kexB Gene Correlates With Aspergillus flavus Keratitis Severity: A Whole-Genome Analysis

Abstract

Purpose: Fungal keratitis caused by Aspergillus flavus (A. flavus) can result in severe inflammation and corneal stromal melting, leading to visual impairment. This study aimed to identify virulent genes correlated with the severity of A. flavus keratitis using whole-genome sequencing.

Methods: Whole-genome sequencing of 21 clinical A. flavus strains from cornea was performed to elucidate the pathogenesis of A. flavus in infectious keratitis, followed by pan-genome analysis and virulence analysis. To further understand the results from the previous analyses, growth phenotypes and virulence effect of mutant strains were validated experimentally, including the spore counting, growth pattern under different conditions, and clinical and pathological evaluation of A. flavus keratitis in mice models.

Results: The A. flavus pan-genome was composed of 17,326 gene clusters with a core genome of 5378 (31.0% of the pan-genome) orthogroups in all 21 isolates. Virulence gene analysis revealed 183 genes contributing to A. flavus pathogenesis and mutation of the kexB gene was associated with the severity of keratitis. The kexB mutant (ΔkexB) strains showed significantly reduced conidia formation and lower growth rates in the presence of the cell-wall perturbing agents. Further, the mice models validated that clinical score, and corneal perforation rates significantly decreased in the group infected by ΔkexB strains. Infiltration of immune cells, gene expression of cytokines, and matrix metalloproteinase (MMP) were also decreased in the mutant group.

Conclusions: The role of kexB gene in A. flavus keratitis was identified through whole-genome sequencing. Its mutation impairs conidia formation, cell wall integrity, and invasion, leading to milder clinical symptoms.

Fungal keratitis annually afflicts millions of people globally, with 10% of cases culminating in eye removal due to delayed diagnosis or suboptimal therapeutic outcomes.¹ Typically, the infection occurs when environmental pathogens invade the cornea via a wound subsequent to corneal trauma. Aspergillus species, with their ubiquitous airborne conidia, are significant causative agents, particularly in tropical regions.² Among them, Aspergillus flavus (A. flavus), as one of the most commonly isolated pathogens in fungal infection, is estimated to constitute 75% of keratitis cases within the Aspergillus genus owing to its invasive propensity and production of mycotoxins.^3,4 The potent pathogenicity and rapid disease advancement of A. flavus can lead to severe vision impairment or even blindness without prompt diagnosis and treatment.⁵ Our previous study had indicated that the incidence of Aspergillus keratitis has been steadily ascending in recent years, imposing a substantial burden on individuals and society.⁶ 

The host immune status is of paramount importance for the prognosis of A. flavus keratitis. Concurrently, genetic and phenotypic variations among strains also exert a considerable influence.^7,8 Factors such as thermotolerance, secondary metabolites, cell wall integrity, stress response, and transcription act as determinants of virulence in both plant and animal models.^9–12 Additionally, Barakat et al. found that AFB1 toxin–encoding gene was detected in more than half of the isolates, and there was a significant correlation between severity of keratitis and the quantity of aflatoxin B1 production.¹³ These findings underscore the connection between strain-level genetics and clinical severity, highlighting the necessity for further investigation into the pathogen features isolated from patients. 

Pan-genome analysis, which centers on comprehensive genetic diversity by utilizing multiple isolates rather than a single strain, has emerged as an effective approach to capture the extensive diversity within a species and conduct in-depth investigation on novel gene clusters and their regulation. Amelia et al. identified genomic variation associated with clinical isolates and triazole resistance and characterized genetic variation in known virulence factors through A. fumigatus pan-genome analysis.⁸ Gangurde et al. provided a complete pangenome framework for A. flavus in agricultural environments along with associated genes for pathogen survival and aflatoxin production.¹⁴ However, research into the specific pathogenic mechanisms of A. flavus in corneal infections, particularly the exploration from the genomic level, remains relatively scarce. 

This study is designed to provide insights into the relationship between genomic characteristics and clinical features by analyzing clinical cases of A. flavus keratitis and exploring the genomic features of isolated pathogens. Additionally, we aim to identify virulent genes correlated with the prognosis of patients, thereby facilitating the development of more effective diagnostic methods and prediction of the outcomes of A. flavus keratitis. 

Materials and Methods

Establishment of A. flavus Keratitis Cohort

Twenty-one consecutive patients (21 eyes) with A. flavus keratitis were included in the study, from the Department of Ophthalmology at Beijing Tongren Hospital, Capital Medical University, between 2019 and 2023. The inclusion criteria were as follows: patients had typical clinical manifestations of A. flavus keratitis and at least one positive etiological examination (corneal scraping and microbial culture). The data encompassed demographic details, time of presentation, predisposing factors, anterior segment photography, administered treatments, and outcomes. Based on the corneal ulcer scoring system, the severity of patients with Aspergillus keratitis was graded as mild, moderate, and severe.¹⁵ The study was conducted in accordance with the Declaration of Helsinki and was approved by the Medical Ethics Committee (TRECKY2021–024). 

Fungal Strains and Growth Conditions

The conidia of 21 strains of A. flavus isolated from corneal scrapings between 2019 and 2023 were stored at −80°C in an ultra-freezer in 25% to 50% (v/v) glycerol for long-term preservation. A. flavus ATCC 204304 served as the quality control strain in this study. These strains were inoculated on potato glucose agar (PDA; Dijing, Guangzhou, China) and cultured at 28°C for at least 3 days. Subsequently, the conidia were collected and resuspended in saline with 0.1% Tween to the desired concentration. The spore titers were determined using a hemocytometer. 

Antimicrobial Susceptibility Testing

To perform the antimicrobial susceptibility testing, 100 µL conidial suspensions were diluted with 10 mL of RPMI-1640 medium (DL Biotech, Zhuhai, China) for the final inoculum density (0.4 × 10⁴ to 5 × 10⁴ CFU/mL).^16,17 Then 100 µL of the suspension was added into each well of a 96-well plate, which was combined with lyophilized antifungal drugs (Terbinafine, Natamycin, Voriconazole, Posaconazole, Itraconazole, and Amphotericin B; DL Biotech, Zhuhai, China). The plates were sealed and incubated at 35°C for 48 hours.^18–20 Resistance was defined according to the clinical breakpoints established by European Committee on Antimicrobial Susceptibility Testing (EUCAST). 

Whole Genome Sequencing and Annotation

Genomic DNA (gDNA) was extracted from each strain using the Quick-DNA fungal/bacterial miniprep kit (Zymo Research) according to the manufacturer's protocol. The concentration and purity of the gDNA samples were evaluated using Nanodrop2000 (Thermo Fisher Scientific, Waltham, MA, USA). Subsequently, the gDNA samples were submitted for whole genome sequencing (WGS) on the Illumina Novaseq platform in the PE150 mode (Novogene, Tianjin, China). The raw data obtained from the WGS were filtered to eliminate low-quality data, including reads containing low-quality bases (>40%), reads containing >10% N, and reads with an overlap between adapters exceeding 15 bp. 

De Novo Genome Assembly and Annotation

Genomes were assembled de novo by SPAdes version 3.15.5 with the reference genome (ASM1411746) as a guide. The quality of the genome assembly was assessed using QUAST version 5.0.2. Contigs that were shorter than 500 bp or possessing >95% identity and coverage overlap with other contigs were removed.²¹ Gene prediction and functional annotation were performed using the Funannotate pipeline, which included MEROPS version 12,²² dbCAN2 release 9.0,²³ BUSCO version 4.1.4.,²⁴ and InterProScan version 5.45–80.0.²⁵ 

Pan-Genome Analysis and Phylogenetic Analysis

Pan-genome analysis, comprising core genes present in all individuals, accessory genes that are not shared between all members, and unique genes found in a single strain, was performed using BPGA version 4.6.6 (using default parameters and 50% sequence identity cutoff).²⁶ The phylogenetic tree based on the whole genome was constructed using iTOL (https://itol.embl.de). Further analysis of the gene accumulation curve and core-unique sequence composition was based on these results. Additionally, clusters of orthologous groups (COGs) functional enrichment of the predicted proteins were performed.²⁷ 

Analysis of Virulence Genes

The protein sequences of 21 genomes were aligned using DIAMOND,²⁸ adhering to the following criteria: a cutoff of 40% identity, 70% coverage, and an E value of 1e-10 against the Pathogen Host Interactions Database (PHI).²⁹ Based on the specific distribution and presence of certain virulence genes, the 21 strains of A. flavus isolated from the cornea were presented using R software version 4.2.2 (The R Foundation, Vienna, Austria). STRING version 11.0 was used to analyze interactions between potential virulence gene and the predicted functional partners.³⁰ 

Experimental Validation of Virulence Genes for Corneal Pathogenicity

Phenotyping of Isolates

Each isolate was cultivated in PDA for 5 days at 28°C, and 1 µL conidia suspension (containing 10⁵ spores) of each strain was inoculated into PDA media, respectively. After incubation for 5 days at 28°C, conidia were resuspended in saline containing 0.1% Tween-20 and counted under microscopy using a hemocytometer. For stress resistance analysis, the conidia were spotted on PDA media supplemented with cell-wall perturbing agent (100 µg/mL Congo red) or osmotic stress agent (1 mmol/L NaCl). All plates were incubated in the dark at 28°C for 5 days and the diameter of each colony was recorded. 

Mice Model of A. flavus Keratitis

Thirty C57BL/6N mice aged 6 to 8 weeks, with an equal number of males and females were purchased from Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) and maintained in a pathogen-free environment for 1 week. Subsequently, they were randomly divided into three groups: the control group (receiving an injection of phosphate-buffered solution into the corneal stroma), the mutant group (corneal infection with spore suspension from the mutant strains), and the wild-type (WT) group (corneal infection with spore suspension from the WT strains). Specifically, after the mice were anesthetized with an intraperitoneal injection of 0.1% sodium pentobarbital (Sigma, USA), 1 µL conidia suspension (containing 105 spores) or sterile PBS was injected into the central corneal stroma of the left eyes using a 33-gauge syringe (Hamilton 21-2062).³¹ Only the left eye of each mouse was utilized for the experiments. The right eye of each mice model was set as absolute control and remained intact throughout the animal experiments. 

Corneal Examination and Scoring

The models with A. flavus keratitis were evaluated using a slit lamp microscope (Topcon SL-D7, Tokyo, Japan) on day 1, day 3, and day 5 (n = 4/group/time). Additionally, 1% sodium fluorescein (Shanghai Yuanye Bio-Technology Co., Ltd., China) was instilled into the inferior conjunctival sac of the mice, and the corneal staining patterns were photographed with a cobalt blue filter. The corneal opacity was assessed using a corneal scale that encompassed 3 criteria: opacity area, opacity density, and surface regularity, with each criterion being graded on a scale from 0 to 4.³² A total score of 0 was given for a normal cornea and 12 for the worst cornea. 

Tissue Collection

On day 3 after infection, the mice were euthanized (n = 6/group/time). Subsequently, the corneas were meticulously removed and half stored at −80°C in the refrigerator, with half in 4% paraformaldehyde (for paraffin sections) for further experiments. 

RNA Extraction and RT-qPCR

Total RNA was extracted from the corneas using the RNA Easy Fast Tissue/Cell Kit (Tiangen, Beijing, China). Then, the RNA was reversed transcribed (RT) to cDNA using the HiScript III All-in-One RT SuperMix (Vazyme, Nanjing, China). Real-time quantitative polymerase chain reaction (RT-qPCR) was performed using Tag Pro Universal SYBR qPCR Master Mix (Vazyme, Nanjing, China) on an ABI 7500 system (Applied Biosystems, USA). The qPCR procedure consisted of an initial step at 95°C for 30 seconds, followed by 40 cycles of denaturation at 95°C for 10 seconds, and annealing and extension at 60°C for 30 seconds. Three corneas in each group were used for real-time PCR and the relative expression of the mRNA was calculated using the 2^−ΔΔCt method. Mice β-actin expression acted as an endogenous reference to normalize relative gene expression. Primer sets used in qRT-PCR are listed in Supplementary Table S1. 

Hematoxylin and Eosin Staining

After deparaffinization and hydration, the corneal paraffin sections were incubated with hematoxylin for 5 minutes and 0.5% acid alcohol differentiation solution for 5 seconds, then washed with tap water for 3 minutes. Thereafter, they were stained with eosin for 30 seconds. 

Immunohistochemistry Staining

For immunohistochemistry (IHC) staining, corneal paraffin sections were deparaffinized twice with xylene for 10 minutes each and hydrated with gradient alcohol. Then, the heat-mediated antigen retrieval was conducted. Endogenous peroxidase was blocked by 3% hydrogen peroxide for 10 minutes. After being permeabilized with 0.1% Triton X-100 for 20 minutes and blocked with goat serum for 1 hour at room temperature, the paraffin sections were incubated with anti-TNF alpha (1:250; ab1793, Abcam), anti-IFN-γ (1:200, bs-0480R, Bioss), anti-MMP13 (1:200; 18165-1-AP, proteintech), anti-MMP3 (1:50; ab52915, Abcam), anti-MPO antibodies (1:200; 22225-1-AP, proteintech), CD68 antibody (1:250; sc-70761, Santa Cruz Biotechnology), or CD3 antibody (1:150; ab16669, Abcam) overnight at 4°C. Subsequently, the sections were washed and incubated with goat anti-rabbit IgG H&L (1:1000, ab205718; Abcam) or goat anti-mouse IgG H&L (HRP; 1:1000, ab205719; Abcam) for 1 hour at room temperature. Finally, they were stained with DAB (ZSGB-BIO, Beijing, China) for 5 to 10 minutes and observed under a microscope with 400-fold magnification. The cells with brown-stained cytoplasm were considered positive. 

Statistical Analysis

Statistical analyses were conducted using Prism software (version 9.0; GraphPad Software Inc., San Diego, CA, USA). Continuous data between the two groups were analyzed by t-test or Mann-Whitney U test, whereas among the three groups, the 1-way ANOVA or Kruskal-Wallis test was used. Categorical variables were analyzed by Chi-square test or Fisher exact test. All the hypotheses tested were 2-tailed, and P < 0.05 was considered statistically significant. 

Results

Clinical Manifestations

A total of 21 patients presenting with A. flavus keratitis during 2019 to 2023 were included. The median age was 50.4 ± 13.5 years (range from 26 to 65 years) and 61.9% (13 of 21) of the patients were men. All of them had unilateral infections. The risk factors associated with A. flavus keratitis were shown in Table 1. Among these cases, 9 of the 21 patients (42.9%) had a history of trauma, 1 patient had a history of long-term dry eye and diabetes, and 1 had a history of corneal surgery. The average duration from the onset of symptoms to arrival at our hospital was 21.2 ± 12.9 days, ranging from 5 to 45 days. At the initial visit, 8 patients presented with moderate (8 of 21) symptoms and 13 had severe symptoms based on the severity scale. After medical treatment, 4 (19.0%) patients showed improvement in clinical manifestations, whereas 17 of 21 (81.0%) required keratoplasty. 

Table 1.

 

Clinical and Genomic Features of 21 Patients With Aspergillus Keratitis

 

The minimum inhibitory concentrations (MIC) results (Table 2) demonstrated that the most effective drugs for A. flavus isolates were Posaconazole (0.06–0.125 µg/mL, MIC₅₀ = MIC₉₀ = 0.06 µg/mL), Itraconazole (0.06–0.125 µg/mL, MIC₅₀ = MIC₉₀ = 0.125 µg/mL), and Terbinafine (0.06–0.125 µg/mL, MIC₅₀ = MIC₉₀ = 0.125 µg/mL), followed by Voriconazole (0.125–0.5 µg/mL, MIC₅₀ = MIC₉₀ = 0.25 µg/mL), and then Amphotericin B (0.5–1 µg/mL, MIC₅₀ = MIC₉₀ = 0.5 µg/mL). For the same isolates, Natamycin (32–128 µg/mL, MIC₅₀ = 32 µg/mL and MIC₉₀ = 64 µg/mL) was found to be the least effective. The MIC of strains isolated from different outcomes was compared. However, there was no significant difference between the MIC and prognosis. 

Table 2.

 

Antifungal MIC Values of A. flavus Strains Isolated From Corneal Ulcers

 

Whole-Genome Analysis Showed Functional Differences of Two Groups

General Genomic Features and Whole-Genome Phylogenetic Analysis

The high-quality data of 21 isolates of A. flavus were sequenced and assembled. FastQC and Trimmomatic were used to control the raw data for an average quality, and reads of low quality were excluded. The main obtained genomic characteristics and annotation information are presented in Table 1. The average genome size was 37.9 Mb (range from 36.8 to 42.4 Mb), with 47.9% of the GC content on average. The mean gene number was 12709, ranging from 11167 to 16101. To further investigate the inter-relationship among the strains, a phylogenetic tree was constructed based on the whole genome. In addition, the clinical prognoses of patients infected by these 21 A. flavus strains were incorporated around the circular whole-genome phylogenetic tree (Fig. 1). Upon meticulous analysis of the phylogenetic tree, it was revealed that these strains, despite eliciting disparate prognoses in patients, did not exhibit significant differences at the whole-genome level. 

Figure 1.

 

Whole-genome phylogenetic analysis of 21 A. flavus strains. The colored ring shows the clinical prognoses of patients infected by these strains.

Pan-genomic and Functional Characterization of A. flavus Strains

Given the lack of research on the pan-genomic characteristics of A. flavus, the core and pan-genome were analyzed for the purpose of comparing the general genetic similarities and diversity among all the strains. Empirical power-law regression and exponential curve fitting were used for extrapolation of the pan- and core genome curves, respectively, as presented in Figure 2A. The number of pan-genomes did not substantially increase after the addition of approximately 20 genomes, indicating that the pan-genomes were in a “close” state. The A. flavus pan-genome was composed of 17,326 gene clusters with a core genome of 5378 (31.0% of the pan-genome) ortho groups in all 21 isolates. Furthermore, the detailed distribution of unique genes in each genome was exhibited (Fig. 2B), and the results showed that strain Asp1784 contained the largest number of unique genes (n = 2428), whereas strain Asp2799 had the smallest number of unique genes (n = 10). Among the COG of Proteins database categories, core genes were found to be significantly abundant in the general function prediction (21.5%). However, the unique genes were more enriched in cell wall/membrane/envelope biogenesis, cell motility, and transcription. 

Figure 2.

 

Pan-genome analysis of 21 A. flavus strains. (A) Genome size evolution curves of the pan-genome (blue) and core genome (orange) grouped. (B) The number of core genes among all strains and unique genes of individual strain. (C) The percentage (%) of COG categories and annotations of predicted genes within the core, accessory, and unique genomes of all strains, strains with poor and good prognosis.

Based on the clinical prognosis, we grouped strains from different backgrounds for COG functional differences analysis (Fig. 2C). In terms of core genes, the COG functional analysis of the two groups showed no significant difference. However, 7 categories (D, M, Q, P, R, and S) of the accessory genome, and 13 categories (M, N, O, T, K, C, G, H, I, Q, P, R, and S) of the unique genome were observed with significant differences. Notably, categories M (cell wall/membrane/envelope biogenesis) and categories S (function unknown) of accessory and unique genomes were highly represented in strains with poor clinical prognosis, which may potentially play a pivotal role in the progression and prognosis of A. flavus. 

Virulence-Related Genes in A. flavus Genomes

One hundred eighty-three virulence-associated genes were identified using the PHI database. This list includes genes involved in cell wall integrity (19 genes), initiation of germ tubes (2 genes), nutrient uptake in invasive growth (11 genes), resistance to host immune response (4 genes), response to stress (15 genes), toxins (9 genes), etc. Overall, the majority of virulence-associated factors were well conserved. 145 of 183 (79.2%) of the genes were present in all genomes. The remaining 38 virulence-associated genes had some degree of genetic variation expected to affect gene function among our 21 genomes, which can be visualized in Figure 3A. Among them, 31 genes were mutated in 4.8% of the isolates (1/21); 4 genes, including Fcr3, kexB, HorA, and tmpL, were mutated in 9.5% of the genomes (2/21). Furthermore, we observed absence of hexA gene, encoding aflatoxin biosynthesis in 6 (28.6%) isolates and the ZafA gene, regulating zinc homeostasis in 9 (42.9%) strains. Compared with the proportion of gene mutant in the clinical good prognosis (2/4, 50%) and with poor prognosis (0/17), we found there was a significant difference in the kexB gene, a gene coding for kexin-like proteinases and affecting cell wall integrity between the two groups (P = 0.002; Fig. 3B). We then analyzed the interactions between the kexB gene and the predicted functional partners using Search Tool for the Retrieval of Interacting Genes (STRING; see Fig. 3C). Functional enrichments showed that the gene functions in kexB network were mainly enriched in proteolysis, metallopeptidase activity, peptidase activity, etc (Fig. 3D). 

Figure 3.

 

Virulence-related genes in A. flavus genomes. (A) Comparative analysis of virulence factor profiles across 21 A. flavus genomes. (B) The proportion of virulence gene mutations in groups with good and poor prognosis. (C) Interactions between kexB gene and the predicted functional partners. (D) GO functional enrichments in the kexB network. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant.

Growth Phenotypes of the ΔkexB Strains

Compared with the ATCC strain, the mean mRNA expression levels of kexB gene decreased significantly in ΔkexB strains, while there was no obvious change in the WT strains, which was 0.67 folds (P = 0.003) and 1.17 folds (P = 0.340), respectively (Fig. 4B). In order to identify the microbiological features between the ΔkexB strains and the WT strains, we inoculated them on PDA agar medium for 5 days (Fig. 4C). The diameter of the colony was measured, which showed that there was little difference of colony morphology between the two groups when they grew on PDA media. To understand the strain’s response to environmental stress, we tested fungal sensitivity to hyperosmotic stress factors and cell wall damaging agents. The number of spores, colony diameter, and growth inhibitory rate of the strains in the presence of the osmotic stress agents NaCl and cell-wall perturbing agents Congo red were analyzed. Conidia counting showed that the ΔkexB strains exhibited a remarkably reduced conidiation at 28°C when compared to the WT strains (Fig. 4D). The mean number of spores in the mutant and WT strains were 475.0/cm² and 105.8/cm² (P < 0.0001), 570.0/cm² and 133.3/cm² (P < 0.0001), 134.8/cm² and 49.8/cm² (P = 0.0004) in PDA, and NaCl and Congo Red media, respectively. Interestingly, the ΔkexB strains also showed higher sensitivities to hyper osmotic stress (1 mol/L NaCl) and cell wall damaging agents (Fig. 4E). Compared to growth in PDA medium alone at 28°C, the growth of ΔkexB strains in the presence of 1 mol/L NaCl decreased by 4.6%, whereas the growth of strains from WT type decreased by 3.2% (P = 0.343). Similarly, cell wall stress imposed by the addition of 100 µg/mL Congo Red to the medium inhibited the growth of both strains. At 28°C, compared to growth in PDA medium alone, the growth rates of the ΔkexB strains and WT strains were significantly lower (by 20.1% and 5.7%, respectively, P < 0.0001). 

Figure 4.

 

Growth phenotypes and sensitivity to cell wall perturbing compounds of the ΔkexB. (A) Representative anterior segment photographs of patients with A. flavus keratitis. (B) Decreased mRNA expression of the kexB gene in the ΔkexB mutant strains. (C) The isolated ΔkexB and WT strains grew on PDA media, with 1 mol/L NaCl and 100 µg/mL Congo Red at 5 days, respectively. (D) Conidia counting of the ΔkexB and WT at 28°C. (E) The growth inhibitory rate between two groups. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant.

Calcofluor White Staining of Wild-Type and ΔkexB Mutant Strain

Fluorescence microscopic examination of the WT and ΔkexB mutant strains following calcofluor white staining demonstrated that the fluorescence intensity of the ΔkexB mutant strain was distinctly lower than that of the WT strain (Fig. 5). The hyphae and spores of the WT strain presented a vivid bright blue fluorescence, whereas those of the ΔkexB mutant strain only showed a pale light blue. 

Figure 5.

 

Fluorescent microphotographs of calcofluor white stained hyphae and conidia. (A, B) WT strains; and (C, D) ΔkexB mutant strains.

Virulence of the ΔkexB Mutant Strains on Mice Cornea Infection

To validate the potential virulence effect of ΔkexB strains on corneal infection, intrastromal injections of 5000 conidia from the ΔkexB strains or WT strains were performed in the corneal stroma of C57BL/6 mice. Photographs of corneal ulcers with and without 1% sodium fluorescein were conducted on day 1, day 3, and day 5 after infection (Fig. 6A). On the first day, the mice inoculated with the mutant strain showed less severe corneal ulceration and a significantly lower clinical score (2.3 ± 1.0) compared to those with the WT strains (5.0 ± 0.8,P = 0.005; Fig. 6B). Regarding the outcome of A. flavus keratitis, the clinical scores of mice inoculated with the mutant strains were lower than those with the WT strains, which were 4.5 ± 2.6 and 8.7 ± 1.0, respectively (P = 0.023). The analysis of the corneal survival rate showed that, after 3 days, 80.0% (8/10) of the corneas infected with WT strains showed descemetocele or corneal perforation, which was significantly higher than those inoculated with the mutant strains (4/10, 40.0%). In addition, the log rank (Mantel-Cox) test showed that the survival curves were significantly different (P < 0.0001, Fig. 6C). 

Figure 6.

 

The A. flavus keratitis model mice infected by ΔkexB and WT strains. (A) The representative anterior segment photographs with and without fluorescein staining under slit lamp microscopy on days 1, 3, and 5 (n = 4). (B) The clinical scores and (C) corneal survival rate was assessed in each group, and four mice were assessed in each group for the indicated time (n = 4). (D) Decreased gene expression of IFN-γ, TNF-α, MMP3, MMP9, and MMP13 in the mutant group. (E) H&E staining revealed a reduction in inflammatory cells and IHC showed decreased protein levels of IFN-γ, TNF-α, MMP3, and MMP13 in cornea tissues of the ΔkexB mutant group. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant.

Compared with the WT group, hematoxylin and eosin (H&E) staining showed decreased infiltrated inflammatory cells in the mutant group. Concurrently, the mutant group exhibited significantly lower mRNA expression levels of IFN-γ, TNF-α, and MMP3, which was consistent with the outcomes of histological examination and clinical manifestations (all P < 0.05; Figs. 6D, 6E). To further investigate the infiltration of immune cells, tissue sections were immunostained with MPO, CD68, and CD3 antibodies to examine neutrophils, macrophages, and T cells, respectively. Our results revealed that the WT strains of A. flavus induced a significantly greater infiltration of neutrophils, macrophages, and T cells compared with the ΔkexB strains (all P < 0.05; Fig. 7). These data indicated that the ΔkexB strains alleviated the inflammation and stromatolysis of the infected cornea, highlighting their potential role in modulating the host immune response and corneal tissue damage. 

Figure 7.

 

Infiltration of immune cells in mice cornea. Representative sections and quantitative analysis of the WT and ΔkexB mutant strains-infected corneas immune-stained with antibodies to neutrophils (MPO), macrophages (CD68), and T cells (CD3). ***P < 0.001, **P < 0.01, *P < 0.05; ns, not significant. Scale bar = 50 µm.

Discussion

In this study, we investigated 21 cases of A. flavus keratitis and found that the clinical presentations and outcomes were notably variable, although there were no significant differences in MIC among A. flavus strains. Subsequently, pan-genome analysis revealed variations in the proportions of certain gene families among different strains. Further virulence gene analysis identified the kexB gene, which encoded a protein involved in cell wall glycan synthesis and was associated with the severity of keratitis. Mutations in this gene led to reduced conidia formation and higher sensitivity to cell wall damaging agents. Moreover, the mice models of A. flavus keratitis were established to validate the potential virulence effect of the ΔkexB strains through clinical scoring and immunohistochemistry. The results showed that ΔkexB strains caused significantly less inflammation and stromal melting compared with the WT strains. Our research uniquely linked A. flavus genomic features to clinical outcomes, providing novel insights into the pathogenicity of this fungus. 

In this cohort, A. flavus keratitis exhibited poorer outcomes compared with other forms of fungal or infectious keratitis, which is consistent with previous cohorts of A. flavus keratitis.³ Louisa et al. investigated the relationship between in vitro susceptibility and clinical outcomes in fungal keratitis, which revealed that a higher voriconazole MIC was correlated with an increased need for therapeutic penetrating keratoplasty among Fusarium-positive cases, but no significant correlations were observed among Aspergillus cases.³³ In this study, no relationships between antifungal susceptibility and treatment outcomes were found in A. flavus, consistent with the findings of Venkatesh et al.³⁴ According to clinical breakpoints, the voriconazole MIC for all A. flavus isolates in our study were at or below the threshold of 1 µg/mL.³⁵ The same trend can be observed for Posaconazole, Itraconazole, and Amphotericin B, with clinical breakpoints of 0.5 µg/mL, 1 µg/mL, and 4 µg/mL, respectively.^19,36 These findings suggested that in vitro antifungal susceptibility testing may be not sufficient to predict treatment outcomes for A. flavus keratitis, underlining the need for further research to better understand their complex relationship. 

Despite leading to diverse prognoses among patients, the strains did not exhibit prominent differences on the phylogenetic tree. The possible explanations could be as follows: although the whole genome contains abundant genetic information, the key gene regions that determine the pathogenicity and prognostic differences of strains during the infection process may only account for a small part of it. Meanwhile, their signals in the whole genome may be masked by the information of other non-key genes. Moreover, even though the variations in the whole genome may affect the biological behaviors of strains in a rather complex way, these differences are difficult to be directly reflected in the phylogenetic tree. The pan-genome analysis focused on the collective set of genes across all strains and revealed that the pan-genome of A. flavus was in a “closed” state.³⁷ This implies that the number of pan-genes did not substantially increase, indicating that a sufficient sample size was utilized to capture the overall genomic diversity of the A. flavus species. The functional distribution among the COG of A. flavus isolated from different clinical contexts suggested that certain categories may be highly associated with its pathogenicity.³⁸ Notably, within the core and accessory genomes of A. flavus, the gene cluster associated with the cell wall, membrane, and envelope biogenesis (category M) showed significant differences between strains isolated from patients with favorable and unfavorable outcomes. 

Virulence genes involved in stress resistance, high growth rate, and nutritional versatility render A. flavus the primary pathogenic mold in fungal keratitis. Among these genes, the kexB gene, which encodes a kexin-like protein, has been proven to have crucial effects on cell wall integrity and virulence in multiple species including Aspergillus oryzae, Candida glabrata, and Candida albicans.^39,40 In this study, the kexB gene was identified in A. flavus, with a higher proportion of mutants observed in the group with good prognosis. Further experiments revealed that the ΔkexB resulted in the lower mRNA expression of kexB, and the mutant strains showed a reduced conidia formation as well as higher sensitivity to cell wall perturbing agents, which was in accordance with previous studies. Wang et al. demonstrated that the deletion of the kexB gene resulted in impaired N-glycan processing, and the activation of the MpkA-dependent cell wall integrity signaling pathway, which further resulted in a retarded growth, reduced conidia formation, temperature-sensitive defect in cell wall integrity, and attenuated virulence.³⁹ 

Calcofluor white stain of A. flavus showed that fluorescence was much more intense in the WT strain than in the ΔkexB strain. This disparity indicates that the ΔkexB mutant strain has a reduced content of chitin and cellulose in its cell wall. This is further supported by the fact that calcofluor white can specifically bind to these components, which are crucial constituents of the cell wall structure.⁴¹ Under the current experimental conditions, the kexB gene does not appear to have a direct influence on the drug sensitivity profile of A. flavus strains as all strains had MIC values within the clinical breakpoints. 

The phenotypes of A. flavus keratitis corresponded to genotype-based groupings of A. flavus strains isolated from the patients. In the group with mutant strains, A. flavus keratitis was milder and showed relatively slow progression. In contrast, the group with WT strains presented with an acute onset on day 1 and exacerbated in the following days, leading to central corneal thinning and perforation in most of the mice models. Compared to the control group, H&E staining revealed more inflammatory cell activation in the lesion area. Furthermore, the relative mRNA expression level of inflammatory cytokines and matrix metalloproteinases also reflected the exacerbation of inflammation condition. Former research has indicated that much of the corneal destruction in microbial keratitis was likely due to an excessive host inflammatory response. Activated neutrophils attempt to destroy the pathogen through the production of reactive oxygen species and the release of enzymes such as MMPs but these mechanisms can also damage surrounding host tissue.^42,43 In our study, it was observed that after the injection of A. flavus, a substantial number of neutrophils were promptly recruited and rapidly amassed at the site of infection. Meanwhile, a limited quantity of T cells was detected infiltrating the area. Although their presence was not as conspicuous as that of the neutrophils, the engagement of T cells implies the initiation of the adaptive immune response. In comparison with the ΔkexB strains, the WT strains of A. flavus elicited a remarkably enhanced infiltration of neutrophils and macrophages, as well as T cells. Thus, A. flavus infection triggers the acute infiltration of immune cells in the cornea, further causing excessive inflammatory response and MMP, which resulted in cornea opacification, corneal thinning, and even perforation. 

Through gene interaction and functional enrichment analysis, we discovered that the kexB gene network was mainly related to proteolytic activity. As secretory proteins are important virulence factors for fungi to overcome the host defense and propagate in the cornea, the downregulation of kexB gene in the mutant strain may affect the function of this network and lead to defect in conidia formation, and cell wall integrity, as well as invasion, thus resulting in a slower course of infection, a milder inflammatory response, and less tissue damage.⁴⁴ 

Our study still had several limitations. First, due to the limited sample size, this study primarily explored the role of kexB gene mutation. Nevertheless, incorporating a larger cohort could identify additional virulence factors associated with the severity of keratitis. Second, we mainly focused on the phenotypic characteristics of the strains, the mechanism regarding how the kexB gene affected the host-pathogen interaction requires further research. Finally, the bioinformatics methods were mainly used to investigate the role of the kexB gene in the regulatory network. Although the algorithms of these methods have been proven accurate and reliable, it would be more beneficial to validate in vivo at the nucleic acid levels using molecular biological techniques. 

In summary, our study correlates the genomic characteristics of A. flavus with clinical manifestations, providing a novel perspective for an in-depth understanding of the pathogenicity of A. flavus. The kexB gene holds promise as a new target for the diagnosis and treatment of A. flavus keratitis. Future studies can further explore the interactions between the kexB gene and other virulence genes, as well as the impact of different host immune responses on kexB gene mutations. 

Acknowledgments

Supported by the National Key Research and Development Program, grant number 2021YFC2301000. 

Author Contributions: J. Liu, M. Kang, J. Pang, and Y. Wei performed the experiments. J. Liu analyzed the data and wrote the manuscript. Z. Zhang and X. Xu performed the bioinformatics analysis. M. Kang revised the manuscript. Yang Zhang, Kexin Chen, Zhiqun Wang, and Xinxin Lu provided materials. Q. Liang designed the experiment, provided materials, and revised the paper. All authors read and approved the final manuscript. All authors have read and agreed to the published version of the manuscript. 

Disclosure: J. Liu, None; M. Kang, None; Y. Wei, None; Z. Zhang, None; J. Pang, None; Q. Chen, None; X. Xu, None; Z. Wei, None; Y. Zhang, None; K. Chen, None; Z. Wang, None; X. Lu, None; Q. Liang, None