Polygenic resistance is associated with altered early immune timing and changes in transcriptome network structure

Polygenic resistance is associated with altered early immune timing and changes in transcriptome network structure | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 28 November 2025 V1 Latest version Share on Polygenic resistance is associated with altered early immune timing and changes in transcriptome network structure Authors : Martin Hádlík , Miroslav Baránek , Kateřina Baránková , and Viera Kovacova 0000-0003-4581-4254 [email protected] Authors Info & Affiliations https://doi.org/10.22541/au.176431171.13982474/v1 174 views 177 downloads Contents Abstract Author Contributions Supplementary Material Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Downy mildew, caused by Plasmopara viticola , is a major threat to grapevine. To investigate how pyramiding resistance loci influences early immune processes, we generated 36 time-resolved transcriptomes (0, 6, 24 hpi) from genotypes carrying single (Rpv12), double (Rpv12+1), or triple (Rpv12+1+3) resistance loci, together with a susceptible control. Aggregated expression divergence revealed that multilocus genotypes showed detectable baseline transcriptional differences prior to infection and distinct temporal trajectories following inoculation. Co-expression analysis identified five higher-order metamodules and uncovered non-additive network restructuring, with the triple-locus line showing markedly reduced co-expression density rather than incremental strengthening. Layer-associated genes related to recognition, signal integration, and defense action displayed genotype- and time-specific expression patterns, suggesting that each locus combination is associated with characteristic shifts in immune-layer dynamics. Transcriptome profiles suggested that Rpv12 involves elevated transcript abundance of several EDS1-associated components despite its CNL-type origin; adding Rpv1 introduced pronounced early changes in recognition- and signaling-related genes; and stacking Rpv3 coincided with rapid, transient induction of transcription factors (e.g., WRKY47, NAC29) and metabolic defense pathways. Together, these results show that resistance loci are associated with non-additive differences in timing, network connectivity, and layer allocation of immune-associated transcriptional programs, revealing systems-level mechanisms underlying multilocus resistance. Title: Polygenic resistance is associated with altered early immune timing and changes in transcriptome network structure Hádlík M. 1 , Baránek M. 1 , Baránková K. 1 , Kovacova V. 2# 1 Mendeleum—Institute of Genetics, Faculty of Horticulture, Mendel University in Brno, Valtická 334, 69144 Lednice, Czech Republic 2 Institute for Biological Physics, University of Cologne, Zuelpicher Str. 77, 50937, Cologne, Germany ABSTRACT Downy mildew, caused by Plasmopara viticola , is a major threat to grapevine. To investigate how pyramiding resistance loci influences early immune processes, we generated 36 time-resolved transcriptomes (0, 6, 24 hpi) from genotypes carrying single (Rpv12), double (Rpv12+1), or triple (Rpv12+1+3) resistance loci, together with a susceptible control. Aggregated expression divergence revealed that multilocus genotypes showed detectable baseline transcriptional differences prior to infection and distinct temporal trajectories following inoculation. Co-expression analysis identified five higher-order metamodules and uncovered non-additive network restructuring, with the triple-locus line showing markedly reduced co-expression density rather than incremental strengthening. Layer-associated genes related to recognition, signal integration, and defense action displayed genotype- and time-specific expression patterns, suggesting that each locus combination is associated with characteristic shifts in immune-layer dynamics. Transcriptome profiles suggested that Rpv12 involves elevated transcript abundance of several EDS1-associated components despite its CNL-type origin; adding Rpv1 introduced pronounced early changes in recognition- and signaling-related genes; and stacking Rpv3 coincided with rapid, transient induction of transcription factors (e.g., WRKY47, NAC29) and metabolic defense pathways. Together, these results show that resistance loci are associated with non-additive differences in timing, network connectivity, and layer allocation of immune-associated transcriptional programs, revealing systems-level mechanisms underlying multilocus resistance. INTRODUCTION The oomycete Plasmopara viticola poses a significant threat to grapevine ( Vitis spp.) species, causing the devastating downy mildew (DM) disease (Viala, 1983; Koledenkova et al., 2022). P. viticola can reduce wine production by a staggering 75 % (Buonassisi, 2017). Currently, copper-based fungicides remain the most effective means of suppressing P. viticola in vineyards; however, their intensive use raises environmental concerns (Buonassisi, 2017). A more ecological approach to reducing downy mildew in vineyards is to cultivate interspecific genotypes, which can effectively limit pathogen development due to their resistance genes. Various grapevine breeding programs have implemented introgressive hybridization of single resistance traits from wild American and Asian non-vinifera Vitis species to mitigate severe yield losses (Eisenmann et al., 2019; Ochssner et al., 2016; Qu et al., 2021). However, a coevolutionary arms race continually drives oomycete counterdefense (Dussert et al., 2019; Fontaine et al., 2021). While pyramiding is a central strategy for durable field resistance (Ciubotaru, 2021; Mundt, 2018; Töpfer, 2011), the fundamental biological principles governing how stacked NLR receptors are associated with early immune signal transduction remain poorly understood. To date, 37 loci of resistance to Plasmopara viticola ( Rpv ) have been described (Ricciardi et al., 2024). The Rpv1, Rpv3-1, Rpv10, and Rpv12 loci are considered the major loci responsible for controlling resistance and are the most commonly used in breeding programs (Heyman, 2021). The Rpv1 locus was identified in the wild North American grape Muscadinia rotundifolia (Feechan, 2013) , the Rpv3 locus was identified in the wild North American grape Vitis rupestris (Bellin et al., 2009; Merdinoglu et al., 2003) , and the Rpv12 locus was identified in the wild Northeast Asian grape Vitis amurensis (Schwander et al., 2012; Venuti et al., 2013) . The Rpv1 locus is considered dominant and highly polymorphic, acting in a qualitative, gene‐for‐gene manner (Feechan et al., 2013). In contrast, Rpv3 and Rpv12 loci confer more quantitative resistance, fine-tuning immune responses without a strictly all-or-nothing HR (Bellin et al., 2009; Merdinoglu et al., 2018). These Rpv loci usually contain major dominant and polymorphic resistance genes that encode NLR proteins, providing full or partial resistance to one or more pathogens (Lacombe, 2010; Schulze-Lefert & Panstruga, 2011). Intracellular NLR protein(s) have a typical structure composed of NB-ARC (nucleotide-binding) domains, which are usually followed by LRR (leucine-rich repeats) domains. The N terminus of NLRs can contain a Toll/interleukin-1 receptor (TIR) domain (proteins are referred to as TNLs), coiled-coil (CC) domain (CNLs), or RPW8 (RESISTANCE TO POWDERY MILDEW 8) domain (RNLs) (Wang et al., 2023). NLRs recognize pathogen-derived molecules or help control signal transduction (Feehan et al., 2020). In Arabidopsis , TIR-NLR (TNL)–mediated immunity strictly requires the EDS1 signaling node, whereas coiled-coil NLRs (CNLs) typically act independently of EDS1 (Lapin et al., 2019; Saile et al., 2021; Lapin et al., 2020). However, the specific signaling architecture for Vitis NLRs and how they are coordinated is less resolved. Pathogen recognition, signal integration, and defense actions are critical components of a tightly controlled, complex cascade of immune mechanisms (Wang, 2019). Traditionally, defense mechanisms are divided into two tiers of the immune response: a broad-spectrum first tier, known as pattern-triggered immunity (PTI), and a more specialized and robust second tier, referred to as effector-triggered immunity (ETI). The evolutionarily older PTI is activated when pattern recognition receptors (PRRs) on the cell surface detect conserved microbe-, damage-, or pathogen-associated molecular patterns (MAMPs, DAMPs, PAMPs). ETI is activated when NLRs intracellularly detect pathogen effectors, often triggering localized cell death (hypersensitive response, HR), which is observed as necrotic lesions. PTI does not involve HR. Despite activation in distinct cell compartments, PTI and ETI overlap in various steps and converge in the downstream outputs (Yuan et al., 2021a). PTI-linked cell surface-located pattern recognition receptors (PRRs) are also required for ETI-linked NLR-mediated plant immunity, thereby merging two central immune signaling cascades into a single functional network (Yuan et al., 2021b; Ngou et al., 2022). Recently, Wang et al. (Wang, 2019) suggest examining plant-pathogenic oomycete infection through immunity layers: the recognition layer, the signal-integration layer, and the defense-action layer. Recognition occurs extracellularly via PRRs or intracellularly via NLRs. The signal integration layer is a complex network encompassing phosphorylation and ubiquitination, protein turnover and trafficking, dynamics of plant hormones and reactive oxygen species (ROS) and nitrogen species, transcriptional regulation, and signals from systemic acquired resistance (SAR). The defense layer encompasses programmed cell death (PCD), the synthesis and secretion of antimicrobial proteins and compounds, RNA interference, and alterations in the cell walls. Successful cooperation of these mechanisms should result in limited mycelial growth, a reduced number of sporangia, or necrotic lesions (Wang, 2019). Transcriptomic studies of the Rpv1 locus revealed that it confers TNL-mediated resistance, linked to calcium signaling, activation of WRKY and MYB transcription factors, leading to the production of ROS and stilbenes, and PCD (Qu et al., 2021). Similarly, the Rpv3 locus confers TNL-mediated resistance, characterized by robust HR (necrotic lesions and PCD) and increased biosynthesis of fungi-toxic stilbenes (Eisenmann et al., 2019; Wairich et al., 2022). The Rpv12 locus confers CNL-mediated resistance with an early onset of PCD. PCD onset in the Rpv3 genotype was delayed and not evident until 28 h post-inoculation (Wingerter et al., 2021). Most metabolite alterations in sugars, fatty acids, and phenols during Rpv12-mediated ETI mirrored the timing and direction in Rpv3-mediated ETI; however, some changes were more prominent or enduring (Chitarrini et al., 2020). Stronger Rpv12-mediated effects are supported by findings that novel Plasmopara viticola isolates surpass grapevine Rpv3.1 and Rpv3.2 resistance but not Rpv12 (Gouveia, 2024). Thus, Rpv1, Rpv3, and Rpv12 differ not only in origin and NLR type, but also in temporal activation kinetics (Wingerter et al., 2021). However, their molecular phenotype is not yet fully characterized. Because grapevine is a long-lived woody perennial, durable defence requires coordination between strong pathogen restriction and long-term physiological balance, making multilocus immune interactions particularly relevant. The effects of combining Rpv loci on transcriptomic regulation remain poorly understood, particularly regarding how pyramiding alters the timing and coordination of immune responses. Although recent work in Arabidopsis shows that PTI and ETI often act concurrently rather than hierarchically (Ngou et al., 2021; Yuan et al., 2021a), it is unclear whether similar coordination underpins polygenic resistance in perennial crops. Here, we use time-resolved RNA sequencing to characterize transcriptomic responses in genotypes carrying single (Rpv12), double (Rpv12+1), or triple (Rpv12+1+3) resistance loci, compared to a susceptible control. To our knowledge, this represents the first transcriptomic study examining Rpv pyramiding, providing transcriptomic insight into how stacked loci influence immune timing, regulatory network architecture, and functional layer allocation. To support broader use in research and breeding, we also provide a curated, Vitis–Plasmopara interactive database of all transcriptional dynamics. RESULTS AND DISCUSSION Wet-lab and RNA sequencing overview To investigate how stacked resistance loci reshape early immune responses, we performed a controlled infection experiment followed by time-resolved RNA sequencing. Three grapevine genotypes carrying single (Rpv12), double (Rpv12+1), or triple (Rpv12+1+3) resistance loci, together with a susceptible control, were inoculated with Plasmopara viticola . Leaf discs were sampled at 0, 6, and 24 h post-inoculation (hpi), yielding 36 RNA-seq libraries (3 replicates × 4 genotypes × 3 time points). Sampling focused on key transition phases (recognition, activation, resolution) previously defined in grapevine– Plasmopara interactions (Wang et al., 2019; Chitarrini et al., 2020; Wingerter et al., 2021), providing temporal granularity sufficient to resolve early immune timing. The experimental design and analytical workflow are summarized in Figure 1. After adapter trimming and quality filtering, we obtained high-quality reads for all samples, detecting expression of 26,169 protein-coding genes. However, sequencing was performed across two batches, and several read mapping metrics (Supplementary Table 1, Supplementary Figure 1) differed significantly between batches (FDR < 0.014), suggesting batch-specific fragmentation or RNA degradation, consistent with known technical variation in multi-run RNA-seq datasets (Sheng et al., 2016). To prevent this technical structure from confounding biological comparisons, we systematically evaluated multiple normalization strategies using principal component analysis (PCA) and correlation between PC1 and batch labels (Supplementary Figure 2 and 3). Size-factor normalization (Love et al., 2014), combined with ComBat (Leek et al., 2012) correction (for amplitude-sensitive analyses such as divergence) and rlog transformation (Love et al., 2014) with ComBat (for correlation-based analyses, including clustering and network inference), most effectively minimized residual batch effects while preserving true biological variability. These quality-controlled, batch-corrected datasets formed the basis for all downstream differential expression, divergence, network, and functional layer analyses. Aggregated expression divergence among genotypes and global transcriptional architecture To assess how stacked resistance loci reshape global transcriptomic states, we quantified aggregated expression divergence (AED)—the mean squared expression difference between each resistant genotype and the susceptible control—and evaluated significance using null distributions of resampled replicate combinations (Figure 2A, Supplementary Table 2). Following Nourmohammad et al. (2017), this aggregate measure summarizes genome-wide expression divergence as an integrated physiological trait, enabling the detection of coordinated or adaptive shifts in regulatory activity, rather than changes in individual genes alone. At 0 hpi, Rpv12 showed only modest divergence (FDR = 0.12), whereas both Rpv12+1 and Rpv12+1+3 were already significantly different from the susceptible baseline (FDR = 0.024). This indicates that resistance pyramiding does not merely amplify inducible defense responses but also shows measurable baseline transcriptional differences prior to infection. The detectable baseline transcriptional differences are consistent with patterns observed in primed systems, where immune receptor accumulation or low-level signalling can establish a sensitized state that enables faster or stronger activation upon attack (Conrath et al., 2015; Mauch-Mani et al., 2017). Whether similar mechanisms underpin the baseline differences observed here remains to be determined, but our data suggest that the magnitude of baseline divergence increases with the number of loci. Divergence further increased in all resistant genotypes, but with distinct temporal profiles. At 6 hpi, Rpv12+1+3 displayed the most substantial shift from the susceptible transcriptome (FDR = 0.024), consistent with earlier shifts in defense-associated transcript abundance. By 24 hpi, Rpv12 and Rpv12+1+3 displayed significant divergence (FDR = 0.024). While Rpv12 displayed late activation, Rpv12+1+3 showed a decrease in AED, suggesting that locus combinations influence not only the magnitude but also the timing and resolution of immune responses. Principal component analysis of all 26,169 genes reinforced these patterns (Figure 2B). While PC1 (42.5% variance) captured the overall resistance complexity gradient, PC2 separated Rpv12 and Rpv12+1+3 from Rpv12+1 and the susceptible genotype, indicating distinct resistance architectures. PC3 further separated susceptible plants from all resistant genotypes, revealing a core transcriptional signature of resistance independent of time. Genes with the highest contribution (Supplementary Figure 4) to PC1 were strongly enriched for signalling functions (FDR = 1.30e-10) forming a dense protein–protein network (FDR = 0.034). PC2 was enriched for immune system reactome pathways (FDR = 0.00036) and PCD-related factors (FDR = 9.20e-05), aligning with effector-triggered immunity pathways (Zhang & Dong, 2022) and suggesting that transcriptional shifts consistent with a primed-like state in Rpv12+1+3 involves the transcriptional patterns enriched in ETI-associated genes. PC3 was enriched in the Plant-type hypersensitive response (FDR = 7.98e-06). PC4 was enriched in the biosynthesis of L-pipecolic acid (FDR=0.0143), a compound that serves as a signaling molecule to initiate SAR in response to pathogen infections (Návarová et al., 2012). Together, these results show that polygenic resistance is associated with progressive genome-wide transcriptional changes and detectable baseline expression shifts indicative of priming-like patterns. Temporal transcriptional dynamics A total of 3,553 genes (Supplementary table 3) were identified as differentially expressed (DE) across all experimental conditions (FDR 1). The distribution of DE genes among individual comparisons is summarized in Table 1. Across resistant genotypes, the extent of transcriptomic divergence from the susceptible background was driven primarily by infection timing rather than genotype identity or direction of change (Table 2; Model A: timing p < 3.4 × 10⁻⁶, genotype p = 0.23, direction p = 0.55), indicating that the phase of infection (0, 6, 24 hpi) explained most variation in DEG counts. DEG counts did not scale with the number of introgressed loci (Model B: loci p = 0.40), revealing non-additivity. Instead, each resistance combination reshaped dynamics in a time-dependent manner. Rpv12 alone showed weak timing effects (Model C – Rpv12: timing p = 0.26), whereas Rpv12+1 produced a sharp early burst at 6 hpi (Model C – Rpv12+1: p < 2 × 10⁻¹⁶) and Rpv12+1+3 was associated with strong transcriptional induction at 6 hpi followed by partial resolution at 24 hpi (Model C/D – Rpv12+1+3: timing p < 2.2 × 10⁻¹⁶). Per-timepoint models displayed progressive genotype divergence (Model D): genotype effects were already detectable at 0 hpi (p = 0.024), intensified at 6 hpi (p = 0.0012), and became highly significant by 24 hpi (p < 2 × 10⁻¹⁶), revealing distinct immune trajectories among stacked genotypes. Although direction was not globally significant (Model A: p = 0.55), it became critical during the activation phase: at 6 hpi, downregulation dominated (Model D – 6 hpi: direction p < 4 × 10⁻¹¹), suggesting suppression of susceptible processes is a core feature of effective defense. Category-level (Supplementary table ST4) modeling of temporal patterns (Model E) revealed that timing is the major driver of DEG proportions (p < 2.2 × 10⁻¹⁶), with transient 6 hpi responses (TSR) strongly enriched (p = 5.1 × 10⁻¹⁰), while stable changes across all time points were rare. Pairwise contrasts (Model F) revealed that Rpv12+1 initially mounted more DEGs than Rpv12 at 6 hpi (p = 0.022). However, by 24 hpi, Rpv12 became the most transcriptionally active genotype (p < 0.0001), indicating a shift in relative responsiveness over time. These nonlinear transitions suggest potential non-additive timing-dependent differences among loci, with early Rpv1-like effects followed by compensatory quantitative responses from Rpv12 at later stages. Such timing-dependent patterns compatible with coordinated PTI–ETI-associated transcript changes and non-additive multilocus effects mirror findings from Arabidopsis , where PTI and ETI reinforce each other dynamically (Ngou et al., 2021; Yuan et al., 2021a), and from rice NLR networks, where multilocus interactions reshape both amplitude and timing of immune outputs rather than summing linearly (De la Concepcion et al., 2021). Together, these models indicate that resistance pyramiding is associated with differences in network connectivity, when and how defense programs are deployed, rather than merely increasing DEG counts. Splicing junction variation Splicing‐junction counts (only junctions supported by ≥10 uniquely mapped reads were analyzed; Supplementary table 5) were modeled with negative binomial GLMs including an offset for library size and an explicit batch term. At the global level, neither infection timing nor batch explained SJ abundance (timing p ≈ 0.48; batch p ≈ 0.45), while genotype effects were modest overall with a specific reduction in the triple‐stacked line (Rpv12+1+3, p = 0.0274). Per-gene NB models across 14,596 SJ-bearing protein-coding loci showed that the most significant terms were genotype-specific (1,948 genes, FDR < 0.05). In contrast, only 23 genes were time-specific and 44 showed genotype×time interactions, indicating that splicing variation is predominantly structured by genetic background. Inspection of presence/absence patterns revealed that 45 SJs appeared restricted to resistant backgrounds, while 12 appeared restricted to susceptible backgrounds. The resistant-restricted set was enriched for NB-ARC/TIR/LRR annotations (FDR < 0.05), highlighting classic NLR resistance loci, while the susceptible-restricted set was enriched for ribosomal protein domains (FDR < 0.05). Given the known impact of reference-mapping bias (Ahmad et al., 2025) and the presence or absence of structural variations on junction discovery, we interpret a substantial fraction of these ”SJs” as allele- or haplotype-specific junctions arising from introgressed NLR-rich regions rather than classical, conditionally regulated alternative splicing per se. This is consistent with the well-documented structural dynamism of Rpv loci (Feechan et al., 2013; Venuti et al., 2013) and with broad evidence that NLR clusters are among the most structurally variable regions in plant genomes (Sarris et al., 2016; Barragan & Weigel, 2021). Although, additional long-read or genomic data would be required for confirmation. In parallel, a small subset of loci (19) enriched in hydrolase activity (FDR<0.05) showed significant genotype, time, and interaction terms without batch effects, supporting bona fide regulated splicing for a minority of genes. Overall, multilocus resistance is associated with transcript profile differences that likely reflect both structural dynamism of introgressed resistance loci, a key evolutionary aspect of plant immunity, and regulatory variation, with only a minority of genes showing evidence of regulated splicing. Gene co-expression network architecture and metamodules To investigate how resistance stacking reshapes regulatory architecture, we built correlation-based coexpression networks from 3,553 differentially expressed genes across genotypes and time points (Supplementary table 6). At 0 hpi, networks exhibited very high density (ρ ≈ 0.126; mean degree ≈ 446) and strong clustering (≈ 0.75), indicative of extensive basal co-expression patterns prior to pathogen contact. Density declined steadily over time (0 → 6 → 24 hpi; linear model p = 0.025; Supplementary Figure 5), reflecting desynchronization as transcription becomes more specialized during defense. Comparing genotypes revealed a striking nonlinearity: Rpv12 and Rpv12+1 formed denser, more connected networks than the susceptible control (ρ ≈ 0.13; mean degree ≈ 468–471), suggesting these loci are associated with higher co-expression density. In contrast, Rpv12+1+3 showed markedly reduced connectivity (ρ ≈ 0.055; mean degree ≈ 194), indicating substantial reorganization of coexpression patterns. Neither linear regression nor Spearman’s correlation supported a monotonic relationship between the number of loci and network density (p ≈ 0.5), suggesting that polygenic resistance alters when and how genes co-vary, not simply how strongly they co-vary. Global clustering remained moderate (0.59–0.77) in all conditions, indicating that modular organization is preserved even when the overall correlation structure is disrupted. Together, these results suggest that stacked resistance loci remodel the temporal and topological architecture of transcriptional networks rather than uniformly amplifying connectivity, consistent with a dynamic reallocation of regulatory control. To identify functional coordination points, we detected 155 high-confidence co-transcriptional modules (containing≥4 genes, with a Spearman correlation against the module eigengene and a Bonferroni-adjusted p-value < 0.05). These modules were enriched for processes central to immunity—including NLR signaling, stress responses, hormone metabolism, cell wall modification, and secondary biosynthesis, underscoring their functional relevance. Degree centrality analysis revealed 56 hub genes (top 1%; 117–1552 partners; Supplementary Table 7), including NAC, bZIP, and WRKY transcription factors (Tsuda & Somssich, 2015), receptor-like kinases (e.g., FERONIA, WAK5-like; Tanaka & Heil, 2021), proteases (e.g., subtilases), metabolic regulators, and multiple canonical R genes (e.g., RPM1, RPP13-like; Kourelis & van der Hoom, 2018). Hub gene identity and connectivity shifted across genotypes, indicating that resistance stacking is associated with genotype-specific differences in hub gene connectivity. Many modules shared large fractions of genes and clustered into five “meta-modules,” representing higher-order regulatory programs (Figure 3A, Supplementary Figure 7). One meta-module (MM2) was consistently downregulated across all resistant genotypes, beginning with the single-locus Rpv12 line, and was enriched for hypersensitive-response (HR) genes. This pattern suggests that HR-related gene attenuation represents a consistent feature of Rpv12-mediated resistance, to maintain balanced immunity. MM2 contained downregulated FERONIA, a receptor-like kinase that positively regulates immune signaling (Guo et al., 2018), and LAZ5, a TNL receptor associated with autoimmune cell death (Palma et al., 2010), supporting the view that coordinated downregulation of this module is consistent with reduced transcription of HR-associated genes. Although direct PTI–ETI gene–gene correlations were rare (as expected for temporally staggered signals), PTI-, ETI-, and shared immunity genes frequently co-occurred within the same metamodule, showing coordinated timing shifts in their expression. This supports the view that PTI and ETI are coordinated at the network level rather than through simple pairwise coexpression (Hatsugai et al., 2012; Tsuda & Somssich, 2015; Yuan et al., 2021a; Yuan et al., 2021b). Overall, multilocus resistance is associated with enhanced early coordination (Rpv12, Rpv12+1) or phase reprogramming (Rpv12+1+3) by rewiring hub genes, merging functional modules, and timing the allocation of defense layers. To dissect how defense layers interact within these restructured networks, we quantified transcriptional dynamics of genes classified as ETI-specific, PTI-specific, or PTI–ETI-shared (Figure 3B; Supplementary Table 8; Tsuda & Somssich, 2015; Lapin et al., 2019; Ngou et al., 2022; Shu et al., 2023). ETI-specific genes showed moderate basal upregulation in Rpv12 and Rpv12+1 at 0 hpi, followed by attenuation over time, whereas the triple-locus line (Rpv12+1+3) displayed overall downregulation across all stages. PTI–ETI-shared genes were associated with similar transient upregulation peaking at 0 hpi, supporting early convergence of immune signaling. In contrast, PTI-specific genes showed weak or inconsistent modulation, suggesting that primary PTI signaling is rapidly subsumed by ETI-dominated responses. These findings align with prior studies showing that ETI triggers more sustained or earlier transcriptomic responses compared with PTI (Mine et al., 2018; Wang et al., 2023; Yu et al., 2024). Together, these profiles reveal that multilocus resistance is associated with the altered timing and amplitude of immunity gene activation rather than uniformly elevating expression, consistent with coordinated but temporally staggered PTI–ETI integration. Layer-specific immune responses Plant immunity emerges from coordinated activity across multiple functional layers—Recognition, Signal Integration, and Defense Action—rather than from individual loci acting in isolation (Wang et al., 2019). Based on literature curation (Wang et al., 2019; Zhou & Zhang, 2020; Coleman et al., 2020, Tanaka & Heil, 2021; Shu et al., 2023), we identified 36 differentially expressed genes (DEGs) with well‐established layer-specific roles (Figure 4). We examined their temporal and genotype-specific expression to evaluate how pyramiding reshapes the architecture of immunity. In Rpv12, layer-associated DEGs were relatively evenly distributed across Recognition (5), Signal Integration (4), and Defense Action (3), consistent with a quantitatively regulated ETI system. Surface co-receptors such as SOBIR1 (Liebrand et al., 2013) and intracellular hubs such as EDS1 (Wiermer et al., 2005) were upregulated, consistent with connections between PTI-associated perception and ETI signaling cascades. Several initial expression variation (IEV) genes were upregulated at 0 hpi, including EDS1 and its partner SAG101, together with SARD4, WRKY55, WRKY51, MIK2, SOBIR1, and the resistance gene RUN1 (Figure 5). EDS1–SAG101 complexes form a core signaling node required for TNL-mediated effector recognition and systemic acquired resistance, whereas EDS1–PAD4–ADR1 modules modulate transcriptional defenses and can interface with CNL-triggered pathways (Lapin et al., 2020; Chhillar et al., 2025). The coordinated transcript changes in genes linked to local and systemic defense processes are consistent with early transcriptional upregulation of both local and systemic defense layers. Thus, although the Rpv12 locus is typically considered as an EDS1-independent CNL-type receptor, we observed elevated transcript abundance of multiple EDS1-associated components, suggesting potential signaling cross-talk between CNL- and TNL-like immune branches. Signal Integration genes were dominated by WRKY transcription factors, supporting convergence on SA/ROS/JA pathways (Tsuda & Somssich, 2015). In the Defense Action layer, SARD4 (Ding et al., 2016) and ACD6 (Lu et al., 2003) expression indicated that Rpv12 primarily triggers a reactive defense program after pathogen ingress, consistent with our DEG and network analyses showing time as the dominant driver of transcriptional divergence. In Rpv12+1, the highest number of layer-specific DEGs occurred in the Recognition (6) and Signal Integration layers (6), with a notable expansion of PRRs and NLRs. The Rpv1 locus, introgressed from Muscadinia rotundifolia , is associated with strong early penetration resistance and high NLR density (Feechan et al., 2010; Feechan et al., 2013), consistent with the early transcriptional activation and higher DEG counts at 6 hpi observed in our data. Significantly, stacking Rpv1 onto Rpv12 did not merely increase the number of DEGs but showed increased co-occurrence of PTI–ETI-associated transcripts, as evidenced by coordinated upregulation of transcriptional regulators such as TGA1, MAPKKK1, MYB12, and PRE6. These factors integrate hormonal, metabolic, and MAPK signaling inputs (Tsuda & Somssich, 2015), consistent with a robust, rapidly deployed, and highly coordinated immune response in Rpv12+1, as reflected by increased network density and hub connectivity. In Rpv12+1+3, the distribution of layer-associated DEGs shifted toward Signal Integration (6) and Defense Action (3), with comparatively few Recognition-layer genes (3). This indicates that adding Rpv3 does not simply increase receptor-level activity but is associated with broader shifts in downstream signalling transcripts, consistent with the reduced coexpression density and altered module organisation observed in our network analyses. Signal Integration was dominated by pronounced induction of RIPK, WRKY47, bZIP53, and NAC transcription factors (e.g. NAC29, NAC JA2L), reinforcing multilayer hormonal and stress signaling. Defense Action genes such as MYB4 (stilbenoids; Kim et al., 2024), EPS1 (SA biosynthesis; Torrens-Spence et al., 2024) and CSLG2 (cell wall remodeling; Xia et al., 2025) suggest a rapid transcription of genes enriched in antimicrobial and structural pathways. Together with elevated baseline expression at 0 hpi, these findings support a model in which Rpv12+1+3 is associated with elevated baseline expression and early activation of several defense-associated genes, consistent with the nonlinear, timing-dependent transcriptional trajectories observed across stacked genotypes. Overall, integrating layer-specific gene behavior with global differential expression, splicing, and coexpression network data further supports that pyramiding does not act additively. Instead, each added locus is associated with differences in timing and relative transcription of immune layer-associated genes: the Rpv12 locus is associated with reactive ETI-linked defense; Rpv12+1 reinforces early recognition and PTI–ETI integration; and Rpv12+1+3 alters signaling hierarchy toward primed, rapid, and extensive activation of defense-related transcriptional programs. These multilocus effects suggest that durable resistance may not arise from increased magnitude alone, but from reconfiguration of immune layer coordination in time and network space, aligning with the emerging view that polygenic resistance is a systems-level property rather than a linear sum of gene functions. This study provides the first comprehensive comparison of transcriptomic responses across three grapevine genotypes carrying successively pyramided resistance loci (Rpv12, Rpv1, Rpv3). By integrating differential expression, splicing, temporal modeling, and co-expression network analysis, we reveal that resistance stacking does not act additively: each locus uniquely reshapes the timing, coordination, and network topology of immune programs. Rpv12 shows transcriptomic patterns consistent with quantitative ETI-like responses with sustained late activity. Adding Rpv1 (Rpv12+1) is associated with earlier upregulation of recognition- and PTI–ETI integration-related transcripts, consistent with its origin from Muscadinia rotundifolia . The inclusion of Rpv3 (Rpv12+1+3) further alters signaling and defense modules, and is associated with elevated baseline levels of defense-related transcripts. Network-level analyses identified five metamodules capturing generalized expression trajectories and genotype specificity, with hub genes enriched in transcription factors, defense regulators, and NLRs. Despite increased resistance, the triple-stacked genotype showed reduced co-expression density, indicating regulatory restructuring rather than amplification. Likewise, splicing junction analysis suggests that multilocus resistance modifies not only transcript abundance but also transcript architecture, reflecting both regulatory shifts and structural divergence at resistance loci. Collectively, our findings support a systems-level view of pyramiding: durable resistance could involve coordinated shifts in the timing and integration of immune-related pathways, rather than from increased expression amplitude alone. Although, causal relationships can not be inferred from transcriptomics alone. These insights highlight the importance of considering gene–gene interactions, timing, and network context when designing knowledge-driven breeding strategies for sustainable disease resistance. The identified hub and layer-specific genes represent candidate biomarkers for assessing immune readiness or resistance durability in breeding populations. Integrating these transcriptomic signatures into marker-assisted selection could accelerate deployment of successful new cultivars harboring multilocus based resistance. As this study is based on transcriptomic data, causal inference and mechanistic conclusions are limited; functional assays and structural-genomic analyses would further clarify the pathways suggested by our data. MATERIAL AND METHODS Plasmopora viticola inoculum P. viticola was obtained in June 2022 as a natural inoculum from Vinselekt Michlovský a.s. vineyards in the phase of oil spots on the leaves. These leaves were stored in the dark for 48 hours at 100% humidity and 22°C until sporulation occurred. The identity of the inoculum as Plasmopara viticola was also confirmed by a PCR test and sequenced (Gessler et al., 2011). Excised P. viticola segments were incubated in water for 4 hours to prepare an inoculum, which was sprayed on sterile-washed ’Pinot Noir’ leaves. Plants were kept at 22°C and 100% humidity until sporulation, after which the densest segments were aseptically excised. Spore concentration was measured using Cellometer K2 and adjusted to 50,000 spores/ml. Plant artificial inoculation and sample collection In January 2022, segments of one-year-old wood were collected from the vineyards of Vinselekt Michlovský a.s and planted. Genotypes with a unique combination of R-loci (Rpv12, Rpv12+1, and Rpv12+1+3) were selected based on laboratory analysis using SSR markers linked to resistance loci (Hádlík et al., 2024). Because resistance loci in grapevine breeding are introgressed through multiple generations of mutual crossing of individual sources of Rpv loci, the resulting lines are not strictly isogenic, and each genotype may carry additional linked genomic regions beyond the targeted loci. Selection process takes about 15 years for a new grape variety (Avia et al., 2023). However, Rpv12, Rpv12+1, Rpv12+1+3 genotypes used in this study represent stable and resistant materials used in breeding programs, although additional linked regions may be present. ’Pinot Noir’ was used as the sensitive control and source of fresh inoculum. At the five-leaf stage, 20 µl of spore suspension was pipetted onto the abaxial side of three leaves on each plant. Leaves were then collected at three time points: 0 hours (immediately after inoculation), 6 hours, and 24 hours post-inoculation (hpi). A one-centimeter-diameter disc was excised from the inoculation spot of each leaf and stored at -80°C until RNA isolation. Samples were collected to ensure three biological replicates for each variant. Total RNA isolation Total RNA was isolated using the Spectrum™ Plant Total RNA Kit (Sigma-Aldrich) protocol. The concentration of each sample was measured using SPECTROstar Nano. RNA integrity was measured using an Agilent 2100 Bioanalyser. The library preparation and sequencing for 36 samples, consisting of four genotypes at three time points with three replicates each, were conducted by Novogene in the UK in two batches. Quality control and mapping The quality of the transcriptomic libraries was assessed using FastQC (version 0.11.9; Andrews S., 2010) and trimmed with Trimmomatic (version 0.39; Bolger et al., 2014). All high-quality reads were mapped onto the reference genome of Vitis vinifera subsp. vinifera (ENSEMBL, ID: PN40024 version 4; 35 134 annotated protein-coding genes) using the STAR mapping tool (version 2.7.4a; Dobin et al., 2013). The featureCounts function (package Subread, version 2.0.3; Liao et al., 2014) estimated the raw counts of mapped reads per gene. Transcripts were annotated by matching gene/protein names from multiple reference resources using BLAST’s blastP (v2.12.0+; Johnson et al., 2008), yielding a conversion table of IDs, names, and symbols. Raw count normalization and batch correction To verify the presence of batch effects, we compared sequencing and mapping metrics between libraries produced in two independent batches. The proportions of forward-only, reverse-only, and correctly paired reads differed significantly between batches (GLM p-values < 0.01; Fig. Sx). Two complementary normalization strategies were used depending on the analysis goals. For computation of Aggregated Expression Divergence (AED), size-factor–normalized counts from DESeq2 (Love et al., 2014) were log₂-transformed and corrected for confirmed systematic technical variation using ComBat (sva package; Leek et al., 2012). This approach preserves the full dynamic range of expression values, enabling quantitative comparison of global expression shifts across genotypes and time points. We used a regularized log (rlog) transformation, followed by ComBat correction, for correlation analyses, clustering, and visualization. The rlog transformation stabilizes variance and improves homoscedasticity, which is desirable for assessing co-expression structure but would artificially compress expression amplitudes if used for divergence quantification. Analysis of Expression Dynamics Aggregated Expression Divergence (AED) was defined to capture the overall transcriptional distance of each tolerant genotype from the susceptible reference at a given time point. For each gene g, we computed the difference between the mean batch-corrected log₂ expression in the tolerant genotype and the susceptible reference. AED was calculated as the mean squared difference across all genes. Significance was assessed by comparing observed AEDs to a null distribution generated from all 220 possible random triplets of samples at each time point. The empirical one-sided p-value was defined as the proportion of null AEDs exceeding the observed AED, and multiple testing across nine genotype–time comparisons was controlled by Benjamini–Hochberg False Discovery Rate (FDR). Differential Gene Expression Analysis (DGEA) Batch-corrected rlog-transformed expression values were used for differential gene expression analysis. Because batch correction was required, and DESeq2 does not allow pre-corrected values as input, we used rlog+ComBat values and t-tests with FDR control. A two-tailed t -test was applied to compare group means and resulting p -values were adjusted for multiple testing using the Benjamini–Hochberg False Discovery Rate (FDR). Genes with FDR 1 were considered differentially expressed. Gene categorization We categorized each differentially expressed gene in every tolerant genotype according to its specific time-dependent expression pattern. At each time point (0, 6, and 24 hpi), the gene’s expression can either be equal to the reference (E or 0), downregulated (D or -1), or upregulated (U or 1). The transcriptional dynamics of each gene in a particular genotype are represented by a combination of three characters (E or 0, D or -1, U or 1). These three character combinations of each genotype (the simplified expression pattern) allow us to categorize the dynamics of transcriptional changes into six categories: 1) Changes detected at 0 hpi or 0 and 6 hpi are called Initial Expression Variations (IEV), 2) Changes detected at 6 hpi and 24 hpi indicate Early Response (ER), 3) Changes detected only at 6 hpi are called Transient Response to Stress (TRS), 4) Changes detected at 24 hpi indicate Late Response (LR), 5) Changes detected in all three time points highlight Sustained Change of expression (SCh), 6) Changes different from the abovementioned are called Complex Patterns (CPs). Grouping of gene categories Categorized genes can be further grouped by comparing the simplified expression pattern among genotypes. If a pattern is listed as shared between two genotypes, the other genotype consistently keeps gene expression levels equivalent to the reference (EEE). We define six groups (Supplementary Figure 7): Group I - gene category shared by all three genotypes, Group II - gene category shared by Rpv12 and Rpv12+1 genotypes, Group III - gene category shared by Rpv12+1 and Rpv12+1+3 genotypes, Group IV - recovered gene category (patterns shared by Rpv12 and Rpv12+1+3 genotypes), Group V - genotype-specific gene categories (changes found just in one genotype), Group VI - complex patterns across genotypes. Group V has three subgroups: (a) Rpv12-specific changes, (b) Rpv12+1-specific changes, and (c) Rpv12+1+3-specific changes. The majority rule approach was used to generalize the expression pattern of a queried group of genes. Analysis of Expression Dynamics To assess how genotype, infection time, and direction of regulation influence the number of differentially expressed genes (DEGs), we applied generalized linear models (GLMs) with appropriate error structures. For DEG counts, we used the negative binomial GLM with a log link. Because all DEGs were defined relative to the susceptible reference (baseline = 0), library-size normalization was already inherent in the design, and no offset was required. To identify specific differences between factor levels (e.g., between genotypes at a given time point or between time points within a genotype), we used estimated marginal means with Tukey’s Honest Significant Difference post-hoc tests (emmeans package), which apply internal multiple-comparison correction. Model assumptions, dispersion, and heteroscedasticity were examined using residual diagnostics, and quasibinomial families were employed when overdispersion was detected. When the sample size within a stratum was low or the dispersion was extreme, we interpreted the results in combination with effect sizes derived from model coefficients (expressed as a fold change on the log scale). Weighted Gene Co-expression Network Analysis (WGCNA) Pearson correlation coefficients were calculated pairwise for all 26169 detected genes (rlog values). Only pairs above the 99.5th percentile of all correlations (r>0.817, p0.817, p<0.01). The module’s eigengene (the first principal component) was computed from the rlog normalized expression values and correlated (a Spearman correlation test) with the simplified expression pattern of the DE gene of interest. Only modules with significant FDR were retained (p<0.05). The WGCNA organizes the most biologically relevant changes into cohesive modules to understand system-level gene functionality (Farber, 2013; Russell et al., 2023). An interaction network (iGraph R package) was constructed from significant modules and queried for hub proteins, which are proteins in the top 1% of degree centrality. A higher-level grouping of modules was achieved by calculating Euclidean distances based on shared proteins between modules and hub proteins. These distances were hierarchically clustered using the complete linkage method, simplifying the dataset into clearer higher-level metamodules. These metamodules were then analyzed for functional enrichment, genotype specificity, and temporal behavior, and integrated with immunity layer annotations to infer how resistance loci rewire signaling architecture at the network level. To examine how transcriptome-wide network architecture changes over time and with increasing numbers of introgressed resistance loci, we constructed correlation-based co-expression networks (iGraph R package) separately per genotype and per time point using the 3,553 DE genes. For each subset of samples (e.g. Rpv12, Rpv12+1, Rpv12+1+3, or susceptible; and separately 0 hpi, 6 hpi, 24 hpi), we computed pairwise Pearson correlations on rlog-normalized expression values. An edge was retained if the correlation exceeded a stringent threshold (r ≥ 0.817; 99.5th percentile of the global correlation distribution; we tested thresholds from r=0.75–0.85 and and observed the same qualitative differences). Logical adjacency matrices were converted into undirected igraph objects with edge weights equal to the corresponding correlation coefficients. Global network metrics were then extracted using igraph: number of nodes, number of edges, density (edge_density), mean degree, and clustering coefficient (transitivity(type=”global”)). To quantify the influence of multilocus stacking on network cohesion, network density was regressed against the number of introgressed loci and evaluated with Pearson and Spearman correlations. In contrast, network density was also regressed against infection timing, revealing temporal trends in co-expression coupling during immune activation. To investigate cross-layer immune interactions, we annotated each gene according to immunity class (PTI-specific, ETI-specific, PTI/ETI-shared, or Other) using a curated literature-based classification. These annotations were mapped onto the nodes of each genotype-specific network. We then quantified functional integration between PTI and ETI layers using a custom “cross-talk” function: for each network, we counted edges connecting PTI and ETI (or shared) nodes, computed their proportion relative to all edges, and compared this to the expected proportion under random pairing based on node class frequencies. An odds ratio (observed / expected) was calculated to test whether PTI–ETI edges were over- or under-represented. Splicing junctions Splicing junction (SJ) counts were obtained from STAR alignments, retaining only junctions supported by ≥10 uniquely mapped reads, ≥20 bp overhang on each side, and detected in ≥3 samples. Junctions were annotated to protein-coding genes, and the total number of SJs per sample was normalized using the log-transformed library size as an offset. For each gene, we fitted a negative binomial generalized linear model of: SJ count ~ genotype × time + batch + offset(log library size). Model parameters (genotype, time, and their interaction, batch design) were tested using likelihood ratio tests (ANOVA with Chi-square test). P-values were adjusted to identify genes with significant genotype-, time-, or interaction-dependent splicing (FDR < 0.05). In addition, we explored extreme patterns where SJs were consistently present in all resistant samples but absent in all susceptible samples (and vice versa). Gene Enrichment Analysis (GEA) Functional enrichment of Gene Ontology (GO) terms, Reactome pathways, UniProt keywords, and protein domains was assessed using the STRING database (Szklarczyk et al., 2025), using Vitis vinifera annotation as well as Arabidopsis homologs. To facilitate exploration of the dataset, we developed an interactive web application using the R Shiny framework (shiny, ggplot2, DT, Biostrings, shinyWidgets packages). The app allows users to query gene-level information, visualize expression profiles, inspect co-transcriptional modules, and download associated data. It integrates normalized expression matrices (rlog- and ComBat-corrected), raw read counts, and reference FASTA sequences of Vitis vinifera (PN40024 v4). Homology relationships with Arabidopsis thaliana were obtained by BLASTP (e-value < 1 × 10⁻⁶) against TAIR10 proteins and are provided unfiltered to retain all potential orthologs. Each gene entry links to UniProt, NCBI, and TAIR10 identifiers, and displays its co-expression partners (Pearson’s r ≥ 0.817) together with Bonferroni-adjusted module statistics. The application, titled Grapevine Guardians: Unraveling Rpv Pyramidization’s Impact on Immunity , is available at https://vierakovacova.shinyapps.io/playing_with_immunity/ and the complete source code is accessible via GitHub ( https://github.com/vierocka/plant_immunity_Vitis ). Data Availability Statement Raw RNA-seq data have been deposited in the NCBI Sequence Read Archive under accession number NCBI BioProject: PRJNA1358055 . An interactive Shiny application that mirrors and facilitates exploration of the transcriptomic results is available at https://vierakovacova.shinyapps.io/playing_with_immunity/. All scripts used for quality control, mapping, differential expression analysis, AED, and network analysis are available at https://github.com/vierocka/plant_immunity_Vitis. Author Contributions VK: conceptualisation of sequencing data analysis and interpretation, data curation and co-writing the original draft of the manuscript; MH: experimental work, co-writing the original draft; MB: conceptualisation of experiment, co-writing the original draft; KB: conceptualisation of experiment, experimental work Acknowledgements We would like to express our gratitude to VINSELEKT MICHLOVSKÝ a.s. for providing the grapevine breeding materials used in this study. Study realization was supported by Mendel University in Brno, Project IGA-ZF/2023-SI1-015, and Deutsche Forschungsgemeinschaft (DFG) CRC 1310, project number 325931972. Conflict of interest disclosure The authors declare no conflict of interest. Ethics approval statement All experimental procedures complied with institutional guidelines for plant research. Permission to reproduce material from other sources All figures and data are original. No material has been reproduced from other sources. References Ahmad, B., Su, Y., Hao, Y., Razzaq, T., Arshad, R., Zhang, Yi, et al. (2025) Mango pangenome reveals dramatic impacts of reference bias on population genomic analyses. Horticulture Research, 12. Avia, K., Schneider, C., Onimus, C., Arnold, G., Dumas, V., Umar-Faruk, A., et al. (2023) 22nd GiESCO International Meeting THE FRENCH GRAPEVINE BREEDING PROGRAM RESDUR: STATE OF THE ART AND PERSPECTIVES. Université de Strasbourg, INRAE, SVQVUMR-A 1131, F-68000 Colmar, France. Barragan, A.C. and Weigel, D. (2021) Plant NLR diversity: The known unknowns of pan-NLRomes. Plant Cell, 33, 814–831. Bellin, D., Peressotti, E., Merdinoglu, D., Wiedemann-Merdinoglu, S., Adam-Blondon, A.F., Cipriani, G., et al. (2009) Resistance to Plasmopara viticola in grapevine “Bianca” is controlled by a major dominant gene causing localised necrosis at the infection site. Theoretical and Applied Genetics, 120, 163–176. Bolger, A.M., Lohse, M., and Usadel, B. (2014) Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics, 30, 2114–2120. Buonassisi, D., Colombo, M., Migliaro, D., Dolzani, C., Peressotti, E., Mizzotti, C., et al. (2017) Breeding for grapevine downy mildew resistance: a review of “omics” approaches. Euphytica, 213. Chhillar, H., Nguyen, H.H., Yeh, P.M., Jones, J.D.G., and Ding, P. (2025) Modular mechanisms of immune priming and growth inhibition mediated by plant effector-triggered immunity. Cell Reports, 44. Chitarrini, G., Riccadonna, S., Zulini, L., Vecchione, A., Stefanini, M., Larger, S., et al. (2020) Two-omics data revealed commonalities and differences between Rpv12- and Rpv3-mediated resistance in grapevine. Scientific Reports, 10. Ciubotaru, R.M., Franceschi, P., Zulini, L., Stefanini, M., Škrab, D., Rossarolla, M.D., et al. (2021) Mono-Locus and Pyramided Resistant Grapevine Cultivars Reveal Early Putative Biomarkers Upon Artificial Inoculation With Plasmopara viticola. Frontiers in Plant Science, 12. Coleman, A.D., Maroschek, J., Raasch, L., Takken, F.L.W., Ranf, S., and Hückelhoven, R. (2021) The Arabidopsis leucine-rich repeat receptor-like kinase MIK2 is a crucial component of early immune responses to a fungal-derived elicitor. New Phytologist, 229, 3453–3466. Concepcioni, J.C.D.L., Maidment, J.H.R., Longya, A., Xiao, G., Franceschetti, M., and Banfieldi, M.J. (2021) The allelic rice immune receptor Pikh confers extended resistance to strains of the blast fungus through a single polymorphism in the effector binding interface. PLoS Pathogens, 17. Conrath, U., Beckers, G.J.M., Langenbach, C.J.G., and Jaskiewicz, M.R. (2015) Priming for Enhanced Defense. Annual Review of Phytopathology, 53, 97–119. Ding, P., Rekhter, D., Ding, Y., Feussner, K., Busta, L., Haroth, S., et al. (2016) Characterization of a pipecolic acid biosynthesis pathway required for systemic acquired resistance. Plant Cell, 28, 2603–2615. Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., et al. (2013) STAR: Ultrafast universal RNA-seq aligner. Bioinformatics, 29, 15–21. Dussert, Y., Mazet, I.D., Couture, C., Erô Me Gouzy, J., Piron, M.-C., Kuchly, C., et al. (2019) A High-Quality Grapevine Downy Mildew Genome Assembly Reveals Rapidly Evolving and Lineage-Specific Putative Host Adaptation Genes. GBE, 11, 954–969. Eisenmann, B., Czemmel, S., Ziegler, T., Buchholz, G., Kortekamp, A., Trapp, O., et al. (2019) Rpv3-1 mediated resistance to grapevine downy mildew is associated with specific host transcriptional responses and the accumulation of stilbenes. BMC Plant Biology, 19. Farber, C.R. (2013) Systems-level analysis of genome-wide association data. G3: Genes, Genomes, Genetics, 3, 119–129. Feechan, A., Anderson, C., Torregrosa, L., Jermakow, A., Mestre, P., Wiedemann-Merdinoglu, S., et al. (2013) Genetic dissection of a TIR-NB-LRR locus from the wild North American grapevine species Muscadinia rotundifolia identifies paralogous genes conferring resistance to major fungal and oomycete pathogens in cultivated grapevine. Plant Journal, 76, 661–674. Feehan, J.M., Castel, B., Bentham, A.R., and Jones, J.D. (2020) Plant NLRs get by with a little help from their friends. Current Opinion in Plant Biology, 56, 99–108. Fontaine, M.C., Labbé, F., Dussert, Y., Delière, L., Richart-Cervera, S., Giraud, T., and Delmotte, F. (2021) Europe as a bridgehead in the worldwide invasion history of grapevine downy mildew, Plasmopara viticola. Current Biology, 31, 2155-2166.e4. Gessler, C.P., Ilaria; Perazzolli, Michele (2011) Plasmopara viticola: a review of knowledge on downy mildew of grapevine and effective disease management. Phytopathologia Mediterranea, 50, 3–44. Gouveia, C., Santos, R.B., Zukic, S., Manthey, T., Malhó, R., Figueiredo, A., and Buchholz, G. (2024) Novel Plasmopara viticola isolate surpasses grapevine Rpv3.1 and Rpv3.2 resistance but not Rpv12. Journal of Plant Pathology. Guo, H., Nolan, T.M., Song, G., Liu, S., Xie, Z., Chen, J., et al. (2018) FERONIA Receptor Kinase Contributes to Plant Immunity by Suppressing Jasmonic Acid Signaling in Arabidopsis thaliana. Current Biology, 28, 3316-3324.e6. Hádlík, M., Baránek, M., Flajšingerová, I., and Baránková, K. (2024) Verification of resistance loci pyramiding in popular interspecific grape varieties using SSR markers. Czech Journal of Genetics and Plant Breeding, 60, 34–41. Hatsugai, N., Igarashi, D., Mase, K., Lu, Y., Tsuda, Y., Chakravarthy, S., et al. (2017) A plant effector‐triggered immunity signaling sector is inhibited by pattern‐triggered immunity. The EMBO Journal, 36, 2758–2769. Heyman, L., Höfle, R., Kicherer, A., Trapp, O., Ait Barka, E., Töpfer, R., and Höfte, M. (2021) The Durability of Quantitative Host Resistance and Variability in Pathogen Virulence in the Interaction Between European Grapevine Cultivars and Plasmopara viticola. Frontiers in Agronomy, 3. Johnson, M., Zaretskaya, I., Raytselis, Y., Merezhuk, Y., McGinnis, S., and Madden, T.L. (2008) NCBI BLAST: a better web interface. Nucleic acids research, 36. Kim, Y.R., Moon, Y.B., Choi, H.B., Han, J.Y., Choi, H.S., Shim, S., and Choi, Y.E. (2024) MYB4 is the best candidate transcription factor involved in pinosylvin stilbene biosynthesis in Pinus strobus L. cells by fungal elicitor treatment. Annals of Forest Science, 81. Koledenkova, K., Esmaeel, Q., Jacquard, C., Nowak, J., Clément, C., and Ait Barka, E. (2022) Plasmopara viticola the Causal Agent of Downy Mildew of Grapevine: From Its Taxonomy to Disease Management. Frontiers in Microbiology, 13. Kourelis, J. and Van Der Hoorn, R.A.L. (2018) Defended to the nines: 25 years of resistance gene cloning identifies nine mechanisms for R protein function. Plant Cell, 30, 285–299. Lacombe, S., Rougon-Cardoso, A., Sherwood, E., Peeters, N., Dahlbeck, D., Van Esse, H.P., et al. (2010) Interfamily transfer of a plant pattern-recognition receptor confers broad-spectrum bacterial resistance. Nature Biotechnology, 28, 365–369. Lapin, D., Bhandari, D.D., and Parker, J.E. (2020) Origins and Immunity Networking Functions of EDS1 Family Proteins. Annual Review ofPhytopathology, 58, 253–276. Lapin, D., Kovacova, V., Sun, X., Dongus, J.A., Bhandari, D., Von Born, P., et al. (2019) A coevolved EDS1-SAG101-NRG1 module mediates cell death signaling by TIR-domain immune receptors. Plant Cell, 31, 2430–2455. Leek, J.T., Johnson, W.E., Parker, H.S., Jaffe, A.E., and Storey, J.D. (2012) The SVA package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics, 28, 882–883. Liao, Y., Smyth, G.K., and Shi, W. (2014) FeatureCounts: An efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics, 30, 923–930. Liebrand, T.W.H., Van Den Berg, G.C.M., Zhang, Z., Smit, P., Cordewener, J.H.G., America, A.H.P., et al. (2013) Receptor-like kinase SOBIR1/EVR interacts with receptor-like proteins in plant immunity against fungal infection. Proceedings of the National Academy of Sciences of the United States of America, 110, 10010–10015. Love, M.I., Huber, W., and Anders, S. (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology, 15. Lu, H., Rate, D.N., Song, J.T., and Greenberg, J.T. (2003) ACD6, a Novel Ankyrin Protein, Is a Regulator and an Effector of Salicylic Acid Signaling in the Arabidopsis Defense Response. Plant Cell, 15, 2408–2420. Malacarne, G., Vrhovsek, U., Zulini, L., Cestaro, A., Stefanini, M., Mattivi, F., et al. (2011) Resistance to Plasmopara viticola in a grapevine segregating population is associated with stilbenoid accumulation and with specific host transcriptional responses. BMC Plant Biology, 11. Mauch-Mani, B., Baccelli, I., Luna, E., and Flors, V. (2017) Defense Priming: An Adaptive Part of Induced Resistance. The Annual Review ofPlant Biology, 68, 485–512. Merdinoglu, D., Schneider, C., Prado, E., Wiedemann-Merdinoglu, S., and Mestre, P. (2018) Breeding for durable resistance to downy and powdery mildew in grapevine. Oeno One, 52, 189–195. Merdinoglu, D., Wiedemann-Merdinoglu, S., Coste, P., Dumas, V., Haetty, S., Butterlin, G., et al. (2003) Genetic Analysis of Downy Mildew Resistance Derived from Muscadinia rotundifolia. In: Proc. VIIIth IC on Grape, pp. 451–456. ISHS 2003, Acta Horticulturae 603. Mine, A., Seyfferth, C., Kracher, B., Berens, M.L., Becker, D., and Tsuda, K. (2018) The defense phytohormone signaling network enables rapid, high-amplitude transcriptional reprogramming during effector-triggered immunity[OPEN]. Plant Cell, 30, 1199–1219. Mundt, C.C. (2018) Pyramiding for resistance durability: Theory and practice. Phytopathology, 108, 792–802. Návarová, H., Bernsdorff, F., Döring, A.C., and Zeier, J. (2013) Pipecolic acid, an endogenous mediator of defense amplification and priming, is a critical regulator of inducible plant immunity. Plant Cell, 24, 5123–5141. Ngou, B.P.M., Ahn, H.K., Ding, P., and Jones, J.D.G. (2021) Mutual potentiation of plant immunity by cell-surface and intracellular receptors. Nature, 592, 110–115. Ngou, B.P.M., Ding, P., and Jones, J.D.G. (2022) Thirty years of resistance: Zig-zag through the plant immune system. Plant Cell, 34, 1447–1478. Nourmohammad, A., Rambeau, J., Held, T., Kovacova, V., Berg, J., and Lässig, M. (2017) Adaptive Evolution of Gene Expression in Drosophila. Cell Reports, 20, 1385–1395. Ochssner, I., Hausmann, L., and Töpfer, R. (2016) Rpvl4, a new genetic source for Plasmopara viticola resistance conferred by Vitis cinerea. Vitis - Journal of Grapevine Research, 55, 79–81. Palma, K., Thorgrimsen, S., Malinovsky, F.G., Fiil, B.K., Nielsen, H.B., Brodersen, P., et al. (2010) Autoimmunity in arabidopsis acd11 is mediated by epigenetic regulation of an immune receptor. PLoS Pathogens, 6. Qu, J., Dry, I., Liu, L., Guo, Z., and Yin, L. (2021) Transcriptional profiling reveals multiple defense responses in downy mildew-resistant transgenic grapevine expressing a TIR-NBS-LRR gene located at the MrRUN1/MrRPV1 locus. Horticulture Research, 8. Ricciardi, V., Crespan, M., Maddalena, G., Migliaro, D., Brancadoro, L., Maghradze, D., et al. (2024) Novel loci associated with resistance to downy and powdery mildew in grapevine. Frontiers in Plant Science, 15. Russell, M., Aqil, A., Saitou, M., Gokcumen, O., and Masuda, N. (2023) Gene communities in co-expression networks across different tissues. PLoS Computational Biology, 19. Saile, S.C., Ackermann, F.M., Sunil, S., Keicher, J., Bayless, A., Bonardi, V., et al. (2021) Arabidopsis ADR1 helper NLR immune receptors localize and function at the plasma membrane in a phospholipid dependent manner. New Phytologist, 232, 2440–2456. Sarris, P.F., Cevik, V., Dagdas, G., Jones, J.D.G., and Krasileva, K.V. (2016) Comparative analysis of plant immune receptor architectures uncovers host proteins likely targeted by pathogens. BMC Biology, 14. Schulze-Lefert, P. and Panstruga, R. (2011) A molecular evolutionary concept connecting nonhost resistance, pathogen host range, and pathogen speciation. Trends in Plant Science, 16, 117–125. Schwander, F., Eibach, R., Fechter, I., Hausmann, L., Zyprian, E., and Töpfer, R. (2012) Rpv10: A new locus from the Asian Vitis gene pool for pyramiding downy mildew resistance loci in grapevine. Theoretical and Applied Genetics, 124, 163–176. Sheng, Q., Vickers, K., Zhao, S., Wang, J., Samuels, D.C., Koues, O., et al. (2017) Multi-perspective quality control of Illumina RNA sequencing data analysis. Briefings in Functional Genomics, 16, 194–204. Shu, L.J., Kahlon, P.S., and Ranf, S. (2023) The power of patterns: new insights into pattern-triggered immunity. New Phytologist, 240, 960–967. Szklarczyk, D., Nastou, K., Koutrouli, M., Kirsch, R., Mehryary, F., Hachilif, R., et al. (2025) The STRING database in 2025: protein networks with directionality of regulation. Nucleic Acids Research, 53, D730–D737. Tanaka, K. and Heil, M. (2021) Damage-Associated Molecular Patterns (DAMPs) in Plant Innate Immunity: Applying the Danger Model and Evolutionary Perspectives. Annual Review ofPhytopathology, 59, 53–75. Töpfer, R., Hausmann, L., Harst, M., Maul, E., Zyprian, E., and Eibach, R. (2011) Fruit, Vegetable and Cereal Science and Biotechnology: New Horizons for Grapevine Breeding. Global Science Books. Torrens-Spence, M.P., Matos, J.O., Li, T., Kastner, D.W., Kim, C.Y., Wang, Z., et al. (2024) Mechanistic basis for the emergence of EPS1 as a catalyst in salicylic acid biosynthesis of Brassicaceae. Nature Communications, 15. Tsuda, K. and Somssich, I.E. (2015) Transcriptional networks in plant immunity. New Phytologist, 206, 932–947. Venuti, S., Copetti, D., Foria, S., Falginella, L., Hoffmann, S., Bellin, D., et al. (2013) Historical Introgression of the Downy Mildew Resistance Gene Rpv12 from the Asian Species Vitis amurensis into Grapevine Varieties. PLoS ONE, 8. Viala (1893) Les maladies de la vigne. Wairich, A., Malabarba, J., Buffon, V., Porto, D.D., Togawa, R., and Revers, L.F. (2022) Molecular characterization of the Rpv3 locus towards the development of KASP markers for downy mildew resistance in grapevine (Vitis spp.). Euphytica, 218. Wang, D., Wei, L., Liu, T., Ma, J., Huang, K., Guo, H., et al. (2023) Suppression of ETI by PTI priming to balance plant growth and defense through an MPK3/MPK6-WRKYs-PP2Cs module. Molecular Plant, 16, 903–918. Wang, J., Song, W., and Chai, J. (2023) Structure, biochemical function, and signaling mechanism of plant NLRs. Molecular Plant, 16, 75–95. Wang, Yan, Tyler, B.M., and Wang, Yuanchao (2019) Defense and Counterdefense During Plant-Pathogenic Oomycete Infection. Annual Review of Microbiology, 73, 667–696. Wiermer, M., Feys, B.J., and Parker, J.E. (2005) Plant immunity: The EDS1 regulatory node. Current Opinion in Plant Biology, 8, 383–389. Wingerter, C., Eisenmann, B., Weber, P., Dry, I., and Bogs, J. (2021) Grapevine Rpv3-, Rpv10- and Rpv12-mediated defense responses against Plasmopara viticola and the impact of their deployment on fungicide use in viticulture. BMC Plant Biology, 21. Xia, W., Su, Y., Song, J., Wang, L., Yang, M., Yang, J., et al. (2025) Cellulose synthase superfamily key in DAMPs-triggered immunity against Sclerotinia stem rot in Brassica napus. Plant Stress, 18. Yang, D.L., Yang, Y., and He, Z. (2013) Roles of plant hormones and their interplay in rice immunity. Molecular Plant, 6, 675–685. Yin, L., Zander, M., Huang, S.C., Xie, M., Song, L., Guzmán, J.P.S., et al. (2023) Transcription Factor Dynamics in Cross-Regulation of Plant Hormone Signaling Pathways. bioRxiv. Yu, X.Q., Niu, H.Q., Liu, C., Wang, H.L., Yin, W., and Xia, X. (2024) PTI-ETI synergistic signal mechanisms in plant immunity. Plant Biotechnology Journal, 22, 2113–2128. Yuan, M., Jiang, Z., Bi, G., Nomura, K., Liu, M., Wang, Y., et al. (2021) Pattern-recognition receptors are required for NLR-mediated plant immunity. Nature, 592, 105–109. Yuan, M., Ngou, B.P.M., Ding, P., and Xin, X.F. (2021) PTI-ETI crosstalk: an integrative view of plant immunity. Current Opinion in Plant Biology, 62. Zhang, X. and Dong, X. (2022) Life-or-death decisions in plant immunity. Current Opinion in Immunology, 75. Zhou, J.M. and Zhang, Y. (2020) Plant Immunity: Danger Perception and Signaling. Cell, 181, 978–989. FIGURES Figure 1: Experimental workflow for artificial inoculation and transcriptomic analysis of four grapevine genotypes (Rpv12 – goldenrod; Rpv12+1 – salmon; Rpv12+1+3 – cornflower blue; and Pinot Noir – black; as a susceptible control). For each genotype, three independent plants were used as biological replicates. At each time point (0, 6, and 24 hpi), three leaves were sampled per plant; leaves from the same plant were processed together to yield one RNA-seq library. In total, 36 libraries were sequenced (4 genotypes × 3 time points × 3 biological replicates). Artificial inoculation was performed by applying a 20 µl drop of spore suspension to the abaxial side of each leaf. Collected samples underwent genomic RNA isolation, RNA concentration and integrity (RIN) assessment, followed by RNA sequencing and subsequent bioinformatic evaluation. Figure 2 : Aggregated expression divergence and global transcriptional architecture across multilocus resistance backgrounds. (A) Empirical distributions of aggregated expression divergence (gray curves) derived from 220 random sample permutations are shown for 0, 6, and 24 hours post inoculation (hpi). Vertical lines represent observed divergence between resistant genotypes ( Rpv12 , Rpv12+1 , Rpv12+1+3 ) and the susceptible control. Colored symbols (goldenrod = Rpv12, salmon = Rpv12+1, and cornflower blue = Rpv12+1+3) denote mean divergence per genotype. Asterisks indicate significant divergence from the null model (FDR < 0.05). (B) Principal component analysis of 36 transcriptomes (three biological replicates per condition). Colors correspond to genotypes ( Rpv12 , Rpv12+1 , Rpv12+1+3 and grey= susceptible) and shapes to infection time (circle = 0 hpi, triangle = 6 hpi, square = 24 hpi). PC1 (42.5%) separates Rpv12+1+3 genotype from the others, PC2 (11.4%) reflects variation in the ETI pathways, PC3 (10.2%) separates the susceptible from the resistant genotypes, PC4 (7.37%) reflects variation in the L-pipecolic acid biosynthetic process (FDR=0.014). Gene Ontology enrichment is indicated for axes associated with major separations. Figure 3 : Temporal activation patterns of hub genes across metamodules and expression dynamics of PTI- and ETI-associated genes across resistant genotypes. (A) Heatmap showing simplified exprssion patterns (relative to the susceptible control) for 56 hub genes representing five higher-order co-expression metamodules (MM1–MM5). Rows correspond to resistant genotypes and infection time points (0, 6, 24 hpi); columns correspond to individual hub genes grouped by metamodule identity. Red and blue indicate up- and down-regulation, respectively (|log₂FC| > 1, FDR<0.05), and gray denotes no significant change. (B) Bar plots show mean log₂ fold-changes (± SE) relative to the susceptible control for genes classified as ETI-specific, PTI-specific, or PTI–ETI-shared. Colors denote resistant genotypes (Rpv12 = goldenrod; Rpv12+1 = salmon; Rpv12+1+3 = cornflower blue). ETI-specific and shared genes were associated with early upregulation in Rpv12 and Rpv12+1 at 0 hpi, followed by attenuation or downregulation at later stages, whereas Rpv12+1+3 showed overall downregulation. PTI-specific genes displayed weaker or inconsistent regulation. These patterns indicate that multilocus resistance reshapes the timing and amplitude of PTI–ETI transcriptional responses, supporting dynamic, rather than additive, integration of immune signaling layers. Figure 4 : Schematic representation of grapevine immune response layers and associated gene activity. The left panel depicts a simplified plant cell divided into three conceptual immunity layers: The recognition layer, which includes both pattern recognition and effector recognition; The signal-integration layer, where key defense-related pathways (SA, JA, ET, ABA, RONS) and transcriptional regulators coordinate downstream responses; and the defense-action layer, involving processes such as programmed cell death, etc. The right panel lists the identified genes/proteins, their assigned immunity layer, and the grapevine genotype showing the strongest expression, with matching color coding (Rpv12 - goldenrod, Rpv12+1 - salmon, Rpv12+1+3 - cornflower blue). Adapted with modification from Wang et al. (2019, Annual Review of Microbiology , 73:667–96, Figure 1). Figure 5. EDS1-centered transcriptional co-expression and predicted protein–protein interaction networks across immune layers. (A) Heatmap showing log₂ fold changes (relative to the susceptible control) of nine Rpv12-associated defense genes (EDS1, SAG101, SARD4, WRKY55, WRKY51, MIK2, SOBIR1, RUN1, ACD6) across genotypes and time points. (B) Pairwise expression correlations among these genes (Pearson r), highlighting tight transcriptional co-regulation within the EDS1–SAG101–WRKY–SARD4 cluster. (C) Co-expression network (r ≥ 0.65) derived from the Rpv12 genotype, colored by functional layer: Recognition (purple), Signal Integration (light blue), and Defense Action (yellow). (D) STRING-db protein–protein interaction (PPI) network for the same genes (confidence score ≥ 0.4; PPI p = 1.12 × 10⁻¹³). Significant enrichments include GO:0009862 – systemic acquired resistance (FDR = 0.0014), GO:0031349 – positive regulation of defense response (FDR = 5.13 × 10⁻⁵), and GO:0080183 – EDS1 disease-resistance complex (FDR = 0.0011). TABLES Rpv12 0 125 246 6 713 130 24 258 341 Rpv12+1 0 49 126 6 749 394 24 14 19 Rpv12+1+3 0 296 152 6 1421 396 24 148 119 Table 1: Numbers of differentially expressed genes (DEGs; FDR 1) across genotypes and time points. The number of DEGs peaked at 6 hpi in all genotypes, with a predominance of down-regulated genes. A Timing, genotype, direction All DEGs across resistant genotypes Timing significant; genotype, direction not p < 3.4 × 10⁻⁶; FDR < 0.05 Infection phase is primary driver of expression change Genotype p = 0.23 (ns) — Direction p = 0.55 (ns) — B Number of loci (1, 2, 3) Aggregated DEGs across genotypes Loci effect non-significant p = 0.40 (ns) No additive effect of stacked loci C Timing (within genotype) Rpv12 Weak timing effect p = 0.26 (ns) Stable moderate response Rpv12+1 Strong timing effect p < 2 × 10⁻¹⁶; FDR < 0.05 Early transcriptional burst at 6 hpi Rpv12+1+3 Strong timing effect p < 2.2 × 10⁻¹⁶; FDR < 0.05 Strong induction at 6 hpi, partial resolution by 24 hpi D Genotype (per timepoint) 0 hpi Genotype significant p = 0.024; FDR < 0.05 Pre-conditioning or baseline priming 6 hpi Genotype significant p = 0.0012; FDR < 0.05 Active immune divergence 24 hpi Genotype highly significant p < 2 × 10⁻¹⁶; FDR < 0.05 Full differentiation of immune trajectories 6 hpi (direction) Up vs. Down p < 4 × 10⁻¹¹; FDR < 0.05 Downregulation dominates during active defense E Timing (groups of gene categories) IEV (0–6 hpi) Groups *** (p I (p = 0.0061); Va, Vb, Vc ≫ I (p < 0.0001) Direction NS (p = 0.79) Strong genotype-specific effects; little up/down impact ER (6–24 hpi) Groups *** (p = 1.2 × 10⁻⁸)**; Vc ≫ others (p = 0.0016) Direction NS (p = 0.17) Double/triple-locus genotypes dominate early response TSR (6 hpi only) Groups *** (p = 7.5 × 10⁻⁵); Direction *** (p = 5.1 × 10⁻¹⁰)** Down ≫ Up 6 hpi burst dominated by downregulated genes in Vb and Vc LR (24 hpi) Groups *** (p < 2 × 10⁻¹⁶)**; IV, Va, Vc ≫ others (p < 10⁻⁶) Direction NS (p = 0.29) Strong late activation, especially in Rpv12 and Rpv12+1+3 SCh (0–24 hpi) Groups *** (p = 2.1 × 10⁻¹⁴)**; None (p ≈ 1.0) Direction NS (p ≈ 1.0) Few sustained DEGs; dominated by transient shifts F Full quasibinomial model All timings & groups Timing *** (p < 2.2 × 10⁻¹⁶); Direction ** (p = 6.0 × 10⁻⁵)**; Groups NS (p = 1.0) — Temporal phase and direction explain DEG proportions; grouping not significant Pairwise contrasts TRS < IEV (p < 0.0001); TRS < LR (p < 0.0001); Sch < TRS (p = 0.0003) Down ≫ Up TRS phase shows maximal transient repression Table 2: Statistical models describing the effects of infection timing, genotype, and direction on DEG dynamics across resistant grapevine genotypes (NS – nonsignificant). SUPPLEMENTARY FIGURES Supplementary Figure 1: Read-mapping quality metrics for all 36 RNA-seq libraries. The proportions of forward-only, reverse-only, dropped, and properly paired reads are shown per sample. Colors indicate sequencing batches. Generalized linear models were used to test for batch-associated differences (FDR values shown). These metrics were used to assess technical variation prior to normalization and batch correction. Supplementary Figure 2: Principal component analysis of the 36 RNA-seq libraries under different normalization and batch-correction steps. Panels show PCA on raw counts, raw counts after ComBat correction, size-factor normalized counts, and size-factor normalized counts after ComBat. Samples are colored by batch (top row) and by genotype with timepoints indicated by shape (bottom row). Reported r-values correspond to the correlation between PC1 and batch. These analyses were used to assess the influence of batch effects and the impact of correction procedures. Supplementary Figure 3: Principal component analysis of rlog-transformed expression values under different batch-handling strategies. Panels show PCA of rlog values without batch modelling, rlog values after ComBat correction, rlog values with batch modelled in DESeq2, and rlog values with both batch modelling and ComBat. Samples are colored by batch (top row) and by genotype with timepoints indicated by shape (bottom row). Reported r-values correspond to the correlation between PC1 and batch. These analyses were used to compare the effectiveness of alternative batch-correction approaches. Supplementary Figure 4: Gene contribution scores for the first four principal components based on rlog-transformed expression values. Each point represents a gene (n = 26,169). Dashed horizontal lines indicate the contribution threshold corresponding to the 50 highest-loading genes for each component. These summaries were used to identify genes with the strongest influence on major axes of variation. Supplementary Figure 5: Network density across timepoints based on pairwise gene–gene correlations above r > 0.817. Points indicate density estimates at 0, 6, and 24 hpi, with lines added for visualization. These values summarize global changes in coexpression connectivity over the infection time course. Supplementary Figure 6: Heatmap showing the proportion of shared genes among hub sets from the five coexpression meta-modules (MM1–MM5). Rows and columns represent hub genes, ordered by hierarchical clustering. Colors indicate the proportion of overlap between hub neighborhoods, with diagonal values reflecting within-hub self-similarity. Module membership is shown by the color bar on the left. This analysis was used to summarize relationships among meta-modules. Supplementary Figure 7: Overview of gene groups showing shared expression patterns across genotypes. Columns represent the major pattern groups (I–VI), with text indicating which genotypes contribute to each pattern and brief notes on their characteristics. Groups Va–Vc denote genes with genotype-specific behavior. This summary was used to guide interpretation of genotype-dependent expression trends. SUPPLEMENTARY TABLES Supplementary Table 1: Summary of sequencing and read-mapping quality metrics for all 36 RNA-seq libraries. Shown are the proportions of forward-only, reverse-only, dropped, and properly paired reads, together with additional mapping statistics used to assess technical variation between the two sequencing batches. These metrics supported the identification of batch-associated differences prior to normalization and correction. Supplementary Table 2: List of differentially expressed genes identified across genotypes and timepoints. For each gene, the table reports log₂ fold changes, statistical test results, and FDR-adjusted P-values, together with basic annotation information. These data provide the full results underlying the differential expression analyses. Supplementary Table 3: Comprehensive gene‐level dataset supporting all differential expression and grouping analyses. The table includes: (i) differential expression statistics for all genes (log₂ fold changes, P-values, FDR), (ii) size-factor–normalized and ComBat-corrected expression values for each sample, (iii) rlog-transformed and ComBat-corrected values, (iv) assignments to shared or genotype-specific expression groups, and (v) functional annotations (Ensembl, NCBI, UniProt identifiers; Arabidopsis orthologs; BLAST scores). Additionally, genes are grouped within each expression pattern group (Groups I–VI) and subdivided them by timing behavior. This dataset provides the full underlying results for all transcriptomic analyses. Supplementary Table 4: List of hub genes identified in the five coexpression meta-modules. For each hub, the table reports module membership, connectivity metrics, and associated annotations (gene identifiers, functional descriptions, and Arabidopsis orthologs where available). These data provide the full gene sets used for module-level summaries and overlap analyses. Supplementary Table 5: Summary of splicing-junction analyses across genotypes and timepoints. The table reports detected junctions, read support, assigned categories, and annotations, together with statistical comparisons and FDR-adjusted P-values. These data provide the full results underlying the splicing analyses described in the manuscript. Supplementary Table 6: Summary of coexpression network properties across genotypes and timepoints. The table includes network density estimates, edge counts, connectivity metrics, and module-level summaries based on the correlation threshold used in the analysis. These statistics provide the complete quantitative results supporting the network comparisons reported in the manuscript. Supplementary Table 7: Information for the five coexpression meta-modules (MM1–MM5). The table includes module sizes, member genes, hub status, enrichment summaries, and functional annotations. These data provide the full gene sets and metrics used in the meta-module analyses described in the manuscript. Supplementary Table 8: Lists of genes associated with pattern-triggered immunity (PTI) and effector-triggered immunity (ETI). The three sheets provide (i) curated PTI- and ETI-related gene sets, (ii) gene-level annotations and ortholog information, and (iii) module membership and expression summaries for these genes across genotypes and timepoints. These data were used to support the interpretation of immune pathway behavior. Supplementary Material File (table_1.xlsx) Download 8.20 KB File (table_2.xlsx) Download 7.69 KB Information & Authors Information Version history V1 Version 1 28 November 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords genome grapevine oomycetes pti-eti pyramiding Authors Affiliations Martin Hádlík Mendeleum—Institute of Genetics View all articles by this author Miroslav Baránek Mendeleum—Institute of Genetics View all articles by this author Kateřina Baránková Mendeleum—Institute of Genetics View all articles by this author Viera Kovacova 0000-0003-4581-4254 [email protected] University of Cologne View all articles by this author Metrics & Citations Metrics Article Usage 174 views 177 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Martin Hádlík, Miroslav Baránek, Kateřina Baránková, et al. Polygenic resistance is associated with altered early immune timing and changes in transcriptome network structure. Authorea . 28 November 2025. DOI: https://doi.org/10.22541/au.176431171.13982474/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.176431171.13982474/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fe8f35d9d8806eb',t:'MTc3OTI1NTAzOQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();