves. In terms of the predicted function of the five groups, three predicted genes (Glyma06G296500, Glyma14G087100, Glyma15G154000) were selected as the best candidate genes that might affect the leaf phenotype because they referred to various biological processes (Supplementary Table S2). Glyma06G296500 encodes a GH3-related gene involved in red light-specific hypocotyl elongation. GO analysis showed that this gene is also involved in response to auxin and light stimulus in Arabidopsis thaliana. Glyma14G087100 encodes cyclin-dependent protein kinase 3 and participates in cell cycle, cell division, and cell cycle regulation. Glyma15G154000 encodes a cullin, which is a component of SCF ubiquitin ligase complexes involved in mediating responses to auxin, and the gene is implicated in more than 10 biological processes, including auxin-activated signaling pathway, cell cycle, ethylene-activated signaling pathway, and response to auxin. In general, these four candidate genes related to leaf should be further studied to gain an in-depth understanding of their mechanisms under shading.

Table 4 Annotation description of five gene groups based on GO analysis.
Full size table

Candidate gene validation

qRT-PCR was performed on total RNA of leaves using gene-specific primers listed in Supplementary Table S3. According to the real-time PCR results, the expression profiles of each gene were shown (Fig. 4), and the expression level of Gmact11 in CK ‘Nandou 12’ (N) at V1 was 1. CCNA was present at the highest level in CK ‘Jiuyuehuang’ (J) and RIL64 (R64) at V0, while in CK N and RIL87 (R87) was lower than T2. CUL1 in N and J had a similar pattern to that of R64 and R87, and the expression of CK N and J had the lowest level at V0, so did the T1 and T2 R64 and R87 at V1 and V2. GH3 differential expression between CK and T2 was significant in J and R64 at V1, and the same as those between CK and T1 in N and R87.

Figure 4
figure 4

Comparison of the relative expression levels of CCNA (a), CUL1 (b), and GH3 (c) in ‘Nandou 12’ (N), ‘Jiuyuehuang’ (J), RIL64 (R64), and RIL87 (R87) under normal light (CK), one layer of shading net (T1), and two layers of shading nets (T2) at 7 days (V0), 15 days (V1), and 23 days (V2) after germination. ‘Nandou 12’, ‘Jiuyuehuang’, RIL64, and RIL87 (from left to right) in the photos were grown under CK, T1, and T2 (from bottom to top), whose fully developed unifoliate leaves, unrolling of first trifoliate leaves, and unrolling of second trifoliate leaves were sampled at 7 days (d), 15 days (e), and 23 days (f) after germination, respectively. The capital letters and the small letters on the top of SD bar denote significance level of P < 0.01 and 0.05 under different treatments, respectively.

Full size image

Supplementary Table S4 showed that the phenotypic values of LDW, SLW, TLA, and SLA of four materials in this section were similar to those in fields. N and R87 is more shade tolerant than J and R64. By analysis of variance, the differences of each material among CK, T1, and T2 treatments were significant (P < 0.05). The above results show that the correlation between three candidate genes and leaf-related traits can be preliminarily determined.

Discussion

In terms of excellent advantages, RILs were selected to construct the soybean genetic linkage map in this study. With innovations, such as new molecular markers, high-throughput sequencing has become the preferred strategy to obtain SNP markers for constructing a high-density linkage map for QTL mapping. Zhang et al.19 obtained 5,785 SNP markers by using 110 RILs (F8) derived from ‘Luheidou 2’ and ‘Nanhuizao’ and constructed a soybean genetic map with a total map distance of 2,255.18 cM and an average marker interval distance of 0.39 cM. Huang et al.20 constructed an asparagus bean map, which consisted of 5,225 SNP markers, spanning a total distance of 1,850.81 cM, with an average distance of 0.35 cM. In comparison with those above maps, the map in the present study contained more families in the population, obtained more SNP markers, and required a larger average marker interval distance.

The results of QTL identification revealed that qCLN17-3, qTLA12-41, qTLA15-99, qLDW12-40, qLDW15-98, and qLDWR5-19 were major QTLs. Three major QTLs have been detected in other studies: qTLA15-99 and qLDW15-98 (12,783,197–15,034,369 bp) were located in 9,396,584–34,757,105 bp on Chr158, and qLDWR5-19 (3,116,523–3,201,469 bp) is found in 209,284–3,735,646 bp on Chr0512. These QTLs are associated with the leaflet area.

Many QTLs related to 43 traits are located near the regions of the three QTL clusters. Chen et al.21 detected a QTL related to leaflet length (7,922,227–41,622,802 bp) was located near qCLN14-100 and qLDWR14-95 (7,440,291–8,071,592 bp) on Chr14, while Kim et al.15 identified two QTLs related to leaflet shape and width (13,755,345–17,021,739 bp) were located near qTLA15-99 and qLDW15-98 (12,793,197–15,034,369 bp) on Chr15. Reinprecht et al.22 and Keim et al.23 reported a QTL associated with seed weight (48,441,504–48,467,115 bp) and first flowering (48,305,238–48,441,504 bp), respectively. The two QTL regions were similar to qTLA6-192 and qLDW6-192, which were detected on Chr06 in 2016 with an interval of 48,180,859–49,430,702 bp. In another study, a QTL associated with seed weight is located in 5,540,205–16,352,945 bp24, and a QTL related to sudden death syndrome is found in 3,063,657–9,434,565 bp25. Both of them were found on Chr14. Interestingly, qCLN14-100 and qLDWR14-95, which respectively controlled the compound leaf number and the ratio of leaf dry weight in our study, were detected within the two regions. Moreover, qTLA15-99 and qLDW15-98 located on Chr15 were within the regions associated with shoot weight, nodule number26, pod maturity27, seed width28, sudden death syndrome25, and lodging29.

According to the results of gene expression, the correlation between leaf-related traits and three candidate genes can be found. In the case of normal light, CCNA expression level in shade-tolerance materials was higher. However, CUL1 and GH3 were highly expressed in all materials under shade, and these expression levels varied both with the shading degree and the shade-tolerance of materials. Glyma06G296500 (GH3) was associated with the auxin responsive GH3 gene family according to Phytozome v12.1 (https://phytozome.jgi.doe.gov/pz/portal.html). Wu et al.30 found there were some auxin response factors regulating tomato leaf shape development. In conclusion, the expression level of these three candidate genes in leaves was related to the shading condition and the shade-tolerance of materials, and the effect of CUL1 and GH3 on leaf-related traits may be contrary to that of CCNA.

Methods

Plant materials and experiment design

The RILs of 200 individuals were developed from a cross between the cultivar ‘Nandou 12’ (female) and the Sichuan local variety ‘Jiuyuehuang’ (male). ‘Nandou 12’ has lodging resistance and strong shade tolerance, which are significantly different from the phenotypic characters of ‘Jiuyuehuang’.

The field experiments of monoculture and relay intercropping were conducted in the Modern Agricultural Base of Renshou County, Sichuan Province, China (104°09′N, 30°00′E). In these experiments, 137 F7 RILs and 200 F8 RILs were grown together with two parents in a randomized block design with two replications on June 15, 2015 and 2016, respectively. In monoculture planting, each plot included 25 plants with 0.5 m between rows and 0.1 m between plants. In relay intercropping with a wide-narrow row cropping pattern (two maize rows with a 0.4 m inter-row distance and two soybean rows with a 0.5 m inter-row distance), adjacent rows of maize and soybean spaced 0.55 m, and 0.16 m and 0.1 m between plants of maize and soybean, respectively. The row length was 2.5 m. ‘Zhenghong 505’ (Zea mays L.) as the semi-compact maize was cultivated in narrow rows on April 1, 2015 and 2016, and harvested on August 5, 2015 and 2016. The soybean was cultivated in wide rows, and the other conditions were the same as those of the monoculture.

According to the field experiment phenotype data, two parents, RIL64 (weaker shade tolerance), and RIL87 (stronger shade tolerance) were analyzed to gene expression by real-time PCR. The above four accessions were sowed under normal light (CK, PPFD = 878.15 μmol/m2/s), one layer of black sun-shade net (T1, PPFD = 336.79 μmol/m2/s), and two layers of black sun-shade net (T2, PPFD = 251.54 μmol/m2/s) with three repeats.

Measurement of leaf-related traits

The soybean samples in field were collected 40 days after germination. Three normal samples with two repeats of each parent and RIL individual were selected to determine their leaf-related phenotype data. The following standards were used to evaluate the six leaf-related traits: CLN; TLA, which use the hole puncher get the round piece (diameter = 1.2 cm), then divide the hole area by the round piece weight and multiply by the total leaf dry weight; SLA, which is the ratio of the area of one side of a small round leaf piece to its dry weight; SLW, which is the reciprocal of SLA; LDW; and LDWR, which is the ratio of LDW to the sum of LDW and stem petiole dry weight.

‘Nandou 12’, ‘Jiuyuehuang’, RIL64, and RIL87 for candidate gene analysis were sampled to measure TLA, SLA, LDW, and SLW. The fully developed unifoliate leaves, unrolling first trifoliate leaves, and unrolling second trifoliate leaves were collected in 7 days (V0), 15 days (V1), and 23 days (V2) after germination under three treatments with repeated three times, respectively.

DNA extraction

The young leaves of 2 parents and 200 individuals of RILs were collected 20 days after germination. DNA was extracted as described by Doyle and Doyle31 with slight modifications. In brief, 3–5 g of fresh leaves were grinded and powdered in a mortar with liquid nitrogen and placed in a pre-cooled 2 mL Eppendorf tube (load 1/3), then mixed into 65 °C 600 μL of preheated CTAB and 20 μL of hydrophobic base ethanol, and incubated in a water bath at 65 °C for about 1 h. The preparation was gently shaken once every 10 min. Afterward, 400 μL of Tris-balance phenol and 400 μL of chloroform iso-amyl alcohol solution were added (24:1), fully mixed and gently shaken for 15 min. The samples were centrifuged for 10 min at 10,000 rpm. Then, 500 μL of the supernatant was transferred into a new Eppendorf tube, finished once suction filter with 400 μL of chloroform iso-amyl alcohol solution (24:1). The resulting mixture was centrifuged at 10,000 rpm for 10 min, and 400 μL of the supernatants was transferred into a new Eppendorf tube and added with 1 mL of pre-cooled isopropyl alcohol. The mixture was blended gently until the flocculent precipitated completely. The flocculent was gently selected from the mixture by using a syringe, transferred into a new Eppendorf tube, washed twice with 70% ethanol, then washed with anhydrous ethanol, blow dried in a clean bench, and dissolved in 200 μL of TE buffer. Finally, NanoVue was used to detect DNA quality, and agarose gel electrophoresis was conducted.

Genotyping

The DNA fragments of each parent were randomly sheared into ~500 bp by ultrasonication, and the sequencing library was constructed by end repair, 3′-plus A, plus sequencing connector, purification, and PCR amplification. The library was sequenced through the HiSeqTM 2500 system (Illunima, USA) sequencing platform after passing quality inspection. A high-density genetic linkage map for 200 RILs was constructed with a SLAF-seq technology. The SLAF marker was identified and genotyped in accordance with the procedures described by Sun et al.32. The original sequencing of SLAF-seq library was PE125bp. After the raw data were filtered, the reads in the Glycine max genome sequence were compared using the BWA software, and sequences mapped in the same position were defined as one SLAF locus. The alleles of each SLAF locus were then defined in accordance with parental reads with a sequence depth of >15X fold, and each genotype sequence contained 30% of the offspring information. Only SLAFs with two to four alleles were identified as polymorphic and considered potential markers. All polymorphism SLAF loci were genotyped with consistency in parental and offspring SNP loci. The marker codes of the polymorphic SLAFs were analyzed on the basis of the RILs. High-quality SLAF markers for the genetic mapping were filtered to construct the map.

Genetic map construction

Genetic map construction was based on 200 F7 RILs and was completed by using HighMap software by Beijing Biomarker Technologies (www.biomarker.com.cn). First, the recombination rate and the modified logarithm of odd (LOD) scores between markers were calculated to infer the linkage phases in chromosomes, through the maximum likelihood method for genetic map construction, according to the map order to rectify and sequence circularly, and a high quality genetic map was obtained.

QTL analysis and GO enrichment

QTL analysis was performed through the inclusive composite interval mapping (ICIM)33 of QTL IciMapping 4.1. An QTL was significant with the LOD score of 2.5. The nomenclature of McCouch et al.34 was used as a reference with slight modifications. The name of the QTLs in this study included four parts. For example, q, CLN, 17, and 3 of qCLN17-3 represented QTL, trait abbreviation, chromosome number, and QTL position in this chromosome, respectively. QTL mapping was drawn by MapDraw35. QTL validation was conducted with the software of QTL.gCIMapping.GUI in data of this study (https://cran.r-project.org/web/packages/QTL.gCIMapping.GUI/index.html)36.

The topGO software was used to conduct enrichment analysis of the related genes between the sample groups annotated into the GO database, and the hierarchical relationship of the nodes significantly enriched in the GO system was demonstrated visually in the form of directed acyclic graph.

RNA extraction, cDNA synthesis and real-time PCR

The leaves for RNA extraction were sampled with liquid nitrogen frozen in 7 days, 15 days, and 23 days after germination. RNA extraction was conducted according to the Plant RNA Kit (Omega Bio-tek, Inc., USA) instruction. The cDNA synthesis and real-time PCR were conducted according to the RT reagent Kit (Takara Bio, Inc., Japan) and SYBR Premix Ex Taq II (Takara Bio, Inc., Japan) instructions, respectively. Primers were designed by Takara Bio and synthesized by Sangon Biotech (Shanghai, China). Real-time PCR conducted with QuantStudioTM 6 Flex System and calculated the RQ value with 2(−∆∆CT) in Real-Time PCR Software v1.3 (Thermo Fisher, USA).

Data statistics

Microsoft Excel 2013 was used for data-calculated analysis and the gene expression profiles. IBM SPSS Statistics 22 was utilized to conduct the ANOVA and correlation analysis of the leaf-related traits. Origin Pro 8.0 was employed to draw the histogram.