Structure and Quantitative‐Structure‐Function Relationships of Wheat High‐Molecular‐Weight Glutenin Subunits: From Allelic Polymorphism to Dough Rheology

ABSTRACT

High‐molecular‐weight glutenin subunits (HMW‐GSs) form the structural backbone of the wheat gluten network, and their compositional and structural polymorphisms strongly shape dough rheological properties. As wheat processing becomes increasingly standardized and specialized, the requirements for dough properties, especially strength and extensibility, are becoming more stringent and diverse. This makes it essential to clarify the structure–function and dose–effect relationships between HMW‐GS and dough rheological properties. On the basis of current evidence, this review focuses on three main aspects: (1) the contributions of HMW‐GS allelic variation to dough strength and extensibility; (2) the molecular mechanisms by which multidimensional structural features of HMW‐GSs determine these rheological traits; and (3) the relationships between multidimensional structural polymorphism and dough strength and extensibility. Subunits, such as Ax1, Ax2*, Bx14 + By15, and Bx17 + By18, can synergistically enhance both strength and extensibility, whereas Bx7^OE + By8 and Dx5+Dy10 generally enhance strength at the expense of extensibility. We propose a multidimensional structure–function framework in which “loop‐train” motifs and helical conformations confer elasticity; disulfide bonds cross‐link elastic units into a gluten network; and non‐covalent interactions (e.g., hydrogen bonding, hydrophobic, and ionic interactions) cooperate to build and stabilize the network. Single amino‐acid substitutions at key residues can shift dough properties by altering local conformations and intermolecular interactions, including cysteine substitutions (e.g., Cys10Ser‐N, Cys40Ser‐N, Ser8Cys‐central repetitive domain [CRD], Tyr612Cys‐CRD, and Cys25Ser‐N) and non‐cysteine substitutions (e.g., Gly244Glu‐CRD for Ax1). However, reported sequence polymorphisms‐rheological traits remain difficult to reconcile with large‐scale sequence datasets, and systematic analyses of higher order structural polymorphisms remain limited. Future work should resolve these gaps to enable more precise quality control and rational design of wheat‐based products.

1. Introduction

Wheat‐based products constitute a major global food staple, ranging from noodles and steamed bread in Asia, bread in Europe, and flatbreads in the Middle East, to biscuits, cakes, and pastries that are widely consumed worldwide (Hoque and Islam 2024). The remarkable diversity of wheat‐based products is not merely a consequence of processing traditions; rather, it is fundamentally enabled by the distinctive viscoelastic behavior of wheat dough, which is largely governed by gluten proteins. In practical terms, gluten confers the capacity of dough to resist deformation while still undergoing controlled extension and flow, allowing it to be mixed, sheeted, molded, fermented, and baked into products with contrasting textures and structures. Consequently, understanding and controlling dough rheology is central to improving processing performance and ensuring end‐use quality in wheat‐based manufacturing.

Dough rheological properties describe how dough deforms and flows when subjected to external stresses during unit operations, such as hydration, mixing, kneading, resting, sheeting, and shaping. These properties are typically quantified using empirical and fundamental instruments, including the mixograph, farinograph, extensograph, alveograph, texture analyzer, and rheometer (Spina et al. 2021). Dough rheological behavior is mainly manifested in properties, such as strength, extensibility, elasticity, and viscosity. Compared with elasticity and viscosity, dough strength and extensibility are more readily characterized and are considered the most critical quality parameters. Strength broadly reflects the resistance of dough to deformation and rupture, whereas extensibility reflects the ability of dough to stretch without tearing, enabling gas retention, sheet ability (pancake, flatbreads, noodles, etc.), and structural development depending on the product type. Different wheat‐based products exhibit markedly different requirements for dough strength and extensibility. For example, biscuits require dough with low strength and high extensibility; noodle production demands a combination of moderate strength and high extensibility, whereas breadmaking requires dough with high strength and moderate extensibility (Liu et al. 2025; Wang et al. 2018; Yang, Wang, et al. 2023).

At the molecular level, wheat gluten proteins consist primarily of glutenins and gliadins, which together assemble into a continuous protein network upon hydration and mechanical work. Glutenins can be further classified into high‐molecular‐weight glutenin subunits (HMW‐GSs) and low‐molecular‐weight glutenin subunits (LMW‐GSs). HMW‐GSs are encoded by three loci, Glu‐A1, Glu‐B1, and Glu‐D1, located on the long arms of the Group 1 homoeologous chromosomes of wheat. Each locus typically comprises two tightly linked subunits: an x‐type subunit with a higher molecular weight and a y‐type subunit with a lower molecular weight (Payne 1987). HMW‐GSs exhibit extensive allelic variation in wheat; since the publication of the catalogue of gene symbols for wheat, more than 100 HMW‐GS alleles have been identified at the Glu‐1 loci (McIntosh 2024). Although the Ay subunit at the Glu‐A1 locus is silent in most bread wheat cultivars, functional Ay alleles have been identified and reactivated. In particular, introgression of the expressed 1Ay21* allele significantly increased grain protein content and breadmaking quality (Roy et al. 2020, 2021).

Using diverse genetic materials, including natural populations, recombinant inbred lines, and near‐isogenic lines (NILs), numerous studies have examined how HMW‐GS allelic variation contributes to dough strength and extensibility (Gao et al. 2016; Guzmán et al. 2022; Zhou et al. 2024). Although HMW‐GSs account for only ∼15% of total gluten proteins, their allelic variation has been estimated to explain a remarkably large fraction (approximately 30%–70%) of the genetic variation in dough rheological traits (Eagles et al. 2002; Gao et al. 2016; Liu et al. 2005; Payne et al. 1988). Although most studies confirm that HMW‐GS allelic variation has a significant impact on dough rheology properties, substantial discrepancies among studies persist, and a consensus has yet to be reached regarding the contributions of specific HMW‐GS alleles to dough strength and extensibility.

The molecular structure of HMW‐GS is characterized by highly conserved domains as follows: an N‐terminal domain comprising 81–104 amino‐acid residues, a central repetitive domain (CRD) rich in glutamine residues that adopts a β‐spiral conformation, and a C‐terminal domain containing 42 amino‐acid residues. Both the N‐ and C‐terminal domains contain at least one cysteine residue, which participates in intermolecular disulfide bond formation, thereby establishing the backbone of the gluten protein network (Shewry et al. 2023). Despite the high conservation of the primary structure of HMW‐GS, substantial differences exist among individual subunits, including amino‐acid substitutions, insertions, and deletions. These variations modulate the extent to which different HMW‐GS contribute to dough strength and extensibility (Du et al. 2018; Li et al. 2004).

Existing high‐quality reviews have provided an important foundation for understanding HMW‐GS, although their emphases differ. Previous reviews have primarily focused on HMW‐GS allelic variation and its relationships with dough quality and breadmaking quality, as well as on the polymorphism of HMW‐GS primary structure (Shewry, Halford, et al. 2003; Lafiandra and Shewry 2022; Shewry and Lafiandra 2022). However, comprehensive reviews that specifically address the two core rheological phenotypes of dough strength and extensibility, while systematically integrating the multidimensional structural features and polymorphisms of HMW‐GS with dough rheological properties, remain limited. In this context, the present review focuses on how the composition and molecular conformation of HMW‐GS govern dough strength and extensibility. The goal is to deepen our understanding of HMW‐GS structure–function relationships and to provide a theoretical basis for breeding wheat cultivars for specific end uses, specialty flour production, quality control of wheat‐based products, and the development of new products.

2. Natural Distribution and Genetic Background of HMW‐GS Polymorphism

In theory, each hexaploid wheat genotype carries six HMW‐GSs, corresponding to the paired x‐ and y‐type genes at the three Glu‐1 loci: Glu‐A1 (Ax + Ay), Glu‐B1 (Bx + By), and Glu‐D1 (Dx + Dy). However, due to factors such as gene silencing, common wheat generally contains only three to five HMW‐GS subunits: zero or one subunit at the Glu‐A1 locus (Ax), one (Bx) or two (Bx + By) subunits at the Glu‐B1 locus, and two subunits (Dx + Dy) at the Glu‐D1 locus. To characterize the distribution of HMW‐GS allelic variation, a literature survey was conducted on the Google Scholar, Web of Science, and PubMed using the keyword combinations “HMW‐GS + allelic,” “HMW‐GS + the designated country,” and “Glu1 + the designated country” with the time frame set from 2000 to the present. The designated countries are the top 20 countries in terms of wheat production. From the retrieved studies, datasets describing the frequency of HMW‐GS alleles in wheat cultivars were extracted, harmonized, and compiled for 16 countries. This set comprised 15 countries ranked among the world's top 20 wheat producers, ensuring broad representation of major production regions, together with one additional country selected because of a notably high frequency of the Dx2.2 + Dy12 allele at Glu‐D1. The collected datasets were systematically integrated and analyzed, and the results are presented in Figure 1 and Table S1.

FIGURE 1.

Distribution patterns of HMW‐GS allelies in wheat cultivars from major wheat‐producing countries: (A) Glu‐A1; (B) Glu‐B1; and (C) Glu‐D1. The data were compiled from the following 15 references: Branlard et al. (2003), Filip (2018), Gianibelli et al. (2002), Izadi‐Darbandi et al. (2010), Jin et al. (2011), Khalid and Hameed (2019), Kozub et al. (2009), Lawrence (1986), Liu et al. (2005), Lukow et al. (1989), Nagamine et al. (2000), Ram et al. (2015), Shan et al. (2007), Sönmez et al. (2023), and Utebayev et al. (2019). This figure was generated using CNSknowall (https://www.cnsknowall.com/#/HomePage).

At the Glu‐A1 locus, three allelic states dominated the surveyed germplasm: Ax1, Ax2*, and Ax‐null. Each of these variants occurred at comparatively high frequencies in 13 of the major wheat‐producing countries examined. Across countries, Ax1 showed frequency ranges of 14.9%–55.1%, Ax2* ranged from 15.5% to 76.1%, and Ax‐null ranged from 12.3% to 73.6%.

At the Glu‐B1 locus, the most common allelic variants were Bx7 + By8 and Bx7 + By9, which occurred at relatively high frequencies in 11 and 12 major wheat‐producing countries, respectively, with frequency ranges of 14.3%–65.8% and 18.2%–85.0%. Other allelic variants, including Bx7, Bx6 + By8, Bx20 + By20, Bx13 + By16, and Bx17 + By18, were present at relatively high frequencies in specific countries. For instance, the Bx7 subunit accounted for 20.5% of wheat cultivars in France; Bx6 + By8 accounted for 24.5%, 21.1%, 41.3%, and 10.7% in France, Turkey, Germany, and Poland, respectively; Bx20 + By20 accounted for 20.6%, 11.7%, and 17.9% in Australia, Pakistan, and Poland, respectively; Bx13 + By16 accounted for 44.2% in Pakistan; and Bx17 + By18 accounted for 35.0%, 30.2%, 28.5%, 16.4%, and 17.4% in India, Australia, Pakistan, Iran, and Argentina, respectively.

At the Glu‐D1 locus, two allelic variants were consistently prevalent across the 15 major wheat‐producing countries: Dx2 + Dy12 and Dx5 + Dy10. Both occurred at relatively high frequencies in all surveyed major producers, with Dx2 + Dy12 ranging from 10.4% to 71.6% and Dx5 + Dy10 ranging from 17.8% to 88.0%. Beyond these dominant alleles, several lower frequency variants displayed geographically restricted distributions. The Dx4 + Dy12 variant was present in wheat cultivars from China (12.3%), France (7%), and Australia (0.8%). The Dx3 + Dy12 variant occurred in cultivars from the United States of America (USA), France, Germany, Iran, and Romania, accounting for 7.2%, 5.0%, 1.9%, 3.0%, and 2.0%, respectively. In addition, the Dx2.2 + Dy12 variant was particularly frequent in Japanese wheat cultivars, with a proportion of 58.8%.

According to current research data (McIntosh 2024), over 100 allelic variations of HMW‐GS have been identified in the wheat gene resource bank. However, globally registered wheat varieties in field production exhibit that the variations in HMW‐GS are predominantly concentrated in just 16 major variant types. Li et al. (2014) demonstrated that both the allelic composition and population‐level frequencies of HMW‐GS variants are significantly associated with the geographic distribution of wheat germplasm, implying that spatially structured selective forces have contributed to their evolutionary differentiation. Such geographic patterning likely reflects a combination of environmental filtering (e.g., adaptation to local agroecological conditions) and, importantly, sustained human‐mediated selection arising from region‐specific end‐use requirements. In particular, regional dietary traditions and processing technologies impose distinct functional demands on dough viscoelasticity, thereby shaping breeding targets and indirectly influencing the retention or enrichment of specific HMW‐GS alleles. For example, bread‐oriented production systems in many Western countries typically prioritize strong dough and high gas‐holding capacity to support fermentation and loaf expansion; accordingly, cultivars harboring the Dx5 + Dy10 subunit combination are often selected and thus occur at relatively high frequencies. In contrast, in East Asian regions where noodles and steamed breads are staple products and optimal quality relies more on balanced extensibility and moderate gluten strength, cultivars carrying the Dx2 + Dy12 combination tend to be more prevalent. Collectively, these observations support the view that geographic differentiation in HMW‐GS distribution is a meaningful signature of long‐term, product‐driven directional selection.

3. Effects of HMW‐GS Alleles on Dough Rheological Properties

Natural populations typically comprise a diverse set of wheat cultivars and are therefore convenient for surveying allelic effects across broad germplasm panels. However, when natural populations are used as experimental materials, estimates of HMW‐GS effects are often confounded by heterogeneous genetic backgrounds, because many nontarget loci simultaneously contribute to dough traits. Although enlarging the sample size can partially dilute background noise and improve statistical power, the inference remains constrained by the uneven occurrence of HMW‐GS subunits—some alleles are rare and thus insufficiently represented for robust effect estimation. In contrast, NILs are developed through repeated backcrossing coupled with marker‐assisted selection, yielding materials with highly uniform genetic backgrounds that differ only at target genomic region, or even at a single gene. Theoretically, after six generations of backcrossing, background similarity can reach ∼99.2%, enabling much cleaner attribution of phenotypic differences to the focal locus. Consequently, NILs are widely regarded as ideal materials for dissecting genotype–phenotype relationships and are indispensable for clarifying gene functions and protein‐mediated mechanisms. Importantly, NILs can compensate for the stochasticity and rarity constraints of natural populations, albeit at the cost of substantial time and labor during development.

This review focuses primarily on the effects of HMW‐GS on dough strength and extensibility. The mixograph, farinograph, extensograph, and alveograph are the most commonly used instruments to characterize dough rheological properties. The mixograph and farinograph evaluate resistance to shear deformation, whereas the extensograph and alveograph measure resistance to tensile deformation in two and three dimensions, respectively (Figure 2). On the basis of the physical interpretation of instrument parameters, the mixograph weakening slope, farinograph stability time, extensograph maximum resistance, and alveograph P are typically used to represent dough strength, whereas extensograph extensibility and alveograph L are commonly used to evaluate dough extensibility (Figure 2).

FIGURE 2.

Schematic illustration of the main dough rheological methods.

3.1. Effects of HMW‐GS Allelic Variation on Dough Strength

To clarify the contribution of HMW‐GS to dough strength and extensibility, we performed literature retrieval using the keyword combinations “HMW‐GS + dough properties,” “HMW‐GS + Mixograph,” “HMW‐GS + Farinograph,” “HMW‐GS + Extensograph,” and “HMW‐GS + Alveograph” in Google Scholar, Web of Science, and PubMed, with the time span set from 2000 to the present. The retrieved literature was then manually screened.

For studies on natural populations, we selected those with relatively large population sizes (at least 100 accessions) and with relatively representative materials, whereas for NILs, we included nearly all available studies. As summarized in Figure 3 and Table S2, previous studies have investigated the relationships between HMW‐GS allelic variation and dough strength using 162, 240, 251, and 2550 wheat cultivars (lines) from Europe, India, China, and the CIMMYT spring wheat breeding program, respectively (Branlard et al. 2001; Guzmán et al. 2022; Liu et al. 2005; Ram et al. 2015). At the Glu‐A1 locus, European, Chinese, and CIMMYT materials consistently showed that dough strength associated with Ax1 and Ax2* was greater than that of Ax‐null, although the relative contributions of Ax1 and Ax2* differed among populations. In contrast, the Indian population exhibited an opposite trend, with Ax‐null conferring higher dough strength than Ax2* and Ax1. At the Glu‐B1 locus, after excluding low‐frequency subunits, such as Bx20 + By20, Bx6 + By8, Bx7, and Bx14 + By15, the results from Indian, Chinese, and CIMMYT materials were highly consistent. Dough strength generally followed the ranking Bx7^OE + By8 > Bx17 + By18 > Bx7 + By8 > Bx7 + By9 > Bx13 + By16. Notably, European materials displayed a markedly different ranking at Glu‐B1: Bx17 + By18 ≥ Bx13 + By16 ≥ Bx7 + By9 ≥ Bx7 + By8. At the Glu‐D1 locus, the relative contributions of Dx5 + Dy10 compared with Dx2 + Dy12, Dx3 + Dy12, and Dx4 + Dy12 were consistent across studies. Dough strength followed the ranking Dx5 + Dy10 > Dx2 + Dy12/Dx3 + Dy12/Dx4 + Dy12, whereas differences among Dx2 + Dy12, Dx3 + Dy12, and Dx4 + Dy12 were minor.

FIGURE 3.

Effect of HMW‐GS allele on the dough strength and extensibility based on natural population and NILs. The color gradient indicates the rank order of dough strength or extensibility, rather than the absolute magnitude of these traits. The numbers denote the frequencies of the corresponding allelic variants in the natural population. Within the same natural population or near‐isogenic line background, different lowercase letters indicate significant differences in dough strength or extensibility among subunits. The data were compiled from the following 17 references: Branlard et al. (2001), Cooper et al. (2016), Deng et al. (2005), Gao et al. (2016), Guzmán et al. (2022), He et al. (2005), Jin et al. (2013), Li et al. (2016, 2020), Liu et al. (2005), Ram et al. (2015), Wang et al. (2018), Zhang et al. (2008, 2010), Zhao, Gao et al. (2020), Zhao, Li et al. (2020), Zhou et al. (2024, 2025). NILs, near‐isogenic lines.

Using NILs developed in wheat cultivar backgrounds from the USA, China, and Australia, numerous studies further evaluated the effects of HMW‐GS on dough strength (Cooper et al. 2016; Deng et al. 2005; Gao et al. 2016; Li et al. 2016, 2020; Wang et al. 2018; Zhang et al. 2008, 2010; Zhao, Gao et al. 2020; Zhao, Li et al. 2020; Zhou et al. 2024, 2025; Jin et al. 2013). At the Glu‐A1 locus, most NILs showed that Ax1 and Ax2* conferred greater dough strength than Ax‐null. However, the relative contributions of Ax1 and Ax2* varied among NILs, consistent with results from natural populations. Notably, in the Australian cultivar Aroona background, Ax2* exhibited lower dough strength than Ax‐null, indicating that genetic background can modulate the effects of specific subunits. It was also noteworthy that expression of the normally silent 1Ay subunit can further enhance dough and breadmaking quality. Roy et al. (2020) demonstrated that introgression of the expressed 1Ay21* allele into bread wheat increased grain protein content, UPP%, and bread volume, indicating that Ay subunits can positively contribute to dough strength and processing performance when expressed. At the Glu‐B1 locus, the integration of results from different NILs indicated the following ranking for dough strength: Bx7^OE + By8 > Bx17 + By18/Bx14 + By15 > Bx7 + By8 > Bx7 + By9 > Bx7* + By8 > Bx6 + By8* > Bx7*. This ranking is largely consistent with findings from natural populations. At the Glu‐D1 locus, Dx5 + Dy10 conferred significantly greater dough strength than Dx2 + Dy12, Dx3 + Dy12, Dx4 + Dy12, and Dx2.2 + Dy12. Differences among Dx2 + Dy12, Dx3 + Dy12, Dx4 + Dy12, and Dx2.2 + Dy12 were relatively small.

3.2. Effects of HMW‐GS Allelic Variation on Dough Extensibility

As shown in Figure 3 and Table S3, previous studies have examined the relationships between HMW‐GS allelic variation and dough extensibility using 162, 126, and 2550 wheat cultivars (lines) from Europe, China, and the CIMMYT spring wheat breeding program, respectively (Branlard et al. 2001; Guzmán et al. 2022; He et al. 2005). At the Glu‐A1 locus, CIMMYT and Chinese materials consistently showed that Ax1 and Ax2* conferred greater extensibility compared with Ax‐null, although the relative contributions of Ax1 and Ax2* varied among populations. In contrast, European materials showed no consistent differences in dough extensibility among Glu‐A1 allelic variants. At the Glu‐B1 locus, substantial differences in subunit composition were observed among the three populations. In general, Bx7 + By8 exhibited higher extensibility than Bx7 + By9 across all three populations, and Bx13 + By16 showed the highest extensibility in both European and CIMMYT populations. At the Glu‐D1 locus, the relative contributions of Dx5 + Dy10 and Dx2 + Dy12 to dough extensibility differed markedly among populations. In CIMMYT materials, Dx5 + Dy10 conferred greater extensibility than Dx2 + Dy12; in European materials, no significant difference was observed, whereas in Chinese materials, Dx5 + Dy10 conferred lower extensibility than Dx2 + Dy12.

Using NILs developed in Chinese and Australian cultivar backgrounds, further evaluations of HMW‐GS effects on dough extensibility were conducted (Deng et al. 2005; Gao et al. 2016; Li et al. 2016; Wang et al. 2018; Zhang et al. 2008, 2010; Jin et al. 2013). At the Glu‐A1 locus, dough extensibility ranked as Ax2* > Ax1 > Ax‐null, providing new experimental evidence for the previously debated differences between Ax1 and Ax2* observed in natural populations. At the Glu‐B1 locus, extensibility rankings varied substantially across genetic backgrounds. Subunits Bx14 + By15 and Bx17 + By18 generally exhibited superior extensibility. Specifically, in the Lumai‐16 and Longmai‐20 backgrounds, Bx14 + By15 and Bx17 + By18 showed greater extensibility than Bx7 + By8 and Bx7 + By9; in the Aroona background, Bx17 + By18 exhibited greater extensibility than Bx7 + By8. Subunits Bx7 + By8 exhibited greater extensibility than Bx7 + By9 in the Xinong‐1330 background. The extensibility of Bx6 + By8* was lower than that of Bx7 + By8 and Bx7 + By9 in the Aroona background. At the Glu‐D1 locus, the extensibility of Dx2 + Dy12, Dx3 + Dy12, Dx4 + Dy12, and Dx2.2 + Dy12 was markedly greater than that of Dx5 + Dy10, except in the Xinong‐1330 background. Differences in extensibility among Dx2 + Dy12, Dx3 + Dy12, Dx4 + Dy12, and Dx2.2 + Dy12 were relatively small. Although NILs control more than 99% of genetic background similarity, the effects of certain subunits on dough strength and extensibility, particularly extensibility, remain background‐dependent. These background‐dependent subunits may exert relatively small effects, such that residual genetic differences, subunit interactions, and environmental factors obscure these effects.

Simultaneous enhancement of dough strength and extensibility is a recurring objective for many wheat‐based products, particularly under industrial processing conditions where dough must tolerate high‐energy mixing, rapid machining, and stringent line stability requirements (Li et al. 2026). In practice, however, improving these traits concurrently is inherently difficult because strength and extensibility are often strongly and negatively correlated, reflecting a trade‐off between resistance to deformation and capacity for elongation before rupture (Nash et al. 2006; Jødal and Larsen 2021). Consequently, identifying HMW‐GS that can shift the strength‐extensibility balance in a favorable direction remains a central challenge for both wheat breeding and flour functionality design. When evidence from both natural populations and NILs is integrated, several subunit‐specific trends become apparent. At the Glu‐A1 locus, Ax1 and Ax2* enhance both dough strength and extensibility relative to Ax‐null. At the Glu‐B1 locus, Bx14 + By15 and Bx17 + By18 enhance both traits compared with Bx7 + By8 and Bx7 + By9; Bx7^OE + By8 primarily enhances dough strength at the expense of extensibility. At the Glu‐D1 locus, Dx5 + Dy10 enhances dough strength but is associated with reduced extensibility. These findings provide valuable guidance for improving dough strength and extensibility. Notably, the Bx13 + By16 subunit exhibits favorable performance in both dough strength and extensibility in European natural populations but performs poorly in dough strength in CIMMYT materials, suggesting potential for simultaneous improvement of both traits. Future validation of its structure–function relationships using NILs across diverse genetic backgrounds is warranted.

The contribution trends of subunit 5 + 10 to dough strength and of subunits 14 + 15 and 17 + 18 to both strength and extensibility are evident; however, the influence of genetic background cannot be overlooked, particularly for extensibility. Therefore, in breeding and downstream food‐processing applications, subunit selection should account for background effects and consider potential interactions among subunits. The observed variations in dough strength and extensibility across different HMW‐GS alleles underscore the necessity to elucidate the underlying structural mechanisms. In the following sections, we systematically dissect the multidimensional structure of HMW‐GS and examine how structural polymorphisms govern rheological performance.

4. Multidimensional Structure of HMW‐GS and Their Contribution to Dough Rheology

4.1. Primary Structure: Conserved Structure, Amino‐Acid Composition, and the Function of Key Site Polymorphisms

4.1.1. Conserved Structure and Amino‐Acid Composition

As shown in Figure 4, the primary structure of HMW‐GS generally exhibits a highly conserved framework, which can be divided into three domains: the N‐terminal domain, the CRD, and the C‐terminal domain. Within the CRD, x‐type subunits are rich in the GQQ tripeptide, PGQGQQ hexapeptide, and GYYPTS(P/L)QQ nonapeptide, whereas y‐type subunits primarily contain PGQGQQ hexapeptides and GYYPTS(P/L)QQ nonapeptides. The amino‐acid composition of HMW‐GS is highly biased toward a limited number of residues (exemplified by Dx5, Figure 4), with glutamine (Gln, Q), glycine (Gly, G), proline (Pro, P), serine (Ser, S), tyrosine (Tyr, Y), leucine (Leu, L), alanine (Ala, A), and threonine (Thr, T) accounting for 36.2%, 20.1%, 13.2%, 5.7%, 5.6%, 4.4%, 3.0%, and 2.9%, respectively, totaling 90.9%. Q, S, Y, and T account for 50.3%, and their side‐chain amide or hydroxyl groups contribute strong hydrogen bond formation abilities. This not only supports the formation of β‐turns but also lays the chemical foundation for subsequent “loop‐train” hydrogen bond networks. Acidic and basic amino acids are rare and mainly localized in the terminal domains. All HMW‐GS contain at least one cysteine (Cys) residue at both the N‐ and C‐terminal domains (Shewry et al. 2023).

FIGURE 4.

The polymorphism of primary structure: (A) conserved domains and their amino acids lengths; (B) the number and distribution of cysteine; (C) the number of repetitive peptides; (D) amino acid composition (number); and (E) amino acid composition (%). Primary structure data of glutenins proteins were obtained from the NCBI database (https://www.ncbi.nlm.nih.gov).

4.1.2. Polymorphism of Sequence Length, Amino‐Acid Composition, and Cysteine Residues

Over the conserved structural framework, the HMW‐GS protein sequences exhibit substantial polymorphism (such as amino‐acid substitutes, insertions, and deletions), primarily reflected in the following three aspects (Figure 4). (1) Length polymorphism: There are significant length differences among subunits, with the shortest being Ay21* at 567 amino acids (aa) and the longest being Dx2.2* at 1003 aa. This variation is mainly due to changes in the length of the CRD, leading to alterations in the number of tripeptide, hexapeptide, and nonapeptide repeats (Gregova et al. 2016; Wan et al. 2005). (2) Amino‐acid composition variability: The content of Gln, Gly, Pro, Ser, Tyr, Leu, Ala, and Thr fluctuates significantly among different subunits, with their quantities ranging from 181 to 372, 100 to 207, 55 to 137, 37 to 72, 28 to 56, 21 to 42, 17 to 29, and 18 to 30, respectively; and their proportions ranging from 31.9% to 37.1%, 17.6% to 20.6%, 9.5% to 13.7%, 5.5% to 9.0%, 5.0% to 7.2%, 2.8% to 5.2%, 2.1% to 4.0%, and 2.8% to 3.8%. (3) Polymorphism in Cys number and position: The number of Cys residues in different subunits ranges from 2 to 8, with significant positional differences. A typical x‐type subunit contains three Cys residues in the N‐terminal domain and one Cys residue in the C‐terminal domain (Shewry et al. 1992). Bx14 and Bx20 subunits have one Cys residue at each terminal domain (Li et al. 2004; Shewry, Gilbert, et al. 2003), whereas the Dx5 subunit contains three Cys residues in the N‐terminal, one in the CRD, and one in the C‐terminal domain (Pirozi et al. 2008). A typical y‐type subunit contains five Cys residues in the N‐terminal, one in the CRD, and one in the C‐terminal domain (Shewry et al. 1992). The Ay21* subunit has five Cys residues in the N‐terminal and one Cys residue in the C‐terminal domain, whereas the Dy12** subunit has five, one, and two Cys residues in the respective domains (Du et al. 2018).

4.1.3. Link Between Key Site Polymorphisms and Dough Strength and Extensibility

4.1.3.1. Cys Residue Number and Position

Cys residues are indispensable for disulfide bond formation and therefore play a decisive role in establishing the glutenin polymers. Because HMW‐GS constitute the backbone of the gluten network, both the number and positional distribution of Cys residues within HMW‐GS can markedly influence polymer size, cross‐link density, and ultimately dough strength (Figure 5). In general, an increased number of Cys is positively associated with dough strength, consistent with the notion that additional thiol groups expand the capacity for intermolecular disulfide crosslinking. For instance, typical x‐type subunits such as Ax2* (Cys10‐N, Cys22‐N, Cys37‐N, and Cys30‐C denote cysteine residues at positions 10, 22, and 37 of the N‐terminal domain and at position 30 of the C‐terminal domain, respectively) and Dx2 (Cys10‐N, Cys25‐N, Cys40‐N, and Cys30‐C) contain four Cys residues, whereas Ax2*^B (Cys10‐N, Cys22‐N, Cys37‐N, Cys287‐CRD, and Cys30‐C) and Dx5 (Cys10‐N, Cys25‐N, Cys40‐N, Cys8‐CRD, and Cys30‐C) possess an additional Cys residue in the CRD. This extra Cys residue is associated with higher dough strength (Gupta and MacRitchie 1994; Juhász et al. 2001). A similar pattern is observed for variants with extra Cys residues in terminal regions. The Dx12** subunit (Cys10‐N, Cys22‐N, Cys44‐N, Cys45‐N, Cys55‐N, Cys403‐CRD, Cys13‐C, and Cys30‐C), which contains an additional Cys residue in the C‐terminal domain, is linked to higher SDS sedimentation values compared with the typical Dy10 and Dy12 subunits (Cys10‐N, Cys22‐N, Cys44‐N, Cys45‐N, Cys55‐N, Cys416/421‐CRD, and Cys30‐C) (Du et al. 2018). Conversely, reduced Cys availability can weaken dough strength. The Bx20 subunit (Cys10‐N and Cys30‐C) lacks two N‐terminal Cys residues relative to the typical Bx7 subunit (Cys10‐N, Cys17‐N, Cys32‐N, and Cys30C), and this deficiency is associated with inferior dough strength, consistent with diminished crosslinking capacity (Pirozi et al. 2008).

FIGURE 5.

Multidimensional structure of HMW‐GS and its relationship with dough rheological properties. The figure was primarily developed on the basis of the findings reported by Field et al. (1987), Belton et al. (1995), Belton (1999), Tamás et al. (2002), Wieser (2007), Li et al. (2015), Wang et al. (2017, 2021), Yang, Chen, et al. (2023), Wei et al. (2025), and Wang et al. (2026). The schematic diagram was created using https://BioRender.com. CRD, central repetitive domain; HMW‐GS, high‐molecular‐weight glutenin subunit.

Beyond sheer Cys number, the positional context of Cys residues is critical because it governs whether a residue preferentially participates in intermolecular crosslinking (thereby strengthening the gluten network) or is sequestered into intramolecular disulfide bonds (potentially reducing network connectivity). Gao et al. (2012) reported a mutant Bx7^OE subunit carrying a Ser‐to‐Cys substitution at residue 533 in the CRD (Ser533Cys‐CRD); this mutant was associated with a lower content of unextractable proteins, implying decreased inter‐chain covalent stabilization, suggesting that the additional Cys preferentially forms intramolecular disulfide bonds, thereby disrupting inter‐chain crosslinking. To dissect Cys functionality more directly, Wang et al. (2017) examined the N‐terminal domain of Dx5 (Dx5‐N) and showed that Cys10‐N and Cys40‐N are crucial for inter‐chain disulfide bond formation. They further found that Cys40‐N can also form an intra‐chain disulfide bond with Cys25‐N. In this framework, a Cys‐to‐Ser substitution at residue 25 in N‐terminal domain (Cys25Ser‐N)—thereby reducing an intramolecular pairing option—produced a positive effect on dough strength relative to the wild‐type Dx5‐N, whereas mutations at other positions did not yield comparable strengthening. More recently, Wei et al. (2025) used reciprocal site‐directed mutants of Dx2 (Ser8Cys‐Rep) and Dx5 (Cys8Ser‐CRD), confirming the importance of Cys8‐CRD in enhancing dough strength. They additionally showed that a substitution at position 612 in the CRD of Dx2 (Tyr612Cys‐CRD) significantly increased dough strength and that this effect was additive with the Ser8Cys‐CRD substitution, indicating that multiple strategically placed Cys residues can contribute cumulatively to network reinforcement.

Taken together, these studies support a coherent mechanistic interpretation: dough strength can be effectively enhanced by increasing the number of Cys residues that are available for intermolecular disulfide bond formation or, alternatively, by reducing Cys residues that are prone to forming intramolecular disulfide bonds that “consume” reactive thiols without increasing polymer connectivity. On the basis of current evidence, Cys10‐N, Cys40‐N, Cys8‐CRD, Cys612‐CRD, and Cys30‐C appear to have substantial potential to participate in intermolecular disulfide bonding, whereas Cys25‐N shows a tendency toward intramolecular pairing. These sites therefore represent promising targets for wheat quality improvement via gene‐editing approaches.

4.1.3.2. Amino‐Acid Length of the CRD

The number and positional distribution of Cys residues, while central to disulfide‐mediated polymerization, are not sufficient on their own to explain the full spectrum of variation in dough strength observed among HMW‐GS alleles and allele combinations. For example, a Dx5' subunit that exhibits the same electrophoretic mobility as the conventional Dx5 subunit on sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS–PAGE) carries a serine substitution at the eighth residue of the CRD (Cys8Ser‐CRD) yet displays dough‐quality performance comparable to that of Dx5 (Ren et al. 2008). Likewise, the subunit combinations Bx20 + By20 and Bx14 + By15 have identical numbers and positional distributions of Cys residues; however, the Bx14 + By15 combination confers markedly greater dough strength than Bx20 + By20 (Gao et al. 2016; Pirozi et al. 2008).

Indeed, the CRD of HMW‐GS displays pronounced length polymorphism, and accumulating evidence supports a close association between repetitive‐region length and dough rheology. Using hordein genes to construct recombinant polypeptides of defined repeat lengths, followed by heterologous expression and chemical integration assays, Tamás et al. (2002) demonstrated that increasing the length of the central repetitive region is positively correlated with dough strength but negatively correlated with extensibility. This trade‐off is consistent with the concept that longer CRDs provide more interaction sites and greater capacity for entanglement and cooperative non‐covalent bonding, thereby reinforcing the glutenins’ network; however, the same increase in interaction density may restrict chain mobility and reduce stretchability. Complementary evidence comes from homoeologous substitution studies in which non‐native HMW‐GS (e.g., DxNS or 1Sl×2.3*/1Sly16*) were introduced into the D‐ and B‐genome loci of common wheat, resulting in marked increases in dough strength in the derived substitution lines (Guo et al. 2021; Wang et al. 2013). Nevertheless, the phenotypic impact of repeat‐length variation appears to be scale‐dependent. In a doubled haploid population, Gobaa et al. (2007) reported that Ax2•• contains only nine additional amino‐acid residues in the central repetitive region relative to Ax2*, yet this modest increment did not produce statistically significant effects on dough strength or extensibility—potentially reflecting a true biological threshold, limited detection resolution in phenotyping, or masking by genetic background and environmental variance.

4.1.3.3. Variations at Other Amino‐Acid Sites

Variations at amino‐acid sites beyond cysteine—particularly at highly conserved residues such as glycine—can also be decisive for the functional performance of HMW‐GS (Wellner et al. 2006) (Figure 5). In contrast to Cys, which primarily affects dough properties through disulfide‐mediated covalent crosslinking, non‐Cys substitutions often exert their effects by reshaping local conformational preferences, altering the balance of non‐covalent interactions (e.g., hydrogen bonding and hydrophobic association), and modulating chain flexibility within the repetitive domain. A representative example is provided by mutations in the Dy12 subunit. Yang, Chen, et al. (2023) reported that replacement of glycine with arginine at position 284 in the CRD (Gly284Arg‐CRD) induced a marked shift in predicted secondary structure, promoting a transition from random coil toward α‐helical propensity. This conformational remodeling was accompanied by reduced electrophoretic mobility on SDS–PAGE, consistent with altered protein shape and/or detergent binding, and—more importantly—by weakened gluten network strength and a reduction in dough extensibility. Importantly, non‐Cys substitutions can also generate favorable outcomes and, in some cases, appear capable of partially decoupling the canonical trade‐off between dough strength and extensibility. Li et al. (2015) showed that substituting glycine with glutamic acid at position 244 in the CRD (Gly244Glu‐CRD) of the Ax1 subunit decreased SDS–PAGE mobility and significantly enhanced dough strength while maintaining extensibility. Additional evidence underscores that substitutions outside the CRD can also be consequential. Wang et al. (2021, 2026) found that a Ser‐to‐Asp substitution at position 21 in the C‐terminal domain (Ser21Asn‐C) of Dy10 permits vacuolar processing enzymes to act on the posttranslational product, generating two distinct polypeptide products and ultimately reducing dough strength. This mechanism highlights that primary‐structure variation can influence dough quality not only through polymer physics, but also through protein processing and maturation pathways that alter the effective subunit composition incorporated into the gluten network.

Overall, substantial datasets describing HMW‐GS primary‐structure polymorphisms—encompassing substitutions, insertions, and deletions—have accumulated. However, much of the mechanistic literature remains focused on Cys residues and disulfide crosslinking. The functional consequences of the broader landscape of non‐Cys variants, and their genotype‐to‐phenotype mappings across different genetic backgrounds and processing contexts, have yet to be systematically characterized. Developing such a comprehensive framework will be essential for translating sequence diversity into predictive rules for dough rheology and for enabling more precise quality‐oriented processing and protein engineering.

4.2. Secondary Structure: Dynamic Equilibrium Between Folds and Turns and Its Contribution to Strength and Extensibility

Protein secondary structure refers to characteristic conformations formed by the polypeptide backbone as it coils or folds around a defined axis, including α‐helices, β‐sheets, β‐turns, and random coils; these conformations are described primarily by backbone geometry rather than side‐chain effects. Early analyses of chemically extracted HMW‐GS mixtures from wheat kernels using circular dichroism (CD) and infrared spectroscopy (IR) showed pronounced β‐turn signatures in the far‐UV region, whereas clear features characteristic of α‐helices or β‐sheets were not observed. In contrast to these experimental observations on extracted mixtures, structure‐prediction analyses have suggested that the N‐ and C‐terminal domains of HMW‐GS are relatively enriched in α‐helical elements (Tatham et al. 1984, 1985; Yang, Wang, et al. 2023). This apparent discrepancy is not necessarily contradictory, because different domains may exhibit distinct structural propensities; terminal regions are more conserved and compositionally diverse, whereas central repetitive regions are enriched in turn‐favoring residues and may dominate the overall spectroscopic signal. It should be noted, however, that HMW‐GS predominantly exist as polymers in the wheat endosperm and are difficult to isolate in their native form; therefore, chemical reagents used during extraction (e.g., acetic acid and β‐mercaptoethanol) may partially disrupt original conformations. Therefore, structural conclusions derived from extracted proteins should be interpreted with caution, particularly when extrapolating to the native polymeric state in situ.

Van Dijk et al. (1998) expressed the N‐terminal and CRD of the Dx5 subunit in Escherichia coli and, using CD and IR, showed that the N‐terminal domain in solution consists mainly of α‐helix (26%–35%) and β‐sheet (33%–36%), whereas the CRD is dominated by consecutive β‐turn structures (Van Dijk, De Boef, et al. 1997). This observation is consistent with both the findings obtained from chemically extracted mixtures of HMW‐GS from wheat kernels and the results of structure‐prediction analyses. The authors further expressed cyclic and linear repetitive polypeptides (e.g., the hexapeptide PGQGQQ and the nonapeptide GYYPTSPQQ), demonstrating that β‐turns preferentially form at motifs such as QPGQ, YPTS, SPQQ, and QQGY, thereby clarifying the primary‐structure basis for β‐turn formation in the central repetitive region (Van Dijk, Van Wijk, et al. 1997). A defining feature of a β‐turn is the hydrogen‐bond interaction between the first and fourth residues; in the HMW‐GS repetitive region, hydrogen bonding may occur between Q–Q, Y–S, S–Q, or Q–Y pairs. Moreover, the side‐chain functional groups of these residues may also participate in hydrogen bonding, potentially playing an important role in stabilizing HMW‐GS super secondary or tertiary structures.

Primary‐structure polymorphism in HMW‐GS is expected to entail concomitant variation in secondary‐structure propensity. At present, owing to the intrinsic difficulty of isolating HMW‐GS in their native polymeric state, researchers largely rely on computational secondary‐structure prediction to infer domain‐ and subunit‐level conformational profiles. For example, Peng et al. (2015) predicted that Dy10, Dy12, Dy12.6, and Dy12.7 comprise five major secondary‐structure categories—α‐helix, β‐turn, β‐sheet, random coil, and β‐helix—and further demonstrated substantial differences in the relative abundance of these elements among the four subunits. Likewise, Pang and Zhang (2008) observed pronounced differences in α‐helical content among Bx13 + By16, Bx14 + By15, Bx7 + By8, Bx7 + By9, and Bx20 + By20; notably, By16 contains seven α‐helical segments, whereas Bx7 contains only five. More recently, Yang, Wang, et al. (2023) provided comparative evidence that x‐type subunits (e.g., Ax1, Bx7, and Dx2) tend to exhibit higher predicted contents of β‐sheets, β‐turns, and random coils than y‐type subunits (e.g., By8 and Dy12). Because β‐turn–rich and coil‐like conformations are often associated with conformational flexibility and increased opportunities for intermolecular contacts, these structural tendencies were interpreted as being more favorable for protein–protein interactions and network formation. Consistent with this interpretation, experimental observations in the same line of work indicated that x‐type subunits generally contribute more strongly to dough strength than y‐type subunits.

Beyond prediction‐based analyses, limited experimental evidence has also begun to connect HMW‐GS composition with gluten secondary structure. Using HMW‐GS NILs, Gao et al. (2016) applied CD spectroscopy to quantify the secondary‐structure distribution in gluten proteins, reporting variation ranges of 48.29%–60.41% for α‐helix, 3.36%–4.87% for intermolecular β‐sheet, 13.71%–24.01% for β‐sheet, and 12.88%–13.44% for β‐turn. Correlation analysis further showed that β‐turn content was significantly positively associated with dough extensibility, whereas β‐sheet content correlated positively with dough strength. These results offer preliminary support for a regulatory mechanism in which interconversion between β‐sheets and β‐turns shifts the balance between strength and extensibility. Nonetheless, current structure–function inferences remain based on relatively limited NIL panels; broader validation across diverse germplasm sets and standardized rheological conditions is still needed to establish robust, generalizable rules.

4.3. Tertiary Structure: Characteristics and Polymorphism and Their Functional Analysis

Protein tertiary structure refers to the three‐dimensional conformation formed when a single polypeptide chain undergoes further folding on the basis of secondary structure through non‐covalent interactions (e.g., hydrophobic interactions, hydrogen bonds, and van der Waals forces), encompassing the spatial arrangement of all atoms and the hierarchical organization of domains. Field et al. (1987) first isolated a single HMW‐GS from the durum wheat cultivar Bidi17 and, based on hydrodynamic measurements, showed that HMW‐GS adopts a rod‐like conformation in 50% 1‐propanol and in trifluoroethanol solutions; CD data suggested that β‐turns in the central repetitive region form a loose helical structure. Miles et al. (1990) likewise observed helical features of HMW‐GS using scanning tunneling microscopy. Integrating evidence from multiple spectroscopic and hydrodynamic approaches, Shewry et al. (2000) proposed a general tertiary‐structure model in which the N‐ and C‐terminal regions are α‐helix‐rich and globular, whereas the central repetitive region forms a spring‐like β‐helix (Figure 5). These studies outline the overall tertiary architecture of HMW‐GS; however, the detailed internal mechanisms—such as how Q/S/Y/T‐rich side chains participate in hydrogen‐bond networks and stabilize local conformations—remain poorly resolved and still require higher resolution structural characterization.

Lafiandra et al. (1999), using transverse urea‐gradient gel electrophoresis, demonstrated marked differences in unfolding behavior between x‐ and y‐type subunits, indicating differences in tertiary structure. Specifically, x‐type subunits exhibited broad, continuous unfolding profiles, whereas most y‐type subunits showed a distinct single inflection point coincident with a sharp mobility decrease. Moreover, urea‐denaturation inflection‐point concentrations differed significantly among y‐type subunits, reflecting differences in structural stability. For example, at the Glu‐B1 locus, stability followed the order By16 > By8/By18 > By15, whereas at the Glu‐D1 locus, the order was Dy10 > Dy12 > Dy12*. Because HMW‐GS are difficult to extract and intrinsically complex, defining their refined higher order structures has long been challenging. With advances in computational power and software ecosystems, molecular dynamics (MD) simulations have become increasingly advantageous for interpreting macroscopic food‐property changes from a molecular perspective. Using MD, Yang, Ge, et al. (2023) analyzed the tertiary structures of five HMW‐GS subunits (Ax1, Bx7, By8, Dx2, and Dy12) and obtained four key parameters: root‐mean‐square fluctuation (RMSF), radius of gyration (Rg), solvent‐accessible surface area (SASA), and the number of hydrogen bonds. RMSF, reflecting conformational flexibility, indicated that Ax1, Bx7, By8, and Dx2 are overall more flexible than Dy12. Rg, reflecting compactness, ranked as Dy12 < Ax1 < Bx7 < By8< Dx2 and did not follow molecular‐weight order. SASA, reflecting solvent‐accessible protein surface area, ranked as Ax1 < Dy12 < By8 < Dx7 < Dx2. The hydrogen‐bond count ranged from 380 to 500, with Bx7 exhibiting the highest number and By8 the lowest. Collectively, these results provide a theoretical framework for further interpreting variation in dough quality.

4.4. Network Structure: Synergistic Effects of Disulfide Crosslinking and Non‐Covalent Interactions

Connectivity of HMW‐GS within the gluten network primarily depends on disulfide‐bond interactions. To date, only three inter‐chain disulfide‐linkage patterns have been identified as follows: (1) between a C‐terminal cysteine of an x‐type HMW‐GS and an N‐terminal cysteine of a y‐type HMW‐GS (Lutz et al. 2012; Tao et al. 1992); (2) between cysteines of two x‐type HMW‐GS subunits (Werner et al. 1992); and (3) between N‐terminal cysteines of two y‐type HMW‐GS subunits (Köhler et al. 1991). Considering the composition and typical x‐/y‐type ratio in native flour (x‐/y‐ ≈ 2.5:1), an updated structural model proposes that the most basic molecular unit of HMW‐GS is a low‐oligomer consisting of two y‐type and four x‐type subunits covalently connected via inter‐chain disulfide bonds (Wieser 2007) (Figure 5). Additional low‐abundance covalent interactions, such as dityrosine crosslinks, have also been reported (Tilley et al. 2001). However, Hanft and Koehler (2005) questioned the significance of this mechanism, noting that total Tyr content in flour is approximately 18.6 mmol/kg and that at most ∼0.1% of Tyr residues participate in dityrosine crosslinking in dough. Thus, in terms of both abundance and contribution, disulfide crosslinking is far more critical than tyrosine‐based crosslinking for gluten‐network formation.

Although non‐covalent interactions (hydrogen bonding, ionic interactions, and hydrophobic interactions) are weaker than covalent bonds, they are likewise indispensable for glutenins aggregation and gluten‐network development. Evidence for hydrogen bonding includes the observation that hydrogen‐bond disruptors weaken gluten‐network structure, whereas replacing water with heavy water (D₂O) as solvent strengthens the network (Shewry, Halford, et al. 2003). The importance of ionic interactions is supported by the strengthening effects of NaCl or zwitterions on the gluten network (Abedi et al. 2018; Yang et al. 2025). Using a prokaryotic expression system, Wang et al. (2015) prepared the N‐terminal domain of Dx5 and showed that its aggregation in aqueous solution depends on hydrophobic interactions; the addition of this domain to flour further indicated that hydrophobic interactions are a key driving force for glutenins aggregation (Wang et al. 2016). Importantly, aggregates formed via hydrophobic interactions can reduce the spatial distance between specific residues (e.g., cysteine residues), thereby facilitating formation of intermolecular covalent bonds, such as disulfide linkages (Shewry et al. 2000; Wang et al. 2015).

4.5. Microstructure of Dough and Gluten

HMW‐GS is a key protein component determining the microstructure of gluten and dough, and variations in its composition, allelic forms, or the absence of specific subunits can substantially alter the spatial architecture, connectivity, and stability of the gluten network. Evidence from SEM and CLSM studies consistently shows that genotypes carrying superior HMW‐GS alleles or introgressed high‐quality exogenous subunits generally form gluten networks with smaller and more appropriately distributed pores, tighter cross‐linking, and longer, more continuous protein strands (Li et al. 2023; Liu et al. 2016). In contrast, inferior HMW‐GS alleles, as well as the deletion or silencing of HMW‐GS, often result in looser network organization, enlarged or unevenly distributed pores, discontinuous protein chains, and fewer junctions, thereby weakening the cross‐linking capacity and structural stability of gluten macropolymers (Wang et al. 2024; Gao et al. 2016). Quantitative CLSM analyses further demonstrate that materials with superior HMW‐GS typically exhibit higher protein area, protein percentage area, junction number, junction density, total protein length, and branching rate, whereas endpoint number and indices of network irregularity are generally lower, indicating greater connectivity, compactness, and cohesiveness of the gluten network (Li et al. 2023; Zhao, Li et al. 2020). These microstructural differences are closely associated with the intrinsic structural features of HMW‐GS molecules, particularly the number and distribution of cysteine residues, their capacity to form disulfide bonds, and their β‐sheet content, all of which determine whether HMW‐GS preferentially acts as a “chain extender” or a “chain brancher” within the network. This, in turn, influences whether the gluten network develops into a more linear fibrous structure or into a highly cross‐linked lamellar/compact network (Gao et al. 2016). At the same time, the gluten network serves as the three‐dimensional framework that entraps starch granules, and variations in pore size and pore distribution can further affect the embedding state and spatial distribution of starch granules and other components within the dough matrix, thereby ultimately influencing dough rheological properties (Gao et al. 2018).

4.6. Structural Models: The Loop‐Train Model and Elastic Structural Units as Explanations for Rheological Differences in Dough

Primary structure determines multidimensional higher order structures (including secondary structure, tertiary structure, and spatial organization within the gluten network), and the multidimensional architecture of HMW‐GS provides a molecular‐level explanation for the viscoelasticity of the gluten network. Belton and colleagues proposed the “loop‐train” model to explain the molecular basis of gluten elasticity. This model posits that, under low‐hydration conditions, glutamine residues form dense interprotein interactions through inter‐chain hydrogen bonds (the “train” regions). As hydration increases, some inter‐chain hydrogen bonds dissociate, and glutamine residues instead hydrogen‐bond with water molecules, resulting in “loop” regions within the protein conformation (Belton 1999). Because gluten proteins are rich in glutamine, the probability that all inter‐chain hydrogen bonds break simultaneously is extremely low; therefore, the proteins do not fully dissolve. Instead, the system reaches a dynamic equilibrium in which hydrated “loop” regions coexist with hydrogen‐bonded “train” regions. This loop‐train dynamic equilibrium provides a conceptual framework for understanding both strength and extensibility of gluten. Upon stretching of the gluten network, “loop” regions extend first, followed by dissociation of “train” regions; this conformational transition enables chain extension and provides a structural basis for elastic energy storage associated with strength. By inference, “train” regions are more likely associated with β‐sheet structures, whereas “loop” regions correspond to β‐turns. IR analyses of HMW‐GS under different hydration states support this inference: upon hydration from the dry state, β‐sheet content first increases and then decreases, with the initial increase attributed to ordering during the glass‐to‐rubber transition; with further hydration, β‐sheet content gradually decreases, whereas β‐turn content increases markedly (Belton et al. 1995).

The rod‐like tertiary conformation of HMW‐GS, approximately 49 nm × 1.8 nm in 50% 1‐propanol and more than 60 nm long in trifluoroethanol, confers polymer‐like segmental flexibility and extensibility (Field et al. 1987). Small‐angle neutron scattering further demonstrated that the CRD of HMW‐GS can elongate, behaving like a “flexible spring” (Van Swieten et al. 2003).

At the network level, HMW‐GS crosslink elastic loop‐train regions and extensible helical regions into a three‐dimensional gluten network via disulfide bonds, thereby imparting pronounced viscoelasticity to dough. Disulfide bonds are rigid covalent linkages and do not directly contribute to reversible elasticity; rather, they serve as network “nodes” that connect elastic structural units (e.g., β‐helices and loop‐train regions), enabling energy storage and release during deformation‐recovery cycles at the macroscopic level. By contrast, gliadins may contain a certain proportion of helical structures but lack effective inter‐chain disulfide crosslinking capacity and therefore cannot construct a stable elastic network (Tatham et al. 1985). This comparison underscores the indispensable role of disulfide bonds in establishing and maintaining an elastic gluten network.

From the general perspective of elastomeric materials, mechanical behavior can be characterized along two principal dimensions: strength and extensibility. Available evidence indicates that distinct structural units contribute differentially to these two properties within an elastic gluten network. “Loop” structures primarily contribute to extensibility, supported by the significant positive correlation between β‐turn content in gluten proteins from HMW‐GS NILs and dough extensibility (Belton 1999; Gao et al. 2016). In contrast, “train” structures and disulfide bonds primarily contribute to strength, as evidenced by significant positive correlations between β‐sheet content and disulfide‐bond content with dough strength (Gao et al. 2016; Wei et al. 2025). The contribution of helical structures to strength versus extensibility depends on the degree of conformational looseness of the helix. Ultimately, the abundance and proportion of these structural units determine the balance between dough strength and extensibility (Figure 5).

We systematically synthesized current knowledge on the multidimensional architecture of HMW‐GS and its intrinsic links to dough rheological properties, and we propose a preliminary structure–function framework spanning “primary sequence → secondary/tertiary structure → HMW‐GS network organization → macroscopic viscoelasticity” (Figure 5). Dough strength is governed primarily by the density of the disulfide crosslinking network and the stability of β‐sheet‐rich “train” segments, whereas extensibility depends more on the deformability of β‐turn‐dominated “loop” regions and the flexibility of helical structures. Single‐amino‐acid substitutions at key sites can fine‐tune higher order structure by altering local conformations or intermolecular interactions, thereby modulating dough rheology. Nevertheless, the fine high‐order structural features of HMW‐GS remain poorly resolved, particularly in Q/S/Y/T‐enriched regions and in local conformations within the CRD. A clear gap persists in quantitatively mapping molecular‐scale structural evolution to macroscopic rheological responses. Future work should integrate advanced structural characterization (e.g., cryo‐electron microscopy, small‐angle scattering, and solid‐state NMR) with computational approaches, such as molecular dynamics simulations, to overcome technical bottlenecks and enable quantitative and visual elucidation of HMW‐GS structure–function relationships.

5. Conclusions

This article systematically reviews the dose‐effect contributions of allelic variation in HMW‐GSs to dough strength and extensibility, together with the underlying structure–function mechanisms. In terms of phenotypic contributions, subunits, such as Ax1, Ax2*, Bx14 + By15, and Bx17 + By18, are expected to synergistically enhance both dough strength and extensibility, whereas Bx7^OE + By8 and Dx5 + Dy10 typically enhance dough strength at the expense of extensibility. Beyond Ax alleles, reactivation of the normally silent Ay locus represents a promising and underutilized strategy for quality improvement. Introgression of the expressed 1Ay21* allele has been shown to significantly increase grain protein content, UPP%, and bread volume without yield penalty, highlighting the considerable but underexploited potential of Ay subunits in wheat quality breeding. From a structure–function perspective, a preliminary multidimensional framework is proposed for HMW‐GSs: “loop‐train” motifs and helical structures impart molecular elasticity; disulfide bonds crosslink HMW‐GSs containing these elastic structural units into an integrated network; and non‐covalent interactions, including hydrogen bonding, hydrophobic interactions, and ionic interactions, synergistically drive the network assembly and stabilization. Dough strength is mainly governed by the density of the disulfide‐crosslinked network and the stability of β‐sheet chain structures, whereas extensibility depends more on the deformability of β‐turn loop regions and the flexibility of helical structures. Single‐amino‐acid substitutions at key sites can modulate dough properties by altering local conformations or intermolecular interactions. Specifically, Cys10‐N, Cys40‐N, Cys8‐CRD, Cys612‐CRD, and Cys30‐C favor inter‐chain disulfide bond formation, whereas Cys25‐N preferentially forms intrachain disulfide bonds and represents a critical site for regulating gluten network strength. Mutations at other key residues can likewise tune dough strength and extensibility, as exemplified by the Gly244 to Glu substitution within the CRD of the Ax1 subunit.

Overall, studies linking HMW‐GS structure polymorphisms to dough quality remain relatively limited. Elucidation of secondary‐ and tertiary‐structure polymorphisms still relies largely on computational prediction, and high‐resolution experimental evidence for fine tertiary architectures is lacking. Functional analyses have also been disproportionately focused on cysteine residues in x‐type subunits, with insufficient attention paid to cysteine residues in y‐type subunits and to the structural and functional impacts of other amino‐acid substitutions, deletions, and insertions. Although modern genomics and proteomics have generated extensive datasets of gluten protein sequences and polymorphisms, the intrinsic mechanisms by which HMW‐GS structural features and their variation influence dough rheology and end‐product quality remain incompletely understood.

6. Future Perspectives

To address the aforementioned research gaps, future studies should further strengthen the integration of wheat breeding, food science, computational biology, synthetic biology, and artificial intelligence. Specifically, valuable genetic materials developed by breeders, such as NILs, together with targeted improved materials generated through gene editing and synthetic biology strategies, can be combined with next‐generation high‐resolution structural characterization techniques, including advanced cryo‐electron microscopy, ultra‐high‐field nuclear magnetic resonance, cross‐linking mass spectrometry, and in situ characterization methods, to enable a more refined dissection of the structural features and assembly mechanisms of HMW‐GS. On this basis, bioinformatics, machine learning, and molecular dynamics simulations should be further integrated to systematically mine polymorphism information from the primary sequences of HMW‐GS and related gluten proteins and to establish multiscale association models linking sequence, structure, molecular interactions, dough rheological behavior, and end‐product quality. The accuracy and interpretability of these models should then be continuously improved through experimental validation. In addition, AI‐guided molecular design strategies can be deeply integrated with the design‐build‐test‐learn closed loop in synthetic biology to enable the rational modification and targeted optimization of key amino‐acid residues, the length and composition of repetitive domains, cysteine distribution patterns, and inter‐subunit interaction modes. At the same time, future research should not be limited to describing the static structural features of HMW‐GS, nor should it focus exclusively on HMW‐GS itself; rather, greater attention should be paid to analyzing the dynamic behavior of polymeric glutenins during processing, while systematically considering the roles of LMW‐GS and gliadins in gluten network formation, rearrangement, stabilization, and final product quality. Such efforts will facilitate the predictable design and precise regulation of gluten protein functionality, ultimately providing a stronger theoretical basis and technical support for the targeted improvement of quality traits across diverse wheat‐based food systems and industrial processing scenarios.

Author Contributions

Hongwei Zhou: formal analysis, methodology, software, validation, visualization, writing – original draft. Yimin Wei: writing – review and editing. Xiaolong Wang: writing – review and editing. Boli Guo: funding acquisition, resources, supervision. Yingquan Zhang: conceptualization, funding acquisition, resources, project administration, supervision, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Material: crf370585‐sup‐0001‐TableS1‐S3.docx