Energy Metabolism and Auxin Signaling Disruption Underlying Stamen Identity Defects in Tobacco Cytoplasmic Male Sterility K326 (CMS K326): Integrated Transcriptomic and Metabolomic Analyses

Abstract

Cytoplasmic male sterility (CMS) provides a natural model for studying nuclear–cytoplasmic interactions, although the details of nuclear–cytoplasmic communication remain poorly understood. In this study, transcriptomic and metabolomic data were integrated to elucidate the molecular and metabolic regulatory networks underlying stamen developmental defects in the tobacco CMS K326 (Nicotiana tabacum). Disrupted energy metabolism, auxin pathways, and floral development gene expression were identified in CMS K326. Metabolites such as glucose-6-phosphate, fructose-6-phosphate, and oxalosuccinic acid decreased, while an accumulation of succinate was observed and auxin IAA was deficient. Our study revealed that disrupted nuclear–cytoplasmic interactions in CMS K326 are associated with concurrent disruptions in early auxin homeostasis and energy metabolism, which collectively lead to the disturbance of the stamen development program. This study provides multiomics-level evidence for understanding stamen identity defects in CMS.

1. Introduction

Cytoplasmic male sterility (CMS) is a phenomenon in which plants are unable to produce functional pollen. It is widely observed in higher plants and exhibits maternal inheritance, resulting from disrupted nuclear–cytoplasmic interactions. This unique genetic mechanism makes CMS not only a valuable tool for heterosis utilization in many crops but also an ideal model for studying nuclear–cytoplasmic interactions [1]. In addition to pollen abortion, CMS is often accompanied by abnormal stamen development, including a range of morphological variations such as the absence of stamens, absence of anthers, filament degeneration, and carpelloid or petaloid transformation of stamens [1]. The carpelloid or petaloid transformation of CMS stamens resembles homeotic mutations and has been termed cytoplasmic homeosis [2].

Mitochondrial genome rearrangement is a common feature of CMS and frequently generates novel open reading frames (ORFs) or chimeric ORFs fused with functional genes [1,2]. The maternal inheritance pattern readily links such changes to the cause of CMS, an association that has been confirmed by CMS research in several crops [3,4,5,6,7,8]. It has also been demonstrated that nuclear-encoded restorer genes can suppress or restore CMS through multiple mechanisms. For instance, pentatricopeptide repeat (PPR) proteins are known to inhibit the expression of CMS-associated genes via post-transcriptional regulation [3]. However, not all novel or chimeric ORFs are directly linked to CMS; many are thought to interfere with mitochondrial function [9,10].

The phenotypic manifestations of CMS suggest complexities in its developmental basis. Pollen abortion CMS phenotypes typically arise from defects during later stages of microspore development, whereas CMS with carpelloid, petaloid, or sepaloid stamens reflects early defects in floral organ identity determination [5,11,12]. It is not known whether the distinct CMS phenotypes (such as late pollen abortion versus early homeotic transformation) merely represent varying degrees of disruption along a common developmental pathway or arise from fundamentally independent mechanisms. Furthermore, it is uncertain whether different types of CMS can be suppressed by common nuclear restorer genes or require distinct sets of restorers. Nevertheless, several facts are well-established in CMS research: first, maternal inheritance confirms that mitochondrial gene function is impaired; second, stamen development is disrupted. According to the floral ABC model, alterations in CMS stamens are associated with B genes. In cytoplasmic homeosis CMS, downregulation of B and C genes, as well as SUPERMAN (SUP) expression, has been detected [2], often accompanied by a loss of the boundary between stamens and carpels. Studies on SUP suggest that stamen defects result from excessive proliferation of stem cells in fourth whorl [13]. SUP regulated the boundary between the third and fourth whorls by controlling auxin biosynthesis [14]. Furthermore, fertility and stamen phenotype were restored in tobacco CMS Nta(rep) plants overexpressing the Arabidopsis SUP gene, which partially yield functional pollen [15]. However, in all CMS cases with altered stamen morphology, the positional identity of stamens within the floral whorls remains unchanged [16], indicating that floral organ identity genes are perturbed while broader floral developmental patterning is less affected [1]. Therefore, stamen abnormalities in CMS can be attributed to the failure of stamen primordium stem cells to properly initiate organogenesis and differentiation, coupled with delayed termination of carpel primordium stem cells [15]. Moreover, given the central role of auxin in establishing and developing floral organ meristems [17], it is considered to significantly influence the formation of abnormal stamens in CMS. Nevertheless, the detailed mechanisms of communication between the nucleus and cytoplasm remain unclear.

Tobacco (Nicotiana tabacum) is an important global economic crop. From a naturally occurring mutant, we developed the cytoplasmic male sterile line CMS K326, which exhibits a series of stamen homeotic transformations, including degeneration, absence, carpelloid, and petaloid traits. In our previous multi-omics study [12], we established that these defects initiate at the early bud stage and are transcriptionally linked to the coordinated dysregulation of floral organ identity genes (e.g., WUS, GLO, and SUP) and auxin-signaling pathways. Furthermore, co-expression network analysis identified a key module correlating stamen development with auxin response and mitochondrial retrograde regulation, strongly suggesting a central, yet correlative, role for auxin imbalance in the CMS phenotype. Building directly on these findings, the present study was designed to move from correlation to mechanistic insight. To achieve this, we integrated metabolomic profiling with transcriptomic data from the same floral tissues. We specifically focused on analyzing expression patterns of genes governing energy metabolism, auxin pathways, and retrograde signaling while directly quantifying the corresponding metabolic intermediates and auxin levels. Through this combined approach, this study confirms and extends the intrinsic link between perturbed energy metabolism, auxin homeostasis, and stamen developmental defects in CMS K326, providing a more comprehensive, multi-omics perspective on the potential mechanisms underlying this CMS phenotype.

2. Results

2.1. Identification of CMS-Associated Co-Expression Modules Using Weighted Gene Correlation Network Analysis (WGCNA)

The selection of an appropriate soft-thresholding power is crucial for constructing a biologically meaningful scale-free co-expression network. After systematically evaluating powers from 1 to 30, we chose a power of 10 by balancing the scale-free topology fit with network interpretability (Figure 1). Although a power of 20 yielded a higher scale-free fit index, the power of 10 was selected because it generated the same number of co-expression modules and, more importantly, avoided assigning an excessive number of known phenotype-associated genes to the biologically uninformative grey module (Figure 1A).

Using the selected power, a weighted gene co-expression network was constructed, identifying eight distinct modules derived from 1590 genes (Figure 1B). Pearson correlation analysis between module eigengenes and CMS phenotypic traits revealed that the black, greenyellow, and turquoise modules showed the most significant associations (Figure 1C). To identify key regulatory genes, Gene Significance (GS) and Module Membership (MM) were integrated to pinpoint highly connected hub genes within these modules. The turquoise module (correlation = −0.793) contained four hub genes (TIM14-3, MTHFR, NAC078, and JOINTLESS) with GS < −0.84 and MM > 0.94, indicating a strong negative association with CMS and suggesting their role as candidate suppressors (Figure 1D). The greenyellow module (correlation = −0.511) featured two hub genes (WUS and IAA6B-Like) with moderate GS values (≈−0.63) but high connectivity (MM > 0.96) (Figure 1E). The black module (correlation = 0.834) included five hub genes (WOX9, ARF8, ARF6, PIN1c, and MADS1) exhibiting high GS (>0.88) and MM (>0.93), highlighting their potential role as promoters of CMS (Figure 1F).

2.2. Functional Analysis of Key CMS-Associated Modules

Given the maternal inheritance of CMS and its characteristic disruption of stamen development, together with the established role of auxin in floral organogenesis, genes from key co-expression modules were classified into four functional categories: (1) energy metabolism, (2) mitochondrial biology, (3) floral development, and (4) auxin-related processes. Previously reported CMS-associated chaperones and transcription factors were also considered.

The turquoise module contained 78 genes (Figure 2A), with functional annotation showing enrichment in energy metabolism (29.5%, e.g., COX2, PDK, and IDH) and auxin-related processes (26.9%, e.g., PIN3, and ARF3), indicating a primary role in maintaining cellular and mitochondrial homeostasis under stress. The greenyellow module comprised 38 genes (Figure 2B), dominated by energy metabolism (39.5%, e.g., ATPB, and COX5C) and auxin-related functions (23.7%, e.g., IAA6B-like, and LAX2), alongside a notable proportion of floral-development genes (21.1%, e.g., WUS, and GLO), reflecting a specialized functional profile. The black module included 120 genes (Figure 2C) and displayed a balanced yet coordinated enrichment in mitochondrial biology (28.3%, e.g., TIM17-2, and TOM20-like) and auxin-related processes (25.0%, e.g., ARF6, ARF8, and PIN1c), suggesting its role in mediating crosstalk between mitochondrial function and auxin signaling within a network that also integrates floral development and energy metabolism.

In summary, the turquoise and greenyellow modules—both negatively correlated with CMS—are functionally enriched in cellular homeostasis and floral organ development, respectively. In contrast, the black module—positively correlated with CMS—shows coordinated enrichment for mitochondrial processes, auxin signaling, and floral development pathways. This functional divergence suggests that CMS may originate from dysregulation within core networks that bridge energy metabolism, hormone signaling, and developmental programming.

Since WGCNA is not limited to hub genes identified by a fixed threshold, we further analyzed the differentially expressed genes (DEGs) within the CMS-associated modules to better evaluate the functional relevance of individual genes. Given the influence of nucleo-cytoplasmic interactions, DEGs at the small bud stage are likely more relevant to the causal mechanisms.

Under the cutoff of |log₂FC| > 1, 41 out of 78 genes (52.56%) in the turquoise module are significantly differentially expressed at the small bud stage (Figure 3A). Among these, four genes are upregulated, including HXK2 and NAC089, while 37 are downregulated, such as JOINTLESS, TIM14-3, and NAC078. However, MTHFR, identified as the hub gene of this module by WGCNA, is not significantly differentially expressed. Notably, one member of NAC078, together with AGL82 and IAA10A5, exhibits consistently high expression across all developmental stages in the maintainer line but is nearly silenced (<10 expression value in all samples) in the CMS K326. This indicates severe suppression of these three genes in CMS.

In the greenyellow module, 6 out of 38 genes (15.79%) are significantly differentially expressed at the small bud stage (Figure 3B). Among them, two genes—COX5C and IAA5—are upregulated, and four, including NDB2, MCU4, WUS, and WOX, are downregulated. The hub gene IAA6B-like, identified by WGCNA, is not significantly differentially expressed. Notably, WUS shows high expression in the maintainer line but is nearly silenced in CMS K326, implying substantial disruption of its expression.

In the black module, 19 of 120 genes (15.83%) are significantly differentially expressed at the small bud stage (Figure 3C). Among these, 17 are upregulated—including members of ATPB, PINc, and SAUR, as well as WOX8, WOX9, CRC, AP2, and MADS1—while only two BiP5 genes are downregulated. The WGCNA-identified hub genes ARF8 and ARF6 are not significantly differentially expressed. Notably, two ATPB genes stand out: they are barely expressed in the maintainer line across all stages but are highly expressed in the CMS line, indicating strong transcriptional activation in the CMS background.

In summary, these switch-like expression changes—such as ATPB turning from off to on and WUS from on to off—strongly suggest that these genes are key candidates responsible for the functional disruption of their respective modules and may ultimately contribute to the CMS phenotype. The qRT-PCR primer sequences and corresponding results are provided in Table S1 and Figure S1 of the Supplementary Materials.

2.3. Metabolic and Hormonal Dynamics During Floral Bud Development in CMS

To validate the relationship between gene expression and metabolic changes, we conducted non-targeted metabolomic analysis on the same batch of materials used for transcriptome sequencing. Given the close association between early metabolic dynamics and CMS, we focused on metabolite and hormone profiles at the small bud stage. Applying thresholds of P_adj < 0.05, VIP > 1, and |log₂FC| > 1, we identified 147 and 137 significantly altered metabolites in positive and negative ion modes, respectively. Based on the magnitude of change and biological relevance, 14 metabolites and one hormone were prioritized for detailed analysis (Figure 4).

At the small bud stage, indoleacetic acid (IAA) was significantly decreased (log₂FC = −1.18). As the primary auxin, its severe deficiency at this critical phase for floral organ development may directly contribute to subsequent developmental defects. Key glycolytic intermediates, including glucose-6-phosphate (G6P) (also an entry point for the pentose phosphate pathway) and fructose-6-phosphate (F6P), were all significantly decreased. This concerted suppression strongly suggests a substantial impairment in early carbohydrate catabolism and ATP production in CMS flower buds. Furthermore, the decrease in glycerol-3-phosphate (G3P), a central component in glycerophospholipid synthesis and the glycerol phosphate shuttle, implies potential disruptions in lipid metabolism and cellular energy balance. Several TCA cycle-related metabolites were also reduced. Oxalosuccinate (log₂FC = −1.20), an intermediate in the conversion of isocitrate to α-ketoglutarate, was reduced, indicating a disruption in the TCA cycle flux. The reduction of 2-methyl-cis-aconitate, an intermediate in branched-chain amino acid metabolism, and 4-hydroxy-2-oxoglutarate, which can feed into the TCA and glyoxylate cycles, further supports a broad suppression of central organic acid metabolism at this early stage. In contrast, succinate was significantly increased (log₂FC = 1.20) at the small bud stage, with a slight increase persisting in later stages. This accumulation could result from insufficient upstream substrate supply (e.g., α-ketoglutarate) or impaired activity of succinate dehydrogenase (SDH), a key enzyme linking the TCA cycle to the mitochondrial electron transport chain (ETC). SDH dysfunction would directly compromise ATP production efficiency. Panthenol, a precursor of the central metabolic cofactor coenzyme A (CoA), was significantly decreased, likely impairing CoA biosynthesis and its dependent processes such as the TCA cycle and fatty acid oxidation. Finally, the decrease in uracil (an RNA building block) and inosine (a purine metabolism intermediate) suggests possible constraints on nucleic acid synthesis, which could directly impact the gene expression and active cell division required during early bud development.

Overall, metabolic differences between the CMS and its maintainer were most pronounced at the small bud stage (<3 mm), where nearly all detected metabolites were substantially decreased. As development progressed to the medium (3–5 mm) and large (>5 mm) bud stages, differences in many metabolites reduced or even reversed (see Figure 4). This dynamic pattern indicates that the metabolic impact of CMS is most severe during the initial phase of floral organ development and gradually attenuates thereafter.

2.4. Mutual Corroboration of Gene Expression and Metabolite Data

Based on an integrated analysis of transcriptomic and metabolomic data, the mutually corroborative evidence for the core disruptions in energy metabolism and auxin signaling in CMS K326 floral buds is presented.

2.4.1. Mutual Validation of Energy Metabolism Disorder

Transcriptomic and metabolomic data jointly depict a systematic impairment of energy metabolism spanning from glycolysis to the TCA cycle. At the gene level, the upregulated expression of HXK2, responsible for phosphorylating glucose to initiate glycolysis, suggests an attempted cellular compensation to enhance glycolytic flux. However, this is contradicted by the concurrent downregulation of GAPDH, a gene encoding a key downstream glycolytic enzyme, indicating a probable bottleneck in the later steps of the pathway. In perfect agreement, metabolomic profiling revealed a significant decrease in the abundance of early glycolytic intermediates, including G6P, F6P, and G3P. This depletion directly reflects the consumption of substrates due to enhanced HXK2 activity and, more importantly, the failure of these intermediates to proceed through downstream reactions efficiently. Furthermore, the significant accumulation of the TCA cycle intermediate succinate provides direct metabolic evidence for a blockage at the downstream steps of the cycle, likely at the succinate dehydrogenase complex or in the ETC. This metabolic perturbation aligns with the observed dysregulation of gene sets related to mitochondrial function and oxidative phosphorylation in the transcriptome (e.g., the CMS-positively correlated black module). Thus, evidence from both gene expression changes and metabolite abundance shifts consistently confirms the existence of a continuous metabolic bottleneck from glycolysis to oxidative phosphorylation in CMS floral buds.

2.4.2. Coordinated Disruption of the Auxin Signaling Pathway

Integrated analysis of metabolomic and transcriptomic data reveals a complex dysfunction in the auxin pathway, characterized by both a ligand deficiency and a breakdown in signal feedback regulation. A significant reduction in the level of IAA occurs during the early bud stage. Transcriptomics further elucidated the anomalous molecular response to this deficiency: under low-IAA conditions, the canonical early-responsive genes SAUR12 and SAUR78, along with the auxin efflux carrier gene PIN1c, were unexpectedly and persistently upregulated instead of being suppressed. The key explanatory clue for this paradox also comes from transcriptomic data, where the genes encoding critical transcriptional repressors in the auxin pathway, ARF17 and ARF18, showed sustained downregulation. Their reduced expression likely leads to the de-repression (removal of transcriptional inhibition) of their target genes, including specific SAURs and PIN1c, allowing for their aberrant activation despite low IAA levels. The aberrant high expression of PIN1c could further disrupt polar auxin distribution, exacerbating the imbalance in local concentration gradients.

In summary, this study obtained mutually reinforcing evidence from different molecular tiers, solidly confirming that metabolic reprogramming and auxin signaling network dysfunction are two early and central molecular underpinnings of stamen developmental defects in CMS K326.

3. Discussion

3.1. Gene Silencing and Activation in Modules Highly Correlated with CMS

This study identified key alterations in CMS K326 affecting energy metabolism, hormonal signaling, and floral development. Notably, the expression patterns of several genes displayed marked divergence between CMS K326 and its maintainer, where they were either nearly silent across all developmental stages in the maintainer line but highly expressed in the CMS line, or conversely, strongly expressed in the maintainer line but suppressed in the CMS line. Based on RNA-seq count data, low expression was defined as having counts below 10 in all samples (considered nearly silenced), while high expression corresponded to counts of several hundred or more, reaching a maximum of 1516 in certain samples.

Within the turquoise module, which was highly significantly negatively correlated with CMS, the genes NAC078 and IAA10A5 were consistently highly expressed in the maintainer line but showed minimal expression in CMS K326. The genes that strongly co-expressed with them (including ARF18, AIR12, HXK1, AGL82, and SOC1) were consistently down-regulated in CMS K326. This pattern suggests that their silencing is not an isolated event, but likely stems from the coordinated suppression of their entire co-expression network. In the greenyellow module (also significantly negatively correlated with CMS), WUS and NDB2 were highly expressed in the maintainer line but were significantly suppressed in CMS K326. Multiple energy-metabolism genes closely linked to WUS (UCP4, MDH, and MPC1) were synchronously down-regulated. Similarly, genes co-expressed with NDB2 (MPC1, COX5C, WUS, and WOX) were all down-regulated. It is noteworthy, however, that a subset of genes (IAA22, IAA29, and COX5C) exhibited upregulation during the small bud stage. The silencing of WUS and NDB2 may be closely associated with altered cellular energy–metabolism states. In the black module, which showed a highly significant positive correlation with CMS K326, the ATPB gene was nearly silent across all stages in the maintainer line but was highly expressed in the CMS line. Multiple nuclear genes displaying high transcriptional synergy with ATPB (PIN1c, ARF6, CRC, ARF8) were all up-regulated in the CMS line. These findings indicate that, within the CMS background, the altered expression of ATPB may represent one component of a broader, coordinated transcriptional reprogramming event, potentially driven by changes in the nucleo–cytoplasmic communication process.

To date, multiple mitochondrial genes functionally associated with CMS have been reported across various plant species, including nad7 [18], nad9 [19], cox1 [20,21], cox2 [22], COX11 [23], atp6 [24,25,26], atp8 [27], and atp9 [28]. These genes typically encode core components of oxidative phosphorylation, and their dysfunction directly compromises ATP synthesis. In alignment with this outcome, our data show a systemic transcriptional downregulation of nuclear-encoded counterparts and partners within the same pathways (e.g., COX5C, NDB2, and HXK1). This supports the view that energy disruption represents a common node in the CMS network. Beyond this conserved metabolic consequence, our study highlights a distinct layer of regulation in CMS K326, with targeted silencing of core floral developmental regulators. Genes such as WUS (involved in stem cell maintenance) and AGL82 (a floral organ identity gene) are not direct components of the electron transport chain. Their coordinated suppression, along with auxin signaling genes (IAA10A5, ARF18, PIN1c, and ARF6), suggests a mechanism for the early developmental arrest and homeotic transformations observed in cytoplasmic homeosis. Furthermore, the concerted upregulation of the mitochondrial ATPB gene with the nuclear auxin transporter PIN1c and response factor ARF6 indicates a potential transcriptional link between mitochondrial status (reflected by ATPB) and auxin-mediated development.

3.2. Hub Genes Identified by WGCNA and Key Genes Identified Through Differential Expression

Transcriptome analysis is a crucial method for elucidating gene function, regulatory mechanisms, and biological processes. The differential expression analysis of genes within CMS-associated modules provided complementary insights that reinforced and refined the functional interpretation derived from WGCNA. Notably, the hub genes identified by WGCNA (e.g., MTHFR in turquoise, IAA6B-like in greenyellow, and ARF8/ARF6 in black) were not necessarily those with the most significant expression changes. This dissociation between topological centrality within a co-expression network and the magnitude of differential expression highlights that the phenotypic impact of CMS may stem less from the drastic dysregulation of individual genes and more from a systemic imbalance in the coordinated regulation of entire functional modules. WGCNA identifies the centrality of genes in terms of their expression synergy/coordination within specific biological processes, rather than the magnitude of expression changes. A hub gene may itself exhibit stable expression levels; however, as a regulatory hub, slight alterations in its expression or post-translational modifications could be sufficient to drive substantial expression changes in a large number of downstream genes within the entire co-expression module, thereby leading to the phenotype. It is noteworthy that several module hub genes identified via WGCNA did not themselves show significant differential expression. This suggests that their role in CMS formation may not involve large-scale changes in their own expression levels, but rather that, as critical regulatory nodes, subtle dysfunctions (such as post-translational modifications or altered protein activity) or spatiotemporal expression abnormalities are adequate to disrupt the stability of their respective networks, consequently causing functional disturbances in downstream pathways.

Performing DEG analysis prior to WGCNA—i.e., constructing the network using only DEGs—may alter the network structure and lead to erroneous conclusions. In contrast, applying WGCNA first to comprehensively parse the entire dataset without altering its topological structure, and subsequently integrating the results with DEG analysis, has no adverse impact on the conclusions [29]. Our findings are consistent with this principle.

3.3. Does Energy Metabolism Affect Stamen Development in CMS K326 via the Auxin Pathway?

The relationship between floral development and auxin is well-documented. Floral organs originate from the floral meristem, which regulates both their initiation and termination [30]. While WUS activity is crucial for establishing and maintaining floral meristematic stem cells [31], the termination of the floral meristem primarily involves the inhibition and silencing of WUS, with AG also being an important regulatory factor [32]. Auxin plays a vital role in initiating floral meristems and supporting subsequent organ development [33]. Typically, floral organ initiation depends on local auxin maxima generated through polar auxin transport [34,35]. To achieve precise spatial positioning of floral organs within a whorl, it is necessary not only to establish and maintain distinct boundary regions between organs, but also to maintain low auxin signaling activity in these areas [36]. In cytoplasmic homeosis, besides abnormal stamen development, the disappearance of the boundary between stamens and pistils is a common accompanying manifestation. Correspondingly, not only is the expression of B or C genes down-regulated, but the expression of the cadastral gene SUP is also reduced [2]. However, studies on SUP mutants show that the carpelloid transformation of stamens is not due to the direct conversion of stamens into pistils, but rather results from the abnormal proliferation of pistil stem cells; SUP functions in this process by regulating auxin biosynthesis [13,14]. The fertility of the tobacco CMS Nta(rep) can be partially restored by overexpressing the Arabidopsis SUP gene [15]. In the CMS system, although stamens develop abnormally, their position within the floral whorl remains unchanged [16]. This suggests that in CMS, genes determining floral organ identity are likely affected, while genes related to floral development per se are relatively less impacted [37]. Histological section observations further confirm that the stamens of tobacco homeotic CMS are disturbed at very early developmental stages [12,38,39]. Furthermore, the activation of the meristem establishment gene WUS requires the formation of a local auxin concentration maximum, and mutations in ARF17 also lead to defects in male reproductive tissues [40]. Therefore, auxin is likely involved in regulating stamen development in CMS.

This study indicates that under extensive suppression of energy metabolism (blocked glycolysis, TCA cycle, and oxidative phosphorylation) in CMS K326, the decreased IAA level, downregulation of ARF17/ARF18, and aberrant upregulation of the efflux carrier PIN1c demonstrate disruption in auxin biosynthesis, transport, and signal transduction. This subsequently interrupts the normal distribution and signal gradient of auxin within the tissues. The silencing of the key meristem maintenance gene WUS may stem from the failure to establish the local auxin concentration maximum required for its activation. The termination of stem cell proliferation and the initiation of differentiation require the function of AG, whose activation depends on WUS. The failure of stem cells to initiate and terminate at the correct time and location inevitably leads to abnormal organ development. Therefore, energy metabolism disorder, by impairing the homeostasis of the auxin pathway and subsequently disrupting the floral meristem maintenance and organ identity determination governed by the WUS-AG axis, is likely a significant upstream trigger for stamen developmental defects in CMS K326.

3.4. Dilemmas in CMS Research and Omics Analysis

It is widely acknowledged that CMS arises from mitochondrial dysfunction that disrupts nuclear gene networks. However, significant controversy persists regarding the precise molecular identity of the factor inducing sterility and whether such mechanisms are universal or specific across different CMS types. The CMS phenotype manifests only in specific tissues and at particular developmental stages. Furthermore, interactions between diverse CMS types and various nuclear genetic backgrounds produce a continuous spectrum of phenotypic outcomes, rendering genetic analysis and mechanistic generalization exceedingly difficult. In practical breeding, the identification of restorer lines still relies heavily on extensive testing/cross-screening, lacking a mechanism-based, rational design and predictive capability. Consequently, for a newly discovered CMS type, rapidly identifying or creating its corresponding restorer line remains a major challenge. At the molecular level, the pathway from mitochondrial signals to nuclear responses may involve multiple regulatory layers, including metabolism, epigenetics, and transcription. The inability to establish a direct and exclusive causal link between mitochondrial aberration and anther developmental failure leaves the starting point for research ambiguous.

Omics analysis serves as a core methodology in functional genomics, particularly suited to systematically unraveling gene regulatory networks and biological processes. Its successful application, however, depends on rigorous experimental design, high-quality samples, robust bioinformatic analysis, and independent experimental validation. In practice, integrating multiple omics approaches or combining them with functional assays is often necessary to achieve a deeper biological understanding. A key weakness of such analyses is their tendency to generate hypotheses without providing validation. Moreover, when they are technology-driven rather than question-driven, their significance becomes limited if the research addresses questions that lack novelty or importance, fails to offer new insights into key field controversies, or merely reconfirms established conclusions.

In CMS research, traditional single-omics analyses often reveal limitations due to inadequate hypothesis testing and weak causal inference, easily relegating the work to a technology-driven descriptive exercise. Meaningful research must originate from a compelling and significant biological question, wherein omics analysis is employed as a tool to address pivotal queries rather than serving as the endpoint itself. For advancing CMS research, omics technologies remain indispensable for pinpointing critical regulatory nodes and identifying potential initiating signals within this complex system.

4. Materials and Methods

4.1. CMS K326 Plant Material, RNA Sequencing, and Bioinformatics Analysis

The cultivation conditions for CMS K326, sampling methodology, RNA extraction, sequencing, as well as subsequent bioinformatics and WGCNA, were performed as described in our previous study [19,21]. RNA-seq analysis was performed with n = 3 biological replicates per group. Metabolomic analysis was performed with n = 6 biological replicates per group. For network construction, the soft-thresholding power was set to 10, and the minimum module size was set to 30 genes. The morphological characteristics of the stamens are presented in Section 2.1 of [12].

In our manuscript, we used different expression metrics appropriately for distinct analytical purposes. For describing gene expression levels in the text and figures, we used FPKM (Fragments Per Kilobase of transcript per Million mapped reads).

For the Weighted Gene Co-expression Network Analysis (WGCNA), we used variance-stabilized transformed (VST) counts generated by the DESeq2 package [41], not raw counts.

WGCNA algorithm requires input data where the variance is independent of the mean. Raw count data (and to a lesser extent, FPKM) exhibit a strong mean–variance relationship, which can distort correlation calculations. The VST transformation effectively stabilizes the variance across the entire dynamic range of expression, making the data more suitable for correlation-based analyses like WGCNA. Using VST counts is a widely adopted and statistically sound approach for WGCNA, as recommended in numerous authoritative tutorials and publications. Building on this foundation, the following modifications were made to the WGCNA. Given the maternal inheritance of stamen malformation in CMS K326 and the role of auxin in floral organogenesis, gene selection was conducted based on the following criteria: a read count of at least 50 in one or more of the 18 samples, P_adj < 0.05, and |log₂FC| > 1. The selected genes were associated with energy metabolism, mitochondrial function, flower development, auxin pathways, and several known CMS-related factors, including retrograde signaling molecules, transcription factors, and chaperonins. WGCNA was used to analyze associations between gene modules and traits, with particular focus on the gene composition and differential expression within the highest-scoring submodule identified with the MCODE plugin (2.0.3) [42] in Cytoscape (3.10.4) [43]. This submodule belonged to the module most significantly associated with the CMS trait.

Furthermore, we performed untargeted metabolomics to profile metabolites related to energy metabolism and phytohormones, with emphasis on auxin. The detailed procedures are described below.

4.2. Untargeted Metabolomics Analysis

4.2.1. Sample Pretreatment

A total of six biological replicates per genotype were processed. Samples were removed from −80 °C and ground in liquid nitrogen, and 60 mg of each sample was weighed. Then, 1 mL of methanol/acetonitrile/water solution (2:2:1, v/v) was added, followed by vortexing for 60 s. The samples were subjected to low-temperature ultrasonication for 30 min, which was repeated twice, and placed at −20 °C for 1 h to promote protein precipitation. After filtration through a filter tube, centrifugation was performed at 14,000 rcf, and the samples were allowed to stand at 4 °C for 20 min. The supernatant was collected, freeze-dried, and stored at −80 °C.

4.2.2. Chromatography–Mass Spectrometry Analysis

Chromatographic Conditions

Samples were separated using an Agilent 1290 Infinity LC ultra-high-performance liquid chromatography (UHPLC) system with a HILIC column (Agilent Technologies, Waldbronn, Germany). The column temperature was maintained at 25 °C, and the flow rate was 0.3 mL/min. The mobile phase consisted of A: water + 25 mM ammonium acetate + 25 mM ammonia, and B: acetonitrile. The gradient elution program was as follows: 0–0.5 min, 95% B; 0.5–7 min, B linearly decreased from 95% to 65%; 7–8 min, B linearly decreased from 65% to 40%; 8–9 min, B maintained at 40%; 9–9.1 min, B linearly increased from 40% to 95%; 9.1–12 min, B maintained at 95%. Throughout the analysis, the samples were kept in an autosampler at 4 °C. To avoid the impact of instrument signal fluctuations, samples were analyzed in random order. Quality control (QC) samples were inserted into the sample queue to monitor and evaluate system stability and data reliability.

2Q-TOF Mass Spectrometry Conditions

After sample detection, an AB Triple TOF 6600 mass spectrometer (AB Sciex, Framingham, MA, USA) was used to acquire the primary and secondary spectra of the samples. The ESI source conditions, following HILIC chromatographic separation, were as follows: Ion Source Gas1 (Gas1): 60, Ion Source Gas2 (Gas2): 60, Curtain gas (CUR): 30, source temperature: 600 °C, and IonSapary Voltage Floating (ISVF) ±5500 V (both positive and negative modes); TOF MS scan m/z range: 60–1000 Da, product ion scan m/z range: 25–1000 Da, TOF MS scan accumulation time 0.20 s/spectra, and product ion scan accumulation time 0.05 s/spectra; secondary mass spectra were obtained using information-dependent acquisition (IDA) in high-sensitivity mode, with Declustering Potential (DP): ±60 V (both positive and negative modes) and Collision Energy: 35 ± 15 eV. The IDA settings were as follows: exclude isotopes within 4 Da, and candidate ions to monitor per cycle: 6.

Raw data were processed using XCMS software (4.6.3) for peak picking, retention time correction, and peak area quantification. Metabolite identification was based on precise mass (mass error < 10 ppm) and by matching MS/MS spectra against the CEU Mass database (https://ceumass.eps.uspceu.es/) (accessed on 22 December 2025) [43,44].

5. Conclusions

In this study, the interconnected regulatory landscape underlying stamen developmental defects in CMS K326 was elucidated through integrated transcriptomic and metabolomic analyses. The results indicate that early nuclear–cytoplasmic interaction is associated with subsequent energy metabolism reprogramming and an imbalance in the auxin signaling pathway. These concurrent upstream disturbances are linked to abnormalities in the core developmental program of stamens, culminating in the loss of their identity. It should be noted that the proposed relationships are primarily based on correlative evidence, and further validation through genetic and molecular biology experiments is required. Utilizing approaches such as gene editing and localized hormone treatment will provide more direct evidence for elucidating the CMS mechanism in tobacco and its potential breeding applications.