A conserved fungal Knr4/Smi1 protein is vital for maintaining cell wall integrity and host plant pathogenesis

Filamentous plant pathogenic fungi pose significant threats to global food security, particularly through diseases like Fusarium Head Blight (FHB) and Septoria Tritici Blotch (STB) which affects cereals. With mounting challenges in fungal control and increasing restrictions on fungicide use due to environmental concerns, there is an urgent need for innovative control strategies. Here, we present a comprehensive analysis of the stage-specific infection process of Fusarium graminearum in wheat spikes by generating a dual weighted gene co-expression network (WGCN). Notably, the network contained a mycotoxin-enriched fungal module that exhibited a significant correlation with a detoxification gene-enriched wheat module. This correlation in gene expression was validated through quantitative PCR.

By examining a fungal module with genes highly expressed during early symptomless infection, we identified a gene encoding FgKnr4, a protein containing a Knr4/Smi1 disordered domain. Through comprehensive analysis, we confirmed the pivotal role of FgKnr4 in various biological processes, including morphogenesis, growth, cell wall stress tolerance, and pathogenicity. Further studies confirmed the observed phenotypes are partially due to the involvement of FgKnr4 in regulating the fungal cell wall integrity pathway by modulating the phosphorylation of the MAP-kinase MGV1. Orthologues of FgKnr4 are widespread across the fungal kingdom but are absent in other Eukaryotes, suggesting the protein has potential as a promising intervention target. Encouragingly, the restricted growth and highly reduced virulence phenotypes observed for ΔFgknr4 were replicated upon deletion of the orthologous gene in the wheat fungal pathogen Zymoseptoria tritici. Overall, this study demonstrates the utility of an integrated network-level analytical approach to pinpoint genes of high interest to pathogenesis and disease control.

Gene co-expression network analysis
RNAseq reads from Dilks et al. (2019) were provided by Dr Neil Brown (European Nucleotide Archive: PRJEB75530). Read quality was assessed with FastQC v. 0.11.9 (Andrews, 2010). Reads were mapped to a combined Fusarium – wheat genome, consisting of v. 5 of the Fusarium graminearum PH-1 genome (King et al., 2017) and the high confidence (HC) transcripts of the v. 2.1 of the International Wheat Genome Sequencing Consortium (IWGSC) Triticum aestivum genome (Zhu et al., 2021). Genome indexing and read alignment were performed using STAR aligner 2.7.8a. Soft clipping was turned off to prevent reads incorrectly mapping to similar regions of the highly duplicated hexaploid wheat genome. Reads were filtered using the filterByExpr function part of the R package Edge R v.3.32.1 (Robinson et al., 2010). Counts were normalised separately for fungal and wheat reads by performing a variance stabilising transformation (VST) using the DESeq2 v 1.30.1 R package (Love et al., 2014) in R (v4.0. 2, https://www.r-project.org/).

The VST normalised counts were filtered to remove any excessive missing values using the function goodSamplesGenesMS in the WGCNA R package (Langfelder and Horvath, 2008). Standard methods were implemented to generate the network using the WGCNA R package, with the following parameters. A signed-hybrid network was constructed using the filtered counts. The soft thresholding power (β) was uniquely selected per network according to scale free model criteria (Zhang and Horvath, 2005), where β = 9 for the fungal network and β = 18 for the wheat network (Figure 2 – figure supplement 2). A deepSplit of 3 was paired with a standard cutheight of 0.25. A minimum module size of 50 was selected to minimise potential transcriptional noise when assigning modules using smaller datasets (Oldham, 2014; Walsh et al., 2016). The function multiSetMEs from the WGCNA package was used to calculate module eigengene expression. Module eigengenes with similar expression profiles were then merged.

Module quality and preservation was calculated using the function modulePreservation present in the WGCNA R package (Langfelder and Horvath, 2008; Langfelder et al., 2011). When calculating module preservation, the original wheat or fungal network was considered the reference network. Then 50 different test networks were created, each built upon randomly resampling (with replacement) a proportion of samples from the original dataset. The average preservation metrics (i.e. Z-score) between the original network and the 50 test networks was calculated for both the fungal and wheat networks.

Module Enrichment an Annotation
Gene ontology (GO) annotations of the v. 5 PH-1 genome (GCA_900044135.1) were generated using Blast2GO v .5 (Götz et al., 2008). Enrichment was calculated using a background set of all genes present in the fungal network. GO annotations for the IWGSC v.2.1 genome were provided by Dr Keywan Hassani-Pak of the KnetMiner team (Hassani-Pak et al., 2021). This was generated by performing a BLASTx search on the NCBI nb database using DIAMOND v 2.0.13-GCC-11.2.0 (Buchfink et al., 2015), then Blast2GO v.5 was used to annotate the BLAST hits with GO terms. GO term enrichment was calculated for each high level GO ontology (Biological Process, Molecular Function and Cellular Component) using the R package topGO v 2.46.0 (Alexa and Rahnenfuhrer, 2009).

Plant Trait Ontology (TO) (Cooper et al., 2024) enrichment analysis was performed using annotations derived from the KnetMiner knowledge graph (release 51) for wheat (Hassani-Pak et al., 2021) and KnetMiner datasets and enrichment analysis notebooks are available at https://github.com/Rothamsted/knetgraphs-gene-traits/. Predicted effectors were determined using EffectorP v.3.0 (Sperschneider and Dodds, 2022). Alongside this, predictions to identify extracellularly localised genes were done using SignalP v6.0 (Teufel et al., 2022). Custom F. graminearum gene set enrichment of the network modules was calculated by performing a Fisher’s exact test using all the genes in the fungal network as the background gene set. A BH correction was calculated for both GO and custom enrichments (Benjamini and Hochberg, 1995). Modules were deemed significantly enriched if P-corr < 0.05.

Gene lists included in GSEA consisted of predicted secreted effector proteins, alongside known gene families associated with virulence, such as biological metabolite clusters (BMCs) (Sieber et al., 2014), polyketide synthases (Gaffoor et al., 2005), protein kinases (Wang et al., 2011) and transcription factors (Son et al., 2011). Due to their well-established importance in F. graminearum pathology, a separate enrichment for genes of the TRI gene cluster was also performed.

Annotation from PHI-base was obtained by mapping genes to version PHI-base (v4.16) annotation using UniProt gene IDs and any through Decypher Tera-Blast™ P (TimeLogic, Inc. Carlsbad, California, USA) (E-value = 0) against the PHI-base (v4.16) BLAST database (Cuzick et al., 2023).