Background

Root hairs are special tubular-shaped outgrowth from root epidermal cells, which vastly enlarge the root surface area and assist in water and nutrient absorption such as NO3−, Cl−, Ca2+, K+, Zn2+, and Mn2+, as well as the interactions with biotic and abiotic stresses [1, 2]. On the other hand, the length, density and morphology of root hair are influenced by various endogenous and environmental factors, including phytohormones and mineral nutrients, especially under phosphate (Pi) limiting condition [3,4,5,6,7]. The development of root hair can be separated into three basic stages: specification of the epidermal cell fate, initiation of the root hair outgrowth, and elongation of the hair via tip growth [2, 8].

As a unique single cell type in plant biology, the development, physiology, and cell biology of root hair have been intensively studied in Arabidopsis [2, 6, 9,27,28,29,30]. In maize, LD decay is rapid due to its extensive genetic diversity. Therefore, maize is recognized as an ideal model plant for conducting association studies [31,32,33,34]. To date, GWAS has successfully used to identify numerous candidate loci/genes controlling sever-al morphological or metabolic traits in maize, such as shoot apical meristem size, husk trait, plant height, kernel weight, drought tolerance, grain drying rate and grain moisture [35,36,37,38,39,40,41]. In this study, we used a maize association panel including 281 inbred lines with tropical, subtropical, and temperate backgrounds to interpret the phenotypic diversity and the genetic basis of root hair development. Several candidate genes putatively involved in root hair development were identified, providing a useful resource for further functional studies to elucidate molecular pathways involved in maize root hair growth and development.

Results

Phenotypic variation among root hair length

The association population in this study consists of a global collection of 281 diverse maize inbred lines [42, 43]. The root hair length of primary roots was measured from 3-day-old plants (Table S1). The noticeable variation in root hair length was represented by B73 and Mo17, two inbred lines commonly used in maize biology (Fig. 1a). The measured root hair trait followed a normal distribution with a slight right skew (Fig. 1b). The trait maximum, minimum, mean, standard deviation and coefficient of variation were listed in Table 1. The ratio of root hair length of inbred lines to B73 control ranged from 0.2 to 1.42 with the mean of 0.95, indicating that the root hair length exhibits broad variations in the association population.

Fig. 1
figure 1

Phenotypic variation of root hair length. a Diagram of B73 (left) and Mo17 (right) root hair length. Bars = 1 mm. b Frequency distributions of root hair length. Classes of trait values are shown on X axis and counts of inbred lines with the phenotypic values for these bins are shown on Y axis. c Boxplot of root hair length distribution in different subpopulations. Kruskal-Wallis test was applied to examine the difference of traits among subpopulations. Different letters indicate significance levels at P ≤ 0.05. No. of inbred lines included in each subpopulation are 117, 30, 22 and 112 for MIXED, NSS, SS and TST, respectively

Full size image
Table 1 Phenotypic statistics of root hair trait for the 281 inbred lines
Full size table

All 281 lines used in this study include three subpopulations and one mixed group, which are referred to TST, SS, NSS, and MIXED [42]. TST subpopulation is of tropical or subtropical origin, consisting of 112 lines. SS and NSS subpopulations are of temperate origin, consisting of 22 and 30 lines, respectively. MIXED subpopulation is inbred lines which were not accurately assigned into the above three subpopulations based on the phylogenic analysis [42, 43]. To investigate the effect of population structure on root hair phenotypes, the root hair length was compared between different subpopulations. Compared with SS, NSS and MIXED subpopulations, the mean value of TST subpopulation was significantly less, suggesting that maize inbred lines from tropical/subtropical origin tend to have shorter root hairs (Fig. 1c).

Associations of root hair phenotype with agronomic traits and metabolic pathways

As root hairs play a crucial role in the plant acquisition of nutrients and water, we postulated that root hair morphology is likely coordinated with other agronomic traits and amino acid metabolism. To verify our hypothesis, the Pearson-correlations were calculated after comparing root hair length with 17 agronomic traits and 18 amino acid contents in maize kernel, which were previously measured in the same association panel [44, 45]. The 17 agronomic traits include seven morphological traits, i.e. plant height (PH), ear height (EH), ear leaf width (ELW), ear leaf length (ELL), tassel maximum axis length (TMAL), tassel branch number (TBN), leaf number above ear (LNAE); seven yield-related traits, i.e. ear length (EL), ear diameter (ED), cob diameter (CD), kernel number per row (KNPR), cob grain weight (GW), cob weight (CW), kernel width (KW); three flowering-related traits, i.e. days to anthesis (DTA), days to silking (DTS) and days to heading (DTH) [44]. The amino acids measured in dry maize kernel include Ala, Arg, Asx, Glx, Gly, Lle, Leu, Lys, Met, Pro, Phe, Val, Tyr, His, Cys, Thr and Ser, and the total concentration of amino acids [45].

Of the 17 agronomic traits examined, 7 were correlated with root hair length (P ≤ 0.05), which were EH, ELW, TBN, LNAE, and all three flowering-related traits (Fig. 2a). Meanwhile, of the 18 amino acids, 3 were correlated with root hair length (P ≤ 0.05), which were Gly, Lys and Arg (Fig. 2b). The Pathway Association Study Tool (PAST) was further performed to elucidate the biochemical pathways likely contributing to root hair elongation [46,47,48]. Under the statistical threshold P < 0.01, a total of 8 and 1 metabolic pathways exhibited significant associations with increased or decreased root hair length, respectively (Table S2 and S3). Overall, the close correlations of root hair length with other agronomic or metabolic traits above-mentioned suggest that the root hair occurrence is correlated to other aspects of plant development and growth.

Fig. 2
figure 2

Correlation coefficients of root hair length with other agronomic and amino acidmetabolic traits. a Correlation coefficients between root hair length with 17 agronomic traits. b Correlation coefficients between root hair length with 18 metabolic traits. *Significant at P ≤ 0.05; **significant at P ≤ 0.01. The correlation level is color-coded according to the color key plotted in the middle

Full size image

GWAS on maize root hair length

Using 1,253,814 SNPs covering the whole maize genome with a minor allele frequency (MAF) ≥ 0.05, we performed GWAS to explore the genetic loci underlying the root hair length. Under the mixed linear model (MLM) [49, 50], which accounts for false positives arising from the population structure (Q matrix) and kinship relationship (K matrix) of the natural variation in the population, a stringent threshold of –log10P ≥ 5.2 was designated as the threshold for calling significantly associated SNPs. The Manhattan plots for the SNPs associated with root hair length were shown in Fig. 3. In total, 18 significant SNPs were detected, which are located on chromosomes 1, 2, 4, 5, 6, and 10, explaining the phenotypic variations ranged from 6.3 to 10.1%. Moreover, GWAS was also performed using the general linear model (GLM). Under the stringent threshold of –log10P ≥ 6.5, a total of 14 SNPs were identified, 12 out of which overlapped with MLM-derived SNPs (Fig. S1). As there were 8 MLM-derived SNPs on chromosome 6 present within the same LD region (r2 ≥ 0.2), and the leading SNP_ 75630277 was selected as the representative of this locus (Table 2), yielding a total of 11 SNPs used for the further analysis.

Fig. 3
figure 3

GWAS showing significant P-values associated with root hair length based on the mixed linear model (MLM). a Manhattan plots of MLM for root hair length. The horizontal dashed line represents the significance threshold -log10(P) = 5.2. b QQ plot of MLM showing the ratio of the observed and the expected P-values for root hair length. The solid diagonal lines represent agreement between observed and expected probability distributions assuming null SNP–trait association

Full size image
Table 2 SNPs, chromosomal position and candidate genes significantly associated with root hair trait identified by GWAS using MLM
Full size table

Genes co-localized with significant SNPs

Protein-coding genes harboring or nearest to the significant SNPs were nominated as the candidates associated with root hair (Table 2). The allelic effect of haplotype coordinated with significant SNPs on root hair phenotypes was assessed (Fig. 4 and Figure S2). The most significant SNP (chr6.S_ 34221678, P-value = 1.70E-07, R2 = 9.0%) locates in the fourth exon of GRMZM2G403003 (Fig. 4a), encoding a TON1 recruiting motif (TRM) protein. The average root hair length for A allele was substantially shorter than T allele (P ≤ 0.01, Fig. 4d). The second significant SNP (chr5.S_198197360, P-value = 1.80E-07, R2 = 9.8%) locates in the third intron of GRMZM2G044851 (Fig. 4b), encoding the nitrate transporter 1.5. The average root hair length for allele with A was greatly longer than allele with C (P ≤ 0.01, Fig. 4e). The third significant SNP (chr6.S_ 75630277, P-value = 3.97E-07, R2 = 9.7%), which is the one with other 7 SNPs within the same LD region, locates at the promoter region of GRMZM5G825276 (Fig. 4c), encoding a GDSL-like lipase. The average root hair length for allele with A was greatly longer than allele with G (P ≤ 0.01, Fig. 4f). It is noted that the haplotype analysis for the other eight genes was depicted in Fig. S2.

Fig. 4
figure 4

The allele effects of significant SNPs located around representative genes for root hair length. a–c Regional plots showing association map** results for SNPs located around GRMZM2G403003 (a), GRMZM2G044851 (b), GRMZM5G0825276 (c). d–f Allele effects of the most significant SNPs for root hair traits. *Significant at P ≤ 0.05; **significant at P ≤ 0.01. (d) GRMZM2G403003, (e) GRMZM2G044851, (f) GRMZM2G825276. Each dot represents an SNP. The horizontal dashed black line represents the significant threshold –log10(P) = 5.2

Full size image

Discussion

Root hairs enlarge the root surface area and thereby play vital roles in plant absorbing water and nutrient, as well as co** with biotic and abiotic stress. In recent years, extensive mutation-based studies have been carried out to dissect the genetic regulatory network determining root hair development, but we lack fundamental knowledge about whether and how root hair is controlled in a plant natural variation population. In this work, we interpreted the natural variation and the associated genetic architecture of root hair length in a maize association panel, and a set of putative candidate genes controlling root hair development were revealed by performing GWAS analysis.

Phenotypic variability and coordination of root hair length with other processes of plant development and growth

Maize was domesticated from its wild ancestor, teosinte (Zea mays ssp. parviglumis) about 8000–9000 years ago. Nowadays, modern maize displays a large geographic distribution from tropical to temperate climate. Therefore, population structure is associated with significant differences in maize morphology [51]. In this study, we observed that root hair exhibits wide length variation with normal distribution. Interestingly, tropical lines displayed shorter root hairs relative to temperate lines, implying that the adaption of maize from tropical to temperate regions was probably accompanied by the character of lessening root hair growth.

In our investigation, the root hair at the length dimension was negatively rather than positively correlated with several morphological and metabolic traits. This is somewhat counter-intuitive. However, it is known that the maize plant is a monocot and has a fibrous root system consisting of massive adventitious and lateral roots, the anatomy of which is markedly different from dicotyledonous model Arabidopsis [52,53,54]. In this context, it is possible that the other aspects of root system may enable maize to have adequate larger surface area to support plant growth under normal soil condition. Meanwhile, it is worthy to be mention that to facilitate high throughput phenoty**, the root hair from the emerging primary roots after 3-days post germination was assessed in this study, thereby whether the growth rate will alter distinctly in different inbred lines is uncertain at this point.

Putative genes involved in root hair morphogenesis

We identified a total of 11 candidate genes associated with root hair elongation, and the homologs of three genes have been reported to influence root hair development in Arabidopsis. GRMZM2G100288 encodes a receptor-like protein kinase, and its closest homologs in Arabidopsis is FERONIA (FER), which is well known as a key hub of cell signaling networks mediating various hormone, stress, and immune responses [55,56,57,58,59,60,61,62,63,http://maizego.org/. The maize lines in this association panel were clustered into three subpopulations, including 22 SS lines, 30 NSS lines and 112 TST lines, and the remaining 117 lines fall into a mixed subpopulation. Detailed information about the origins of these lines was described in previous studies [42, 43]. The 1.25 million high-quality SNPs (MAF > 5%) and the estimated population structure and kinship were assessed using the method previously described [99, 100]. The 165,248 markers in approximate linkage equilibrium with each other were found by PLINK (window size 50, step size 50, r2 ≥ 0.2) [101]. Then, we used the uniform Bonferroni-corrected thresholds at α = 1 for MLM and α = 0.05 for GLM as the significance cutoffs as reported in the previous studies [37, 44, 102]. Finally, the suggestive P value was computed by 1/n and 0.05/n (n = 165,248), and we obtained the suggestive threshold 6.05 × 10− 6 for MLM and 3.03 × 10− 7 for GLM, respectively.

To estimate the phenotypic variance explained by each significant SNP, we used ANOVA to construct linear models of Y = αX + βP + ε (1) and Y = βP + ε (2). In this model, Y is the phenotype, X is the SNP genotype, P is the matrix of three subpopulations (NSS, SS and TST), α is the SNP effect, β is the subpopulation effects, and ε is random effects. Thus, the R2 of each significant SNP after adjusting for the population structure effects were reported as previously described [37].

Pathway analysis

The pathway analysis was performed in https://maizegdb.org/past. The resulting SNP-trait association data and effects data generated by TASSEL were implemented in the pathway analysis. During the process of loading data, the LD data is filtered to drop rows where the loci are not the same, and then unneeded columns from the TASSEL output are dropped [103]. Only pathways with five or more mapped genes were considered in the analysis. Significance of the enrichment score was determined by permutation analysis (1000 random permutations of the effect values).

Prediction of candidate genes

To search the candidate genes underlying associated SNPs, we selected the most significant/leading SNPs within the same LD block (R2 < 0.2) to represent the locus, and candidate genes were nominated by the leading SNP positioned. The physical locations of the SNPs were recorded according to the B73 RefGen_v2 (www.maizesequence.org). The corresponding genes were annotated based on the literatures describing the function of their homologs in other species or the information retrieved from conserved domain database (CDD).