Two putatively novel loci associated with spike length and their breeding potential in Chinese endemic wheat

Abstract

Spike length (SL) is a key determinant of wheat spike architecture and yield potential. In this study, SL was evaluated across multiple environments in a panel of 182 Chinese endemic wheat accessions. Substantial phenotypic variation and high broad-sense heritability were observed, indicating strong genetic control. Genome-wide association study (GWAS) was performed using 38,490 high-quality SNPs under a mixed linear model incorporating population structure and kinship (Q + K). Four loci significantly associated with SL were identified on chromosomes 2B, 3B, 4A, and 7A, explaining 2.53–8.06% of the phenotypic variation. Two loci on chromosomes 4A and 7A co-localized with previously reported SL QTL within the linkage disequilibrium decay distance, confirming their genetic stability. In contrast, loci on chromosomes 2B (QSL.sau.2B) and 3B (QSL.sau.3B) reside in genomic regions not previously associated with SL, suggesting potentially novel genetic factors. Allelic effect analysis based on BLUP values revealed pleiotropic effects of SL-associated loci on spike density, spikelet number, plant height, and anthesis date. Notably, QSL.sau.3B was consistently associated with SL across environments and further supported through effect analysis in an additional wheat population of 220 accessions, confirming its stability and breeding relevance. These results advance understanding of the genetic architecture of SL and provide valuable resources for wheat improvement.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13201-z.

Introduction

Bread wheat (Triticum aestivum L.) is a cornerstone of global food security, providing more than 20% of the world’s dietary calories and protein [1]. With the global population projected to exceed nine billion by 2050 [2], agricultural production must increase by at least 70% to meet future food demands [3]. Achieving this goal is increasingly challenging due to diminishing arable land, water scarcity, and the adverse effects of climate change on crop productivity. Therefore, enhancing the genetic yield potential of wheat remains a critical objective for sustainable food and nutritional security.

Grain yield in wheat is a complex polygenic trait determined by the coordinated effects of multiple yield components [4]. Among these, spike architecture plays a pivotal role. Spike length (SL) is a key morphological trait that directly influences the number of spikelets per spike and, consequently, grain number, thereby determining sink capacity and yield potential [5, 6]. As a typical quantitative trait, SL is controlled by multiple loci with small effects, and its phenotypic expression is often influenced by genotype × environment (G × E) interactions [7].

Numerous quantitative trait loci (QTL) for SL have been identified through biparental mapping and genome-wide association studies (GWAS) [7–9]. Additional SL-associated QTL have been mapped across nearly all 21 wheat chromosomes [10–13]. However, only a few QTL are consistently detected across diverse genetic backgrounds, mainly on chromosomes 1 A, 1B, 2D, 4B, 5 A, 7 A, and 7B.

Despite extensive QTL mapping, relatively few genes underlying SL have been functionally characterized, largely due to the complexity of the wheat genome. In wheat, WAPO1, Ppd-D1, TB1, Rht8, Rht24b, TaMOC1, and WFZP have been shown to influence SL and/or spike morphology [3, 10, 12, 14–19]. Evidence from rice and barley further supports the involvement of conserved regulatory pathways in inflorescence architecture [3, 20–22]. Collectively, these studies highlight the need to identify stable SL-associated loci and link them with causal genes to advance wheat improvement.

Chinese endemic wheat landraces and historical cultivars represent a valuable yet underexplored source of genetic diversity. Shaped by long-term adaptation to diverse agroecological conditions and traditional farming systems, these accessions harbor unique alleles for agronomic traits, including yield components and stress tolerance [23]. Despite their importance, the genetic architecture of SL in these germplasm resources remains poorly characterized because previous studies have primarily focused on modern elite wheat cultivars, whereas Chinese landraces and endemic wheat resources have been comparatively underutilized in GWAS analyses.

GWAS provides a powerful complementary approach to linkage mapping by exploiting historical recombination in diverse populations to achieve higher mapping resolution. GWAS enables the efficient identification of marker–trait associations and candidate genes underlying complex traits, thereby facilitating the discovery and utilization of favorable alleles in modern breeding programs [8].

In the present work, a GWAS was performed to dissect the genetic basis of SL in a diverse collection of Chinese endemic wheat accessions using high-density SNP genotyping. This study aimed to uncover loci controlling natural variation in SL, assess their correspondence with previously reported genomic regions, and identify putative candidate genes through integrative in silico analyses. The findings reveal both conserved and previously unreported genomic regions influencing SL and provide a valuable foundation for the genetic improvement of wheat spike architecture and yield potential.

Materials and methods

Plant materials

A discovery panel consisting of 182 Chinese endemic wheat accessions was assembled for this study, comprising 109 Tibetan semi-wild wheat, 61 hulled wheat accessions from Yunnan, and 12 Xinjiang rice wheat accessions. These materials were sourced from the Triticeae Germplasm Resources Bank of Sichuan Agricultural University and the Chinese Crop Germplasm Resources Bank, as described in previous studies [23, 24]. The same association panel has been utilized in earlier study investigating other agronomic traits [24]. In the present study, it was employed to investigate the genetic architecture of SL. In addition, an independent validation panel of 220 Sichuan wheat cultivars and landraces was employed to assess the robustness of the identified loci [25, 26].

Phenotypic assessment of spike length

SL phenotypic data used in this study were obtained from previous evaluations of the same Chinese endemic wheat association panel conducted at Chongzhou (CZ; 30°33′37.3″ N, 103°38′45.4″ E) and Wenjiang (WJ; 30°43′ N, 103°51′ E), Sichuan, China, across three wheat growing seasons (2022, 2023, and 2025), hereafter denoted as 2022CZ, 2023CZ, 2025CZ, and 2025WJ [24]. Each accession was sown in a single 2-m row with 0.1-m plant spacing and a 0.3-m inter-row distance, following standard experimental protocols [27]. Basal fertilization consisted of 80 kg ha⁻¹ superphosphate and 100 kg ha⁻¹ nitrogen applied at sowing. Standard local agronomic practices were maintained throughout the growing period to ensure optimal development [28, 29].

The field trials were conducted using a single-row plot design with one plot per accession in each environment. For phenotypic evaluation, eight representative plants within each row were randomly sampled at physiological maturity, and their mean value was used as the phenotypic estimate for each accession. These measurements represented within-line subsampling rather than independent biological replicates. SL was measured from the base of the spike to the tip, excluding awns. Mean SL values for each accession were determined by averaging measurements across the sampled plants.

Phenotypic data collection and analysis

Phenotypic data for SL across four environments (2022CZ, 2023CZ, 2025CZ, and 2025WJ) were obtained from previous evaluations of the same Chinese endemic wheat association panel [24]. In addition to SL, phenotypic data for spikelet number per spike (SNS), plant height (PH), anthesis date (AD), and effective tiller number (ETN) were obtained from the same study and used for trait comparison and correlation analyses [24]. Although SL was included in the previous study for phenotypic characterization and correlation analyses, its genetic architecture was not investigated. Therefore, the present study integrated these phenotypic data with genome-wide marker information to perform GWAS, identify associated loci, and predict candidate genes underlying SL variation. Spike density (SD) was calculated by dividing SNS by SL [1]. Phenotypic data for the 220 accessions in the independent validation population were obtained from a previously published study [26].

Best linear unbiased prediction (BLUP) values were estimated using a linear mixed model implemented in SAS 9.1.3. The model was specified as:

Where yij is the observed phenotypic value of genotype i in environment j, μ is the overall mean,Gi is the random effect of genotype, Ej is the random effect of environment, (GE)ij is the genotype-by-environment interaction effect, and ɛ ij is the residual error. Genotype, environment, and genotype × environment interaction effects were treated as random effects. BLUP values for each genotype were estimated using combined multi-environment phenotypic data.

Broad-sense heritability (H²) was estimated based on the formula , where represents genetic variance, is the genotype-by-environment interaction variance, is the residual error variance, and e is the number of environments [30].

For each individual environment as well as for the BLUP dataset, the minimum and maximum values of SL were recorded, and descriptive statistics including the mean and standard deviation were calculated. The phenotypic distribution of SL was visualized using boxplots generated in GraphPad Prism version 9.5.1 (GraphPad Software, San Diego, CA, USA), and all statistical analyses were conducted using R software (R Core Team, 2023).

Genotyping analysis

Genomic DNA was isolated from leaves of two-week-old seedlings using a modified cetyltrimethylammonium bromide (CTAB) protocol [31]. Genotypic information for the discovery panel was sourced from a previously reported wheat 55 K SNP array dataset generated for Chinese endemic wheat accessions [23]. Quality control filtering excluded markers with more than 5% missing genotypes or a minor allele frequency below 5%, resulting in a final dataset comprising 182 accessions used for downstream analyses [24]. Genotyping of the independent validation panel is documented in previous study [32].

Population structure, kinship and LD analysis

Population structure was inferred using STRUCTURE v2.3.4 [33] under an admixture model with K values from 1 to 10, based on five independent runs with 100,000 burn-in and 100,000 MCMC iterations [23]; the optimal K was determined using the ΔK method in STRUCTURE SELECTOR [34]. Kinship was estimated in TASSEL v5.2.38 using the identity-by-state approach [24]. Linkage disequilibrium (LD) was calculated as r² between intra-chromosomal SNPs, and markers within the LD half-decay distance were considered to define a single QTL region, enabling the identification of high-confidence loci while accounting for population structure and relatedness [23, 24].

Genome-wide association analysis

Genome-wide association mapping for SL was conducted using 182 wheat accessions. The analysis was performed in R using the rMVP package [35] under a mixed linear model [36] that simultaneously accounted for population structure (Q) and kinship (K) to reduce spurious associations. SNP markers retained after quality-control filtering and possessing defined physical positions in the reference genome were used for genome-wide association analysis.

Marker–trait associations were declared significant at a threshold of − log₁₀(P) ≥ 3. To ensure robustness, only loci repeatedly detected in at least two independent environments or in BLUP-based analyses were considered reliable SL-associated signals. These high-confidence loci were subsequently used for comparative analysis and candidate gene identification. Identified QTL were named according to standard nomenclature: Q + trait name + research institution + chromosome number + QTL ordinal number. Here, "SL" denotes spike length and "sau" Sichuan Agricultural University.

Candidate gene prediction

Candidate genes underlying SL–associated QTL were identified by retrieving annotated genes located within ± 1 Mb of the lead SNP for each locus using the interval query function of the WheatOmics platform [37]. The size of this window was determined according to the linkage disequilibrium decay observed in the association panel, ensuring that genes in strong linkage with the associated markers were captured. Only protein-coding genes with functional annotations were retained for further analysis.

To refine candidate selection, gene expression patterns within each QTL interval were evaluated using publicly available RNA-seq data from the WheatOmics expression atlas. Particular emphasis was placed on genes showing relatively high or preferential expression in spike tissues, given the direct relevance of spike development to SL determination. Genes combining close physical proximity to lead SNPs, spike-enriched expression profiles, and putative biological functions related to growth, development, or cellular regulation were prioritized as candidate genes. While functional validation was beyond the scope of this study, this integrative in silico approach provides a focused set of candidates for future genetic and molecular analyses of SL regulation in wheat.

Results

Phenotypic variation, heritability and correlation

SL exhibited substantial phenotypic variation across environments (Table 1). SL showed normal distribution among different environments and followed almost similar distribution (Fig. 1A). In the 2022CZ environment, SL ranged from 8.61 to 15.93 cm, with a mean of 12.54 cm and a relatively low coefficient of variation (CV = 12.58%). The broad-sense heritability was high (H² = 0.958), indicating that genetic factors predominantly controlled SL expression in this environment.

Table 1.

Phenotypic assessment of 182 accessions of Chinese endemic wheat for spike length at multiple environments and BLUP datasets

Environment	Min	Max	Mean	STDEV	CV	H2
2022CZ	8.61	15.93	12.54	1.58	12.58%	0.958
2023CZ	7.29	17.55	12.41	2.10	16.91%
2025CZ	7.7	20.1	13.84	2.32	16.78%
2025WJ	7.65	15.11	11.37	1.44	12.64%
BLUP	7.75	17.13	12.54	1.70	13.55%

2022CZ Chongzhou 2022, 2023CZ Chongzhou 2023, 2025CZ Chongzhou 2025, 2025WJ Wenjiang 2025, BLUP The Best Linear Unbiased Prediction, STDEV Standard Deviation, CV Coefficient of Variation, H² Broad Sense Heritability

Fig. 1.

A Density plot showing phenotypic distribution of spike length in 182 accessions of Chinese endemic wheat in 4 different environments (2022CZ, 2023CZ, 2025CZ, and 2025WJ) and BLUP. B Correlation matrix for spike length in 4 different environments (2022CZ, 2023CZ, 2025CZ, and 2025WJ) and BLUP

In 2023CZ, SL showed a wider range (7.29–17.55 cm) and greater phenotypic dispersion, as reflected by a higher CV (16.91%). Similarly, the 2025CZ environment exhibited the largest variation among all trials, with SL ranging from 7.70 to 20.10 cm, a mean of 13.84 cm, and a CV of 16.78%, suggesting stronger environmental effects or enhanced genotype × environment interactions.

In contrast, the 2025WJ environment showed comparatively lower variability (CV = 12.64%), with SL ranging from 7.65 to 15.11 cm and a mean of 11.37 cm. When BLUP values were estimated across environments, SL ranged from 7.75 to 17.13 cm, with a mean of 12.54 cm and moderate variability (CV = 13.55%), indicating stable genetic performance after minimizing environmental noise.

Correlation analysis revealed that SL was significantly correlated among environments (Fig. 1B) and was also associated with several agronomic traits based on BLUP-derived phenotypic values (Table 2). SL was positively correlated with SNS (r = 0.186, P < 0.05), indicating that longer spikes tended to carry more spikelets. In contrast, SL exhibited a strong and highly significant negative correlation with SD (r = − 0.893, P < 0.01), suggesting that increased SL was associated with reduced spike compactness.

Table 2.

Correlation coefficients among spike length and related agronomic traits based on BLUP values across environments

Trait	SL	SNS	SD	ETN	PH
SNS	0.186*
SD	−0.893**	0.218**
ETN	−0.064	0.056	0.052
PH	0.334**	0.033	−0.319**	0.143
AD	0.256**	0.298**	−0.112	0.214**	0.226**

SL Spike length, SNS Spikelet number per spike, SD Spike density, ETN Effective tiller number, PH Plant height, AD Anthesis date

^*, and ** Correlation is significant at the 0.05, and 0.01 level respectively (2-tailed)

SL also showed a significant positive correlation with PH (r = 0.334, P < 0.01), implying partial co-regulation or genetic linkage between spike elongation and overall plant stature. Additionally, SL was positively correlated with AD (r = 0.256, P < 0.01), indicating that genotypes with longer spikes tended to flower slightly later. No significant correlation was detected between SL and ETN suggesting that spike elongation is largely independent of tillering capacity.

Genome-wide association study

GWAS was conducted for SL using 38,490 high-quality SNPs and an MLM incorporating Q and K as covariates (Fig. 2). Population structure, kinship relationships, and LD decay had been comprehensively evaluated for this population in previous studies using the same genotypic dataset, providing a robust foundation for the GWAS framework applied in the present study [23, 24]. Previous analyses of this association panel reported a genome-wide LD half-decay distance of approximately 1.87 Mb [23, 24], which was used in the present study to define QTL intervals and assess co-localization with previously reported loci. Significant marker-trait associations were identified using a threshold of − log₁₀(P) ≥ 3.

Fig. 2.

Manhattan and quantile–quantile (Q-Q) plots showing the results of genome wide association studies (GWAS) for spike length using Q + K, MLM method in 2022CZ, 2023CZ, 2025CZ, 2025WJ, and BLUP datasets. The positions of SNP on chromosomes are indicated on the x-axis relative to their -log₁₀(P) values on the y-axis. The red dotted line refers to the threshold of -log10(P) = 3.00

Four loci, significantly associated with SL were identified on chromosomes 2B, 3B, 4 A, and 7 A (Table 3). The detected loci showed LOD values ranging from 3.07 to 4.22 and explained 2.53% to 8.06% of the phenotypic variation, indicating moderate but meaningful genetic effects.

Table 3.

List of identified QTL for spike length at multiple environments and BLUP datasets

QTL	SNP	Chromosome	Position	Allele	LOD	Marker R2	Environment
QSL.sau.2B	AX-109980437	2B	46,571,250	T/C	3.22–3.55	5.59–6.17	2022CZ, 2025CZ, BLUP
QSL.sau.4A	AX-111615703	4A	636,102,135	G/T	3.13–4.22	5.62–8.06	2023CZ, 2025CZ, BLUP
QSL.sau.7A	AX-109525924	7A	57,978,272	T/C	3.35–3.52	4.54–4.74	2023CZ, 2025CZ, BLUP
QSL.sau.3B	AX-111079520	3B	24,541,870	C/A	3.07–3.51	2.53–3.09	2023CZ, 2025CZ, BLUP

2022CZ Chongzhou 2022, 2023CZ Chongzhou 2023, 2025CZ Chongzhou 2025, 2025WJ Wenjiang 2025, BLUP The Best Linear Unbiased Prediction

Among these loci, QSL.sau.2B, represented by SNP AX-109980437 at 46.57 Mb on chromosome 2B, was consistently detected in the 2022CZ and 2025CZ environments as well as in BLUP values. This locus exhibited LOD scores ranging from 3.22 to 3.55 and explained 5.59% to 6.17% of the phenotypic variation, indicating a stable genetic effect across environments. The T allele at this locus was associated with increased SL.

Similarly, QSL.sau.4A, represented by SNP AX-111615703 located at 636.10 Mb on chromosome 4 A, was detected in the 2023CZ, 2025CZ, and BLUP analyses. This locus showed LOD values ranging from 3.13 to 4.22 and explained 5.62% to 8.06% of the phenotypic variation, further supporting its environmental stability and relatively stronger effect on SL.

In addition, QSL.sau.7A (AX-109525924) on chromosome 7 A and QSL.sau.3B (AX-111079520) on chromosome 3B were identified as SL-associated loci in individual environments. These loci exhibited LOD values ranging from 3.35 to 3.52 and 3.07 to 3.51, respectively, and explained 4.54% to 4.74% (QSL.sau.7A) and 2.53% to 3.09% (QSL.sau.3B) of the phenotypic variation, suggesting environment-specific or minor-effect QTL contributing to SL variation.

Comparative analysis with previously reported loci indicated that two of the identified SL-associated loci overlapped with known SL QTL. Specifically, QSL.sau.4A, detected at approximately 636 Mb on chromosome 4 A, was located close to a previously reported SL QTL at ~ 639 Mb and was therefore considered to be co-localized with the known locus [38]. Similarly, QSL.sau.7A, identified at approximately 57.98 Mb on chromosome 7 A, was located near previously reported SL loci in the 62–67 Mb region and was regarded as overlapping with known SL–associated regions [38].

In contrast, QSL.sau.2B and QSL.sau.3B were detected in genomic regions where no SL QTL have been reported in previous studies and may therefore represent potentially novel loci controlling SL, to the best of our knowledge.

Overall, the identification of both co-localized and potentially novel loci highlights the complex genetic architecture of SL in wheat and demonstrates the effectiveness of the MLM (Q + K) approach in detecting robust and biologically meaningful associations.

Allelic effects of identified SL–associated Loci

Allelic effect analysis based on BLUP-derived trait values revealed distinct pleiotropic effects of the identified SL-associated loci (Fig. 3A-3D). The T allele of AX-109980437, representing QSL.sau.2B, significantly increased SL, ETN, PH, and AD, while significantly reducing SD. Similarly, the T allele of AX-111615703, corresponding to QSL.sau.4A, was associated with significant increases in SL, SNS, and PH, accompanied by a significant decrease in SD.

Fig. 3.

Boxplot showing the effects of QSL.sau.2B (AX-109980437, A, QSL.sau.4A (AX-111615703, B QSL.sau.7A (AX-109525924, C and QSL.sau.3B (AX-111079520, D by comparing two allelic groups. SL, spike length; SNS, spikelet number per spike; SD, spike density; ETN, effective tiller number; PH, plant height; AD, anthesis date. *, ** and ***, significances at the 0.05, 0.01, and 0.001 probability levels, respectively

For QSL.sau.7A, represented by AX-109525924, the T allele significantly increased SL, ETN, PH, and AD, while significantly reducing SD. In contrast, the C allele of AX-111079520, representing QSL.sau.3B, significantly increased SL, SNS, and PH, and was also associated with a significant reduction in SD.

To evaluate the combined contribution of favorable alleles across the identified SL-associated loci, we grouped the 182 wheat accessions according to the number of favorable alleles carried (0–4). SL increased progressively with the number of favorable alleles (Figure S1A). Linear regression analysis revealed that each additional favorable allele increased SL by 0.8 units (95% CI: 0.577–1.023; R² = 0.218; p = 3.19 × 10⁻¹¹) (Figure S1B). No significant non-additive interactions were detected (F = 1.00, p = 0.3943), indicating that the identified QTL act primarily in an additive manner. Summary statistics for each allele group are provided in Table S1.

Collectively, these results indicate that favorable alleles at SL-associated loci not only promote spike elongation but also influence multiple spike architecture– and development-related traits, suggesting pleiotropic genetic effects underlying SL variation in wheat. These findings also highlight the potential utility of pyramiding favorable alleles in breeding programs to enhance SL.

Independent assessment of the potentially novel QTL QSL.sau.3B

To further assess the reproducibility of the potentially novel SL locus identified in this study, the allelic effects of QSL.sau.3B were examined using publicly available phenotypic and genotypic data from an independent wheat panel reported previously [26]. The allelic effects of the lead SNP AX-111079520 were examined across datasets to assess consistency.

The results showed that the favorable allele at AX-111079520 was consistently associated with increased SL across datasets, supporting the stability of QSL.sau.3B (Fig. 4). Although the proportion of phenotypic variance explained by this locus was relatively small, the reproducibility of its allelic effect across independent phenotypic evaluations indicates that QSL.sau.3B represents a reliable genetic factor contributing to SL variation.

Fig. 4.

Boxplot showing the effects of QSL.sau.3B (AX-111079520) by comparing two allelic groups in an independent natural population. SL, spike length. ***, significances at the 0.001 probability levels

Prediction of candidate genes from two potentially novel QTL

Candidate gene analysis for SL was conducted using the IWGSC RefSeq v2.1 annotation. QTL flanking regions were defined using the LD half-decay distance, and a ± 1 Mb window was applied to retain biologically relevant candidates. A total of 140 annotated genes were identified within the intervals of QSL.sau.2B and QSL.sau.3B (Table S2). Gene prioritization was further refined using transcriptome expression data (TPM values) across multiple wheat tissues obtained from a wheat expression database (Table S3). No genes within the QSL.sau.2B interval exhibited relatively high or preferential expression in spike tissues and therefore no candidate gene was prioritized for this locus. In contrast, one gene within the QSL.sau.3B interval, TraesCS3B02G046700, showed relatively higher expression in spike tissues (Table S3) and was therefore retained as a putative candidate gene for further analysis.

Discussion

SL is a key component of wheat spike architecture and contributes to yield potential through its influence on spikelet number, spike compactness, and reproductive development [39]. By integrating multi-environment phenotyping, BLUP-based trait estimation, and GWAS using a robust MLM (Q + K) framework, this study dissected the genetic basis of natural variation in SL while effectively controlling population structure and kinship effects.

The consistently high broad-sense heritability of SL across environments indicates that this trait is predominantly under genetic control, despite the presence of strong genotype × environment interactions. The stability of BLUP-derived SL values further supports the reliability of the phenotypic dataset used for association mapping. Correlation analyses provided important biological context for the identified loci. The strong negative association between SL and SD reflects a well-documented developmental trade-off between spike elongation and compactness [39, 40], whereas positive correlations with spikelet number and PH suggest partial co-regulation of spike architecture and overall plant growth. In addition, the positive correlation between SL and AD implies that spike elongation may be linked to developmental timing, a relationship that warrants consideration in breeding strategies aiming to balance spike architecture and phenology.

GWAS identified four loci associated with SL, each explaining a modest proportion of phenotypic variance. This genetic pattern is consistent with previous reports indicating that SL is a typical quantitative trait governed by multiple loci with small to moderate effects rather than by major-effect genes. The detection of both environmentally stable and environment-specific loci highlights the complex genetic architecture underlying SL and reflects the sensitivity of spike development to environmental conditions.

Comparative genomic analysis showed that two SL–associated loci co-localized with previously reported QTL within the LD decay distance (~ 1.87 Mb) of this population. Specifically, QSL.sau.4A was located close to a reported SL locus on chromosome 4 A, and QSL.sau.7A mapped near known loci on chromosome 7 A [38]. The re-detection of these genomic regions across independent studies supports their genetic stability and reinforces the robustness of the GWAS results. In contrast, QSL.sau.2B and QSL.sau.3B were identified in genomic regions where SL QTL have not been reported previously, suggesting that they may represent uncharacterized loci.

Among these, QSL.sau.3B was further assessed using previously published genotypic and phenotypic datasets from an additional wheat population. The consistent direction of allelic effects across datasets supports the reproducibility of this locus, despite its relatively small effect size. Minor-effect loci such as QSL.sau.3B are expected for complex traits like SL and may collectively contribute substantially to phenotypic variation. Importantly, loci with small but reproducible effects across datasets are particularly valuable for genomic selection and for pyramiding favorable alleles to achieve cumulative genetic gains [41].

Allelic effect analysis revealed that SL–associated loci influenced multiple agronomic traits, suggesting pleiotropic effects or tight genetic linkage. The coordinated effects on SNS, SD, PH, and AD indicate that SL is embedded within a broader developmental network regulating spike architecture and plant growth. From a breeding perspective, such pleiotropy highlights the importance of balanced selection, as alleles enhancing SL may also influence flowering time or plant stature.

The consistent reduction in SD associated with favorable SL alleles provides genetic evidence for the trade-off between spike elongation and compactness. This insight is particularly relevant for ideotype breeding, where optimal spike architecture rather than maximal SL alone is often the breeding objective. Nevertheless, SL alone may not directly translate into higher grain yield because wheat productivity is determined by complex interactions among multiple yield components and plant architectural traits. The additive effect observed suggests that pyramiding favorable alleles across the identified QTL could progressively enhance SL, providing a clear strategy for marker-assisted selection.

Within the ± 1 Mb interval of QSL.sau.3B, TraesCS3B02G046700, encoding a putative leucine-rich repeat receptor-like kinase (LRR-RLK) containing PF08263 (Leucine-rich repeat N-terminal domain) and PF00069 (Protein kinase domain), exhibited relatively higher expression in spike tissues and was therefore considered a putative candidate gene. LRR-RLKs are important regulators of plant growth and developmental signaling and have been implicated in the regulation of inflorescence architecture and meristem activity in wheat and other cereals [42]. The extracellular leucine-rich repeat domain is involved in ligand perception, whereas the intracellular kinase domain mediates downstream signal transduction pathways controlling cell division and tissue development [43]. Therefore, variation in the expression or function of TraesCS3B02G046700 may influence spike elongation through modulation of spike meristem development and rachis growth.

Although the predicted candidate gene is supported by their physical positions within SL–associated QTL intervals and by their spike-preferential expression patterns, direct functional evidence linking it to SL regulation is currently lacking. Therefore, it should be regarded as putative candidate genes, pending further validation through fine mapping, expression analysis across developmental stages, and functional characterization.

Nevertheless, the present study did not include experimental molecular validation or functional characterization of the identified loci. Therefore, further validation using marker-assisted assays, fine mapping, functional analyses, and independent genetic populations will be necessary to confirm the causal relationships underlying SL variation.

In summary, this study provides a comprehensive dissection of the genetic basis of SL in wheat through the integration of multi-environment phenotyping, GWAS, allelic effect analysis, and validation of a potentially novel locus. The identification of both co-localized and underscore unreported loci underscores the polygenic nature of SL and advances current understanding of the genetic factors shaping spike architecture. Future work focusing on fine mapping, functional validation, and the integration of yield-related traits such as thousand-kernel weight and grain yield will be essential to translate these findings into practical tools for wheat improvement.

Conclusion

Two novel loci were identified controlling SL in wheat in 182 genetically diverse accessions of Chinese endemic wheat. Furthermore, one candidate gene for SL was predicted based on the expression level in spike tissues providing useful genetic markers and elite germplasm for future breeding programs for future crop improvement.

Abbreviations

AD

Anthesis Date

BLUP

Best Linear Unbiased Prediction

CZ

Chongzhou

CV

Coefficient of Variation

ETN

Effective Tiller Number

GWAS

Genome Wide Association Study

H²

Broad-Sense Heritability

LD

Linkage Disequilibrium

PH

Plant Height

QTL

Quantitative Trait Locus

SNP

Single Nucleotide Polymorphism

SL

Spike Length

SD

Spike Density

SNS

Spikelet Number per Spike

STDEV

Standard deviation

WJ

Wenjiang

Authors’ contributions

PYD, HPZ, CW: analyzed the data, wrote the draft manuscript, and interpreted the entire results. WTZ: measured and analyzed the data. MNL: participated in field trials, data measurement, and GWAS analysis. JTC, XY, HMFU, MJJ: managed field work and participated in data collection. YFJ, GYC: discussed the results. QX and JM: designed the experiments, directed the entire study, participated in data analysis, and substantially revised this manuscript. All authors participated in the research and approved the final manuscript.

Funding

This research was supported by National Key R&D Program of China (2023YFD1201900), Sichuan Science and Technology Program (2026YFHZ0010), Mianyang Teachers’ College Research Initiation Project (QD2023A12), National Natural Science Foundation of China (32341027), Natural Science Foundation of Jiangsu (BK20221509), and Jiangsu Key R&D Program (BE2023370).

Data availability

The variation data reported in this paper have been deposited in the Genome Variation Map (GVM) [44] in National Genomics Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation [45], under accession number GVM001351.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.