Evolutionary reconstruction, nomenclature and functional meta-analysis of the Kiwellin protein family

Crop diseases caused by pathogens critically affect global food security and plant ecology. Pathogens are well adapted to their host plants and have developed sophisticated mechanisms allowing successful colonization. Plants in turn have taken measures to counteract pathogen attacks resulting in an evolutionary arms race. Recent studies provided mechanistic insights into how two plant Kiwellin proteins from Zea mays mitigate the activity of the chorismate mutase Cmu1, a virulence factor secreted by the fungal pathogen Ustilago maydis during maize infection. Formerly identified as human allergens in kiwifruit, the biological function of Kiwellins is apparently linked to plant defense. We combined the analysis of proteome data with structural predictions to obtain a holistic overview of the Kiwellin protein family, that is subdivided into proteins with and without a N-terminal kissper domain. We found that Kiwellins are evolutionarily conserved in various plant species. At median five Kiwellin paralogs are encoded in each plant genome. Structural predictions revealed that Barwin-like proteins and Kiwellins cannot be discriminated purely at the sequence level. Our data shows that Kiwellins emerged in land plants (embryophyta) and are not present in fungi as suggested earlier. They evolved via three major duplication events that lead to clearly distinguishable subfamilies. We introduce a systematic Kiwellin nomenclature based on a detailed evolutionary reconstruction of this protein family. A meta-analysis of publicly available transcriptome data demonstrated that Kiwellins can be differentially regulated upon the interaction of plants with pathogens but also with symbionts. Furthermore, significant differences in Kiwellin expression levels dependent on tissues and cultivars were observed. In summary, our study sheds light on the evolution and regulation of a large protein family and provides a framework for a more detailed understanding of the molecular functions of Kiwellins.


WEBLOGO OF CONSENSUS SEQUENCES
: Weblogo of aligned consensus sequences (signal peptide trimmed) with secondary structure information of the Kiwellin groups Kissper-Kwl1, Kwl1 (Kiwellins without Kissper domain), Kwl2, and Kwl3 and a set of 391 BL proteins for reference. Green represents beta-sheets and blue alpha-helices. Numbered, yellow circles specify the cysteine residues forming disulfide bounds.

C)
A) B)
of the initial 620 proteins could be re-identified in the current release of UniProt. Next SignalP v5.0b (Almagro Armenteros et al., 2019) was used to predict and trim a leading signal peptide. If no signal peptide was found the leading residues were removed one by one until either a signal peptide was found or less than 50% of the original protein is left. If no signal peptide was found the protein is left unmodified. Alphafold v2.0.0 (Jumper et al., 2020) was used to predict the 3D structure using the database bfd database bfd metaclust clu complete id30 c90 final seq, mgnify database mgy clusters 2018 12 and references pdb70, uniclust30 2018 08 and uniref90. Finally pyMOL v2.5.2 (DeLano and Bromberg, 2004) was used for manual inspection and visualization.
In the following, we refer to the manually verified set of the 235 Kiwellins, 117 BL, and 59 unrelated proteins as round 1 and e.g. with round kissper 1 the set of Kissper-Kiwellins of round 1.

Advanced search (round 2 )
In the following, we frequently investigated if a certain descriptor falls into a range defined by round 1 . To allow more atypical values, we introduced a tolerance parameter of 25% to soften cutoffs, which is used throughout this pipeline.
Building on the knowledge from round 1 we scanned the UniProt database once more with a sophisticated pipeline named find kwl. This tool can be subdivided into 3 main steps: 1. Pre-filtering and Pre-processing 2. Collect descriptors 3. hmmsearch and filtering

Pre-filtering and Pre-processing
To reduce the search space and save computation time we filtered proteins first by sequence length. The signal peptide trimmed Kiwellins identified in round 1 contain between 150 and 227 amino acids (Kiwellins with and without Kissper domain) and at least 3 cysteine residues (including the BL proteins). To account for a possible signal peptide the length limit is extended by 90. Therefore, only proteins of lengths 150 − 317 ± 25% and at least 3 cysteine residues were initially considered. Those sequences were then trimmed using SignalP as described for the sequences of round 1 . The trimmed sequences were filtered again by length: 150 − 227 ± 25%.

Collect descriptors
Besides the sequence length and the number of cysteine residues, we wanted to evaluate the 3D structure. Using AlphaFold 3D structure predictions were generated for all proteins passing the pre-filter. As AlphaFold predictions do not include secondary structure, the dssp routine of the R package bio3d v2.4-1.9 (Grant et al., 2006)  Furthermore, we rated the 3D structure with a special focus on the different domains (barrel, kissper, clamp) of the putative Kiwellins. Thus, a set of reference structures from different species of round 1 was hand curated, i.e. multiple reference structures were used to combat a possible underfitting: Han et al. (2019) The extracted structures for the kissper and clamp domain were reduced to the first 60 residues and the one for the barrel domain to the 6 β-sheets β1, ...β6.
The structure prediction was superimposed using PyMOL with the set of references to calculate the RMSD (the lower the better) and the number of matching atoms (MA, the higher the better). The RMSD can be arbitrarily small even for unrelated proteins with short overlaps (low MA) and a low RMSD does not necessarily follow from a high number of matching atoms. Therefore we combined both values and define The lower RMSDPMAS the better the superimposition as shown in Fig. S4. With that, the smallest RMSDPMAS was determined for each set of reference domains (later referred to as kissper, clamp, and barrel RMSDPMAS). To ensure comparability with the kissper and clamp domain only the leading 60 residues of a protein were compared to the reference structures. As shown in Fig. S5 this leads to a reliable identification of the Kiwellins with and without a kissper domain using the kissper RMSDPMAS and the clamp RMSDPMAS respectively. Additionally, the barrel and ZmKWL1a RMSDPMAS were used to exclude the unrelated proteins. Figure S4: Superimposition of Kwl3-1b (A0A1D6GNR3, crystal structure of Han et al. (2019)) in red and blue the proteins A) W1PCT0, B) Kwl2-2c (T1NDN2), C) A0A3B6PSY1 and D) A0A7J9C558. A) shows an alignment with a high RMSD and low MA values resulting in a high RMSDPMAS. In contrast that the very similar structures of B) result in a low RMSDPMAS. In comparison the non-optimal alignments of C (short overlap with high similarity) and D (long overlap with low similarity) with similar RMSD or MA values, respectively both result in a higher RMSDPMAS.  Although the Kissper-Kiwellins contain a clamp domain the clamp RMSDPMAS is usually orders of magnitudes worse than the Kiwellins without the kissper domain since the first 60 residues do usually not contain the clamp.
To summarize, we collect the following descriptors for any protein that passes the pre-filter checks of 4.2.1: • length of sequence after signalp trimming • number of cysteine residues • the number of continuous regions of β-sheets b regions • the structure scores RMSDPMAS against the set of reference structures

hmmsearch and filter
For each of the curated sets of round 1 (Kiwellins, Kissper-Kiwellins, and BLs) first an alignment was generated using a muscle. Next, those columns were removed that almost only consisted of gaps (> 90%), and a hidden Markov model (HMM) was assembled using HMMer v3.2.1 (Eddy, 1998). hmmsearch was used to query the 3 models against all reference proteomes of UniProt. The resulting E-value of a match (m, p) between a model m and a protein p is denoted by E (m,p) . To determine a suitable E-value cutoff for each model the 59 as unrelated identified proteins of round 1 were used as a negative control set: Next, we wanted to assess if a descriptor (e.g. number of cysteine residues) is untypical compared to the values of round 1 with a predefined tolerance parameter of t = 25%. For that we will say that a descriptor is part of round 1 if the value v lies in the range [mi, ma] of values of round 1 extended by the tolerance parameter t: On the same note we defined that a descriptor is at least or below round 1 for only the left or right inequality respectively. Furthermore, we defined a set of filtered matches as the subset of all reported matches (hmmsearch) that fulfill the following set of rules: • the E-value below the defined model specific cutoff: E (m,p) ≤ c m For each protein among the set of filtered hits, we report the model with minimal E-value as the best match for that protein.
In total 683 new Kiwellins were found, i.e. 59 with and 589 without a kissper domain. The steps 2 − 3 of the pipeline were repeated with the new extended set of Kiwellins as the input (alignment, HMM model, descriptor ranges), and 15 further Kiwellins were identified. In a final step, this set of Kiwellins was checked by hand again and almost all entries (98.2%) could be verified as correctly classified. We removed only 17 entries. Most of which were faulty Kiwellins (missing β7). A final set of 915 Kiwellins (62 with and 772 without a kissper domain) were reported and are denoted as round 2 in Fig. S5. Han et al. (2019) Figure S6: Flow chart of the Kiwellin identification pipeline. More details on the blue-marked processing steps are described in the respective chapters.

DETAILED RNA-SEQ RESULTS
To get an idea of the functions of Kiwellins we re-analyzed publicly available RNA-seq data sets from the NCBI SRA. 70 experiments were obtained using the following filter characteristics: RNAseq, RNA, stress keywords (pathogenic, symbiotic, water, ...), and the scientific name of the plants. The 70 data sets were checked on quality parameters of the raw data using fastQC (Andrews et al., 2010) and on data integrity (at least 2 replicates, associated publication, unambiguous sample naming, and experimental descriptions) resulting in 31 data sets. To determine if a Kiwellin was significantly regulated an FDR threshold of 5% was used. We consider an entry to be differentially expressed if the absolute log2 transformed fold changes (L2FC) is above 1 and the P-value is below the above FDR threshold. Furthermore, we define a Kiwellin group as strongly expressed if the baseMean (baseM) is at least 80 (a proxy for the overall expression strength; log 10 (80)≈1.9). Finally, we grouped the experiments by experimental parameters (pathogenic, symbiotic, abiotic, tissue-specific responses  Figure S7: Sankey diagramm illustrates from left to right the workflow for re-evaluated RNA-seq data. P: pathogenic, S: symbiotic, A: abiotic, T: tissue specific responses. QC: Quality check. More information can be found in the chapter 5. The number of analysed experiments per species are given on the right side. Brackets indicate non signifiant interactions not significant (e.g. S(+P): significant response only to the symbiotic and not to pathogenic partner).

NCBI SRA
In the following chapters, we shortly describe the 31 analyzed case studies and examine the regulation of the Kiwellins groups. Groups were formed from indistinguishable Kiwellins concerning the associated transcript(s) as described in the Material and Methods. A Kiwellin group can include multiple proteins as well as transcripts. For example lets consider the kiwellin group 'Kwl2-2t,2v,2s' of T. aestivum shown in PRJNA743515. This group includes the three almost identical Kiwellin proteins Kwl2-2t, Kwl2-2v, Kwl2-2, that share 188/197 identical residues. With the help of proteinortho two similar transcripts (XM 044541154.1, XM 044500936.1) were identified from the respective transcriptome. Since the proteins as well as the transcripts are almost indistinguishable, we combine the results of this group into one entry. All identified groups of transcripts are listed in the table "Nomenclature-KWL" and the used transcriptomes in "transcriptome sources" of Supplementary Data 3 (proteinortho transcripts).
For each experiment, a heatmap is shown to visualize the L2FC on the left panel from blue (downregulated) to red (up-regulated). Gray is shown if the comparison does not exhibit significant changes. The middle panel specifies the log10 transformed baseM. A * symbol in the name indicates that the Kiwellin group surpasses the baseMean threshold of 80 and thus is considered to be strongly expressed. More details can be found in the Material and Methods. The right panel shows average normalized counts as well as standard deviations between the replicates of all conditions.
The results of the 31 analyzed experiments were divided into sections according to Fig. S7

Triticum aestivum (PRJNA743515)
Bipolaris sorokiniana is a hemibiotrophic fungus responsible for several plant diseases. The study Zhang et al. (2022) aimed to investigate how genes are regulated when Triticum aestivum is infected by pathogenic fungus (TAB). Uninfected plants (TA) served as control groups. Plants were soil-inoculated and samples of root and basal stems were harvested 5 and 15 days after infection. RNA was isolated from the samples and sequenced.
We found 4 strongly expressed groups (*-prefix), three from Kwl1 and one from Kwl3, and 6 further weak expressed groups of Kwl2. We found that one group of the weakly expressed Kwl2 to be differentially regulated 5 dpi (3 L2FC). One group of Kwl1 showed a slight down-regulation late in the infection stage (−1.2 L2FC) and remained unchanged at 5 dpi.

Oryza sativa (PRJNA325291)
In Huang et al. (2017), rice was inoculated with the fungus Magnaporthe oryzae (with=Guy, without=Before). It is known that nitrogen fertilization increases the effects of many diseases. The authors studies whether the external addition or omission of nitrogen led to differentially expressed genes during infection in both species to explain Nitrogen-Induced Susceptibility (NIS). For this purpose, rice plants were infected with water or the fungus and 0 dpi or 2 dpi shoot tissue of the plants were harvested. Nitrogen was omitted from the fertilizer in one series of experiments (0N) and added in the form of ammonium nitrate in another (1N). Subsequently, mRNA was isolated from the obtained tissue and analyzed.
We found 2 groups of two Kwl1 and one Kwl3 to be highly expressed. The kwl3 group shows no regulation in response to the infection. One Kwl1 group was down-regulated (4 L2FC), and in the second group, a slight up-regulation upon infection under nitrate treatment was found. No ammonium nitrate-specific response was observed among all groups.

Cucumis sativus (PRJNA285071)
In Burkhardt and Day (2016), a resistant strain (PI197088) and an susceptible strain (Vlaspik) of Cucumis sativus were infected with the fungus Pseudoperonospora cubensis and water (mock), respectively. Leaves of the plant were harvested 1, 2, 3, 4, and 6 dpi and mRNA levels were determined in each case.
We found a group of Kissper-Kiwellins (Kwl1) to be strongly expressed in the susceptible strain and moderately in the resistant strain. Furthermore, we found a strong up-regulation upon infection (≈ 4 − 6 L2FC) in the susceptible strain throughout infection, while the resistant strain showed a slight dampening of the differential response (≈ 3 − 4 L2FC). Furthermore, the response vanishes at 6 hpi for the resistant strain.

Glycine max (PRJNA412201)
It is known that silicon can protect plants from biotrophic and hemibiotrophic pathogens. To better understand this mechanism, Glycine max was infected with Phytophthora sojae in Rasoolizadeh et al. (2018). Silicon was added in one case (SiPlus) and omitted (SiMinus) in the other plants. After 21 days of infection, root samples were collected and mRNA was isolated and sequenced.
We found one group of Kissper-Kiwellins (Kwl1) to be highly expressed with a differential response to the infection (≈ 1 − 3 L2FC). The silicon treatment slightly reduced effect (≈ 1 L2FC less).  In Gamez et al. (2019) seedlings of Musa acuminata were inoculated with two species of growthpromoting rhizobacteria: Bacillus amyloliquefaciens (Ba) and Pseudomonas fluorescens (Pf). 1 hpi, 2 dpi and 4 dpi whole seedlings were collected, and the mRNAs were isolated. These data sets were compared with water-inoculated seedlings.
We detected three Kiwellins groups of Kwl2. Kwl2-1b was the only strong expressed group and showed a weak up-regulation upon infection with Pf after 1 hpi and remains inconspicuous otherwise. The other two Kiwellins showed no differential response to the infection.

Zea mays (PRJNA506746)
In Shen et al. (2020), the cadmium tolerance of Zea mays roots was investigated, which previously treated with the endophyte Exophiala pisciphila. Roots of three-day-old maize seedlings were first inoculated with the fungus (with=DSE, without=nDSE). 10 days later, plants were fertilized with cadmium for 31 days (with=Cd, without=nCd). Plants not treated with cadmium and/or the fungus served as the control. Finally, roots were harvested and mRNA was extracted and analyzed.
We found one group of Kwl3 to be strongly expressed. In the case where the fungi were absent, this group show a strong down-regulation with cadmium (4 L2FC) but no change upon fungal treatment was detectable. If cadmium is absent, the infection does not significantly impact expression but if cadmium is introduced into the system we see an up-regulation (3.5 L2FC) in infected plants.

Cucumis sativus (PRJNA445328)
To better understand Trichoderma-induced plant resistance to many plant pathogens, cucumber plants were infected with Botrytis cinerea in the presence or absence of Trichoderma in Yuan et al. (2019). At the three-leaf stage, plants were inoculated with Trichoderma and 24 hours later B. cinerea was injected into the leaves. Samples of the leaves were harvested 96 hours later and examined for differential gene expression.
We detected 3 groups of kwl1 Kissper-Kiwellins. One group was highly expressed and showed a significant up-regulation in response to the symbiont and pathogen (≈ 2.5 L2FC).

Triticum aestivum (PRJEB21874)
Triticum aestivum was infected with the bacterial pathogenic Xanthomonas translucens in Fiorilli et al. (2018). It was tested whether the mycorhizal fungus Funneliformis mossae influenced the infection. After plants were colonized by mycorrhiza for 49 days, plants were inoculated with the pathogenic bacterium. One day after infection, samples of roots and leaves were isolated and the mRNA levels of the three species were examined. We found 8 highly expressed Kiwellin groups most of which belong to Kwl2 but and to Kwl3. The Kwl3 group showed no significant response and the results for Kwl2 were mixed. In roots, we observed one group of Kwl2 to be down-regulated (≈ 2 L2FC) and one group to be up-regulated (1 L2FC) in response to the pathogen and symbiont. Differences between roots and leaves can be observed for Kwl1, Kwl2, and Kwl3. Overall the expression strength in roots was observed to be higher compared to leaves.

Medicago truncatula (PRJNA524006)
In Sańko-Sawczenko et al. (2019), the Fabaceae Medicago truncatula was evaluated for their response to water stress when the roots were inoculated by nitrogen-fixing bacteria Sinorhizobium meliloti. After the roots were successfully colonized by the bacteria, the plants were subjected to water stress. For this, the plants were not watered for up to 4 days after colonization. At the end of the four days, root nodules were harvested from watered and non-watered plants. Uninfected plants served as control, here the roots were harvested. The mRNA was isolated from the collected samples and analyzed.
We found one highly expressed group of Kwl3 that shows a strong up-regulation after 4 days of water withdrawal (2 L2FC).

Solanum lycopersicum (PRJNA795851)
Biological control agents (BCA) play a major role to combat plant pathogens. Singh et al. (2021) aimed to investigate the transcriptonal response to treat with the BCA fungus Chaetomium globosum (Cg) on plants infected with the pathogenic fungi Alternaria solani (As). First, 21-day-old tomato plants were inoculated with the BCA. Another 24 hours later, the plants were spray-inoculated with the pathogen. After five days of infection, infected leaves were harvested, and RNA was isolated and sequenced. In total, four data sets resulted from this experiment: plants not infected (CONTROL), plants infected with both fungi (Cg As inoculated), and plants infected with only one fungus each (Cg inoculated, As inoculated).
We found one group of Kwl3 to be highly expressed. An up-regulation could be observed in case of infection with the BCA fungus C. globosum (≈ 4 L2FC). Furthermore, a slight down-regulation (below 1 L2FC) was observed upon infection with the pathogen A. solani. In response to combinatorical treatment with the pathogen, a slight up-regulation (below 1 L2FC) was observed.

Medicago truncatula (PRJNA79233)
In Boscari et al. (2013), the transcription of developing nodules on Medicago truncatula was investigated. For this purpose, plants were infected with their symbiont Sinorhizobium meliloti and samples were taken from different stages of the nodules/roots and the mRNA was isolated and sequenced. Roots of the plant that were not infected were collected 4 days after infection (developing nodules), and 12 days after infection (matured nodules) were examined. Furthermore, a nitric oxide scavenger (cPTIO) was added to an infected plant and the effect on nodules was studied. Thus, a total of four data sets were obtained and compared: unified roots (MtRoots), infected roots (MtInoc), nodules (MtNod), and infected roots treated with cPTIO (MtInocCPTIO). All samples except the nodule sample were collected and analyzed 4 days after infection or mock infection.
We found 2 groups of Kwl3 to be strongly expressed but no difference between infected and non-infected roots (cPTIO independent) was observed. Remarkably, Kiwellins were almost exclusively found in nodules.

Musa acuminata (PRJNA417328)
Benzothiadiazole (BTH) is an inducer of plant resitance that stimulates the defense response in bananas and protects against pathogen infection. In Cheng et al. (2018), via RNA-seq, the effect of BTH was investigated at the gene expression level by spraying young plants with a BTH solution. For this purpose, plant samples from roots (RT) and leaves (LF) 1 and 3 days post-infection with the fungal pathogen Fusarium oxysporum were compared with their respective controls (0 dpi). We found one group of highly expressed Kwl2 members (in roots but almost absent in leaves). Furthermore, no significant changes were observed in response to BTH. Kiwellins were almost exclusively found in roots (compared to leaves).   ). For this purpose, plants were treated with or without ASM and inoculated with the bacterium or buffer. 3, 24, and 48 hpi samples were obtained, and the mRNA was isolated from ground tissue and analyzed.
We found 4 groups of Kiwellins of which 2 are Kissper-Kiwellins (Kwl1) and two belong to Kwl3. The groups of kwl1 are moderately expressed in one group and we see an ASM-specific up-regulation of ≈ 1 − 2 L2FC. In all comparisons without ASM, no effects are were observed throughout the infection.  We found a group of two Kissper-Kiwellins from Kwl1 (here called kissper-kwl0nsp). Although this group was not strongly expressed we found a response to the infection at both time points (2 − 3 L2FC).

Zea mays (PRJNA529541)
Garcia-Ceron et al. (2021) infected Zea mays with Fusarium graminearum and examined the change in gene expression. For this purpose, the leaves of the plants were injured and disk-infected with the fungus. Leaf tissue was collected after 3, 5, and, 7 days respectively, and the mRNA was examined. The Comparison was made with uninfected plants and fungi grown in axenic culture. No Kiwellin was found to be highly expressed but a down-regulation for a group of Kwl2 and a group of Kwl3 was observed upon infection (≈ 2 − 4 L2FC).

Zea mays (PRJNA415355)
In Li et al. (2018b), mRNA levels were examined organ-specifically during tumor development of Ustilago maydis on Zea mays. For this purpose, data sets of bundle sheaths and the mesophyll of fungusinfected plants were compared with water-inoculated plants (mock). We found no strongly expressed Kiwellin groups but an up-regulation (≈ 2 − 4 L2FC) of Kwl3 in infected bundle sheath and mesophyll tissue. For this purpose, Glycine max was infected with the bacterium Bradyrhizobium diazoefficiens. The infected root tissue on which nodules were formed was harvested at 5 − 7 days after infection (emerging nodules) or 14 − 16 days after infection (mature nodules). Root tissue was collected above and/or below the nodules as control groups.
We found no strongly expressed Kiwellin group but differences were found e.g. between emerging and mature nodules as well as between mature nodules and uninfected roots (NA root).

Cucumis melo (PRJEB15551)
Two strains of Cucumis melo were infected with Fusarium oxysporum f. sp. melonis Snyd. & Hans race 1.2 (FOM1.2) in Silvia Sebastiani et al. (2017). One of the melon lines is the NAD strain, which is capable of early recognition of pathogens and developing resistances. The second melon genotype Charentais (CHT) is susceptible to the fungus. Plantlets of the two strains were infected with the fungus and 1 and 2 days. Stems of the small plants were harvested and mRNA levels were determined and compared.
In our reanalysis, we found one highly expressed group of Kwl1 (Kissper-Kiwellin). None of the groups are significantly differentially expressed.

Solanum tuberosum (PRJNA755645)
Alternaria solani is a necrotrophic fungus that infects potatoes and other crops. The aim of Brouwer et al. (2021) was to investigate how the transcriptome of Solanum tuberosum changes during infection with this fungus. For this purpose, the leaves of six-week-old potato plants were infected with the pathogen. Samples of infected leaves were collected 1, 6, 12, 24, and 48 hours after infection, and the mRNA was analyzed and sequenced. Uninfected plants served as the control group (0 hpi).
Our reanalysis was able to identify Kiwellins 3 groups of Kissper-Kwl1. One group can be considered as strongly expressed but no response to the infection was observed.   Tian et al. (2019) investigated the differences between wild rice (Oryza rufipogon) and cultivated rice (Oryza sativa) inoculated with the arbuscular mycorrhizal fungus Rhizoglomus intraradices upon infection with the pathogen Magnaporthe oryzae. For this purpose, ten-day-old rice plants were first inoculated with the mycorrhizal fungus, and after another 45 days, the leaves of the plants were spray-inoculated with the pathogen. After another seven days, the roots of the plants were harvested and the RNA was isolated and analyzed.
For both Oryza species we found a group of Kwl3 to be highly expressed but no significant differential changes were observed. Roots of Glycine max strain EN1282 (nfr1-mutant -a strain lacking in a Nod factor receptor) was infected with the symbiotic bacterium Bradyrhizobium elkanii USDA61 in Ratu et al. (2021). The wild-type strain of the bacterium was compared with a T3SS (Type 3 Secretion system) deletion strain. The roots of the seedlings were harvested 30 days after infection and mRNA levels of the bacterium and the plant were measured.
We found 3 groups of Kiwellins of which 2 belong to Kwl3 and one Kissper-Kiwellins to Kwl1. The Kwl3 groups were weakly expressed and Kissper-Kiwellin showed a moderate expression. Furthermore, all groups in this experiment showed no significant response to the infection.

Glycine max (PRJNA396797)
To investigate transcriptional changes associated with nodule formation genes in soybean, Glycine max roots were infected with Bradyrhizobium japonicum in Hayashi et al. (2012). Two strains of the bacterium were used and compared: a wild-type strain and a NodC strain that cannot synthesize Nod factors. The infected roots were harvested at 2 dpi, the mRNA was analyzed and compared.
We found a Kwl1 (Kissper-Kiwellins) and one Kwl3 group. No group was strongly expressed or differentially regulated in this experimental setup.

Glycine max (PRJNA579169)
Using the Rj2 allele, soybean plants can exclude poorly nitrogen-fixing or less useful rhizobia such as B. japonicum USDA122 or Rhizobium fredii USDA257 from a symbiotic relationship. Host immunity is mediated by the secretory rhizobium type-III-protein NopP and the previously described host resistance protein Rj2. In Shine et al. (2019) transcriptional changes in leaves of Rj2 virus-silenced plants, each infected with buffer, or one of the two rhizobacteria, will be used to better understand the mechanism of systemic resistance induced by incompatible rhizobia. For this purpose, infected roots were harvested and the mRNA was isolated and analyzed.
We found 2 groups of weakly expressed Kiwellins (Kissper-Kiwellins of Kwl1 and Kwl3). But in all comparisons, no differential regulation was detected.

Solanum lycopersicum (PRJNA531604)
In Li et al. (2018a) the influence of the endophyte Pochonia chlamydosporia on the response of Solanum lycopersicum was investigated. For this purpose, plants were infected with the fungus, and samples of the roots were harvested 4, 7, and 21 days after infection. From these tissue samples, mRNA was isolated and analyzed.
No Kiwellin was found to be highly expressed and no differential regulation was observed in response to the fungi.

Musa acuminata (PRJNA287860)
Roots of two-month-old banana seedlings were infected with Fusarium oxysporum Race 4 (FocR4)-C1 HIR in Munusamy and Zaidi (2021). Infected root samples were harvested at 2, 48, and 96 hours. The 2 hpi root sample represents the control with which the other two samples were compared.
Our reanalysis found 3 weakly to moderately expressed Kiwellins belonging to Kwl2. Neither of these groups shows differential regulation in this experiment.

Zea mays (PRJNA551023)
Pantoea stewartii is the causal agent of Stewart's bacterial wilt of corn and is investigated in Doblas-Ibáñez et al. (2019). With the help of a mutation in the pan1 gene, it is possible to create resistant corn plants to bacterial disease. Consequently, heterozygous and homozygous (related to the pan1 gene) maize lines were created by crosses and infected with the bacterium. Subsequently, infected material was harvested one day post-infection, mRNA was isolated, and differences in transcription levels between the different maize lines infected or mock-infected were analyzed.
We found Kwl3-1a and Kwl2-2e but both were neither strongly expressed nor showed any differential response.

Solanum lycopersicum (PRJNA487149)
In Fawke et al. (2019), the influence of glycerol-3-phosphate acyltransferases on the resistance of Solanum lycopersicum to its host Phytophthora infestans was investigated. For this purpose, tomato wild-type plants and plants with a loss-of-function mutation in the gpat6 gene were infected with the fungus. Three days after infection, the leaves were harvested, the RNA isolated, reverse transcribed and the data analyzed.
We found one member of the Kwl3 group that was neither strongly expressed nor showed a differential response.

Chenopodium quinoa (PRJNA720675)
In Rollano-Peñaloza et al. (2021), the authors aimed to investigate the influence of Trichoderma on Chenopodium quinoa. For this purpose, two strains each of the fungi Trichoderma afroharzianum (T22) and Trichoderma harzianum (BOL-12) and the plant (Chenopodium quinoa Kurmi and Chenopodium quinoa Real) were co-cultured with each other. Subsequently, RNA was extracted from the roots and sequenced to determine the differentially regulated genes in the 4 strains.
We found 2 Kiwellin groups of Kwl3 but no group was found to be highly expressed and no differential regulation was observed.

Triticum aestivum (PRJEB8798)
In Rudd et al. (2015) wheat was infected with its fungal pathogen Zymoseptoria tritici and mRNAs were isolated from leaves 1, 4, 9 and 14 dpi (infected=Z.tritici, mock innoculated=M). This data set was compared with mRNAs from buffer-infected plants and fungus growing in liquid culture.