Ancient Liverwort Study Reveals No Universal Cell-Death Program Across Stresses

When plants face stress, they must decide, cell by cell, whether to fight or die. That decision, known as regulated cell death, is a genetically encoded process that plants deploy during development, immunity and environmental adaptation, eliminating damaged or unwanted cells with surgical precision. But a new study of the liverwort Marchantia polymorpha, published in Frontiers in Cell Death, suggests there is no single master switch behind it. By integrating publicly available transcriptomic data from three independent laboratories, researchers found that genes associated with regulated cell death are not switched on uniformly across different stresses. Instead, they are recruited selectively, in combinations that depend on the precise nature of the challenge, whether cold, darkness, high light, nutrient starvation or bacterial infection.

The research team, led by Aakanksha of the University of Delhi and Bal Govind Yadav of Academia Sinica in Taipei, performed a comparative reanalysis of 36 RNA-seq samples drawn from three published studies. The datasets covered seven abiotic stresses from a Marchantia cross-stress transcriptomic atlas, including cold, darkness, heat, high light, nitrogen deficiency, osmotic stress and salt, an independent high-light experiment examining B-GATA-mediated light responses, and a study of Pseudomonas syringae infection across evolutionarily divergent host plants. Rather than pooling published statistics, the authors reprocessed all raw sequencing reads through a single uniform pipeline, using FastQC for quality assessment, fastp for adapter and low-quality-base removal, and a decoy-aware Salmon index with GC-bias correction for transcript quantification. Differential expression was tested separately for each of the nine stress-versus-control comparisons using DESeq2, with log2 fold changes shrunk by the apeglm method to stabilise effect-size estimates for low-count genes. Genes were called differentially expressed at an adjusted P-value below 0.05 and an absolute log2 fold change of at least 1.

The scale of transcriptional change varied dramatically across conditions. The infection comparison produced the largest response, with 4,789 differentially expressed genes, of which 2,606 were upregulated and 2,183 downregulated. Darkness followed with 4,353 genes, heat with 3,793 and cold with 2,921. At the other end of the spectrum, osmotic stress yielded only 1,252 genes and salt 1,282, a hierarchy consistent with the original cross-stress atlas, which found that broad environmental stresses produce more transcriptional change than media stresses such as salinity and mannitol treatment. Every comparison retained three biological replicates for both stress and matched control conditions, and all libraries passed quality control with mapping rates above 80 percent.

To find a conserved core response, the researchers intersected the nine differential expression lists. The result was strikingly small: only 34 genes were shared across all nine comparisons, and just 12 of those changed in the same direction in every condition. Three were upregulated throughout, including a mitochondrial protein import gene, Mp7g17400, and nine were downregulated, including a lipoxygenase, Mp2g16000, involved in lipid-peroxidation signalling. The remaining 22 genes, among them two transcription factors, an AP2/ERF-family factor, Mp5g22160, and a bHLH factor, Mp2g00890, showed opposing directions between comparisons. Nine of these discordant genes were upregulated in infection and in the Schröder high-light experiment but downregulated in all seven comparisons from the Tan et al. atlas. Mp5g22160 showed the largest swing, rising by log2 fold change of 6.5 in infection and falling by 7.3 in osmotic stress. Because the seven Tan et al. comparisons share a common control, the authors caution that this pattern may reflect baseline expression in that control rather than a clean biotic-versus-abiotic divide.

The centrepiece of the analysis was a curated panel of 14 Marchantia genes previously linked to regulated cell death, spanning thionins, metacaspases, NADPH oxidases, autophagy proteins and the ZOU/ICE family of bHLH transcription factors. Ten of the 14 were significantly differentially expressed in at least one comparison, but their behaviour was overwhelmingly stress-specific. In the infection comparison, five genes met the significance criteria, all upregulated: the respiratory burst oxidase MpRBOHA, the type I metacaspase MpMCA-I, the autophagy proteins MpATG1 and MpATG7, and the AVRRPT2-INDUCED GENE2-like marker MpAIG2. No curated gene reached significance in the independent Schröder high-light comparison, despite that experiment producing a robust transcriptional response. Cold, darkness, heat, Tan high light and salt each detected one or two responsive genes, while nitrogen deficiency and osmotic stress detected none. Notably, none of the 14 genes appeared among the 34 shared across all nine conditions.

One gene broke the pattern. MpZOU2, an ancient paralogue of MpZOU1, was the only panel member significantly regulated in more than one comparison, appearing in four independent stress contexts. It was upregulated in cold, darkness and Pseudomonas syringae infection, and downregulated in the Tan et al. high-light treatment. MpZOU1, by contrast, was downregulated only in salt, while MpICE1 responded only to darkness. The recurrence of MpZOU2 across otherwise distinct experimental contexts marks it as a compelling candidate for functional investigation, though the authors are careful to stress that transcript-level changes alone do not establish cell-death execution. MpZOU1 and MpICE1 are known to promote terminal differentiation and programmed death of pegged rhizoids, but no equivalent function has yet been demonstrated for MpZOU2, which may participate in broader stress-responsive transcriptional networks or respond to stress independently of cell death altogether.

Because the datasets came from different laboratories, sequencing platforms and reference pipelines, the authors applied batch correction with ComBat-seq before integrated analyses, preserving biological condition information while minimising study-associated variation. Principal component analysis showed that control samples from different studies formed separate clusters before correction and overlapped substantially afterwards, with the variance explained by PC1 rising from 39.9 to 43.5 percent. The convergence was biologically meaningful: high-light samples from the two independent studies clustered closely together and shared 1,211 differentially expressed genes, demonstrating that the high-light response is reproducible across differences in light intensity, treatment duration, growth conditions and laboratory of origin. Salt and osmotic stress also clustered together but shared only 741 genes, suggesting that salt engages a substantial ion-specific response beyond its shared osmotic component. Darkness remained a distinct outlier in the expression space both before and after correction, reflecting a strong condition-specific program rather than residual technical variation.

Weighted gene co-expression network analysis added a genome-wide layer beyond the curated panel. After filtering, 18,030 genes were included in the network, which resolved into 64 co-expression modules ranging from 33 to 1,931 genes, built with a soft-thresholding power of 14 chosen by the scale-free topology criterion. Two modules showed significant stress associations. The turquoise module, the largest in the network, correlated with infected samples and was enriched in protein ubiquitination, proteasome-mediated degradation, ER-associated degradation, autophagy and intracellular membrane trafficking, painting infection as a coordinated activation of protein quality control and cellular recycling pathways. The green module correlated with nitrogen deficiency and osmotic stress and was enriched in oxidative stress responses, hydrogen peroxide catabolism and cell-wall organisation, consistent with known roles of reactive oxygen species in redox homeostasis and cell-wall integrity during osmotic and nutrient stress.

The authors frame these findings within the broader evolutionary context of bryophyte cell death. Marchantia has already yielded genetically regulated developmental examples, including the MpZOU1/MpICE1 module in rhizoid differentiation and MpMLO1-dependent calcium signalling during sperm discharge, showing that conserved cell-death regulatory modules exist outside angiosperms. The new analysis suggests the same principle applies to environmental responses: conserved cell-death-associated components are retained as part of an ancestral regulatory toolkit but are recruited selectively by different upstream signalling networks rather than through a universal pathway. The pattern parallels cross-stress comparisons in rice, which similarly identified a small set of shared genes alongside larger stress-specific signatures, hinting that limited transcriptional convergence across stress types may be a general feature of plant stress biology.

The study carries important caveats. Transcript abundance does not demonstrate increased protein abundance, enzymatic activity or cell-death execution, a limitation particularly relevant to metacaspases, autophagy proteins and ROS regulators, which can participate in acclimation as well as death. The five genes induced during infection occur in a system where the phytotoxin syringomycin independently promotes tissue necrosis, but pathogen-associated necrosis can arise from host defence, toxin activity or tissue damage, and should not automatically be equated with canonical regulated cell death. The 14-gene panel is a candidate set rather than a complete catalogue, so limited overlap should not be read as evidence that cell death is absent from the remaining stresses. Distinguishing whether recurrent candidates such as MpZOU2 execute cell death or merely respond to stress will require loss- and gain-of-function experiments, cell-type-specific expression studies and direct measurements of membrane integrity, ROS accumulation, protease activity and cell viability. Until then, the work stands as a hypothesis-generating map of how an ancient plant deploys its cell-death toolkit, one stress at a time.

Subject of Research: Cross-stress transcriptomic integration in Marchantia polymorpha reveals stress-specific expression of regulated cell-death-associated genes

Article Title: Cross-stress transcriptomic integration in Marchantia polymorpha reveals stress-specific expression of regulated cell-death-associated genes

Article References: Aakanksha, & Yadav, B. G. (2026). Cross-stress transcriptomic integration in Marchantia polymorpha reveals stress-specific expression of regulated cell-death-associated genes. Frontiers in Cell Death, 5, Article 1985736. https://doi.org/10.3389/fceld.2026.1985736

Image Credits: AI Generated

DOI: 10.3389/fceld.2026.1985736

Keywords: Marchantia polymorpha, regulated cell death, RNA-seq, transcriptomics, WGCNA, plant stress, MpZOU2, bryophytes, Cross-stress, transcriptomic, integration, Marchantia