Infectious bovine keratoconjunctivitis (IBK) is the most common ocular disease in cattle (Brown et al., 1998). IBK has a substantial economic impact including costs associated with treatment as well as decreased weight gain in affected animals and impacts animal welfare by causing pain and blindness (Killinger et al., 1977; Dewell et al., 2014). Moraxella bovis (M. bovis) is the only bacterium that has reproduced IBK-like lesions in a variety of experimental models (Henson and Grumbles, 1960; Aikman et al., 1985). Other bacteria are often associated with IBK, but so far none have produced disease experimentally. The most notable and frequently isolated of these associated bacteria is Moraxella bovoculi (Angelos et al., 2007b). While M. bovoculi has been unsuccessful at inducing IBK experimentally (Gould et al., 2013), it is more frequently isolated from IBK lesions when compared to M. bovis, using both aerobic culture and molecular detection techniques (Loy and Brodersen, 2014; Zheng et al., 2019). Despite the lack of proven causation for M. bovoculi in IBK to date, in 2017 the USDA approved the first conditionally licensed M. bovoculi based vaccine product to be marketed for the prevention of IBK (USDA CVM code: 2A77.00, Addison Biological Laboratory). Recently, whole genome sequencing of M. bovoculi has revealed a large degree of diversity within the species that led to the characterization of two distinct genotypes (genotype 1 and genotype 2) separated by over 23,000 single nucleotide polymorphisms (Dickey et al., 2018). To date, only genotype 1 M. bovoculi have been isolated from IBK lesions while both genotypes have been recovered from animals without clinical signs.

Strains of M. bovis have been shown to produce an exotoxin belonging to the repeats-in-toxin (RTX) class of exotoxins that is cytopathic to bovine erythrocytes and neutrophils (Clinkenbeard and Thiessen, 1991; Angelos et al., 2001). In M. bovis, this RTX toxin, encoded by mbxA within an operon, is often referred to by several names including cytolysin, hemolysin, or cytotoxin. RTX toxins are known virulence factors in a variety of veterinary pathogens, including species within the family Pasteurellacea (Linhartova et al., 2018). A well-studied example of this is the leukotoxin produced by Mannheimia haemolytica (Frey, 2019). These RTX toxins are secreted in a calcium-dependent manner via a type I secretion system (T1SS) responsible for translocating the toxin from the cytosol to the exterior (Linhartova et al., 2018) Like M. bovis, some M. bovoculi strains also contain a complete RTX operon that produces an RTX toxin, cytotoxin A encoded by mbvA (Angelos et al., 2007a). Like mbxA produced by M. bovis, mbvA of M. bovoculi is responsible for hemolytic and lytic activity on bovine cells (Cerny et al., 2006; Angelos et al., 2007b). Within M. bovoculi, only the disease associated genotype 1 strains have been shown to possess the RTX operon, although not all do (Dickey et al., 2018). Besides RTX toxins, both M. bovis and M. bovoculi express a type IV pilus, a known virulence factor in other bacterial pathogens (Angelos et al., 2021). These similar potential virulence factors between the two Moraxella species highlight the relevance of continued investigation into the diversity within M. bovoculi and how this diversity may impact IBK pathogenesis.

Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is an approach that is increasingly being applied to the identification of prokaryotes and eukaryotes. (Seng et al., 2009; Clark et al., 2013; Khot and Fisher, 2013; Karger et al., 2019). MALDI-TOF MS has replaced or supplemented more traditional biochemical identification methods as it is faster and increasingly more accurate as databases become mature. Using spectrum profiles generated from MALDI-TOF MS, biomarker models have been developed that are capable of differentiating subspecies and genotypes within a given bacterial species (Loy and Clawson, 2017; Mani et al., 2017; Perez-Sancho et al., 2018). Recently, a MALDI-TOF biomarker model was developed that accurately distinguishes M. bovoculi genotypes 1 and 2 (Hille et al., 2020). This model allows for the screening of strains for potential disease associated genotypes without the need for genome sequencing.

Genotype 1 M. bovoculi strains isolated from IBK associated eyes have been found with a higher frequency of the RTX operon than those isolated from unaffected eyes (Dickey et al., 2018). While it can be suspected based on a hemolytic phenotype, confirming the presence or absence of an RTX operon requires PCR or genomic sequencing (Angelos et al., 2003). In this study we evaluated the utility of MALDI-TOF MS biomarker models to accurately classify M. bovoculi strains regarding the presence or absence of an RTX operon among strains that had previously undergone whole genome sequencing and thus whose RTX operon status was known. Such a model would allow for an additional screening tool to characterize strains more likely to represent disease associated strains.

In addition to traditional tryptic soy agar (TSA) with 5% sheep blood agar culture conditions, we also compared biomarker model accuracies using three additional growth agar formulations to determine if agar formulation could improve model accuracy. The hemolytic activity of hemolysin produced by M. bovis has been shown to decrease when extracellular calcium is rendered unavailable (Clinkenbeard and Thiessen, 1991; Billson et al., 2000). Calcium has also been shown to promote efficient post translational modification and excretion of RTX toxins from the cell via the T1SS in Bordetella pertussis (Bumba et al., 2016). We hypothesized calcium may be a limiting factor in the production of RTX for M. bovoculi using the traditional culture conditions. Additionally, in the authors’ experience, Moraxella sp. isolated from cattle grow subjectively better on agar that utilizes bovine red blood cells vs traditional sheep’s blood agar.

Here we have described a group of biomarker models developed using ClinProTools 3.0 software that are capable of correctly phenotyping most M. bovoculi genotype 1 strains for RTX. Across all models, the RTX phenotype specificity ranged from 26.7% - 86.7% while the sensitivity ranged from 46.7% - 93.3%. The most accurate model overall was the SVM model developed using strains grown on TSA + 5% bovine blood supplemented with 10 mmol/L CaCl₂, which had a specificity and sensitivity of 73.3% and 93.3% respectively. Regardless of the biomarker model algorithm used, the usage of agar incorporating bovine blood and CaCl₂ substantially outperformed sheep blood agar. Since secretion of RTX toxins utilizes a calcium-dependent T1SS, the extra calcium may result in an overall increase in RTX production, although confirming this would require more investigation and was outside the scope of this study. Additionally, mammalian erythrocytes vary in the composition of their membranes (de and Van Deenen, 1961), and the species of erythrocyte has shown to affect activity of other toxins (Bhakdi et al., 1984). This increase in hemolysis may increase nutrient or other factor availability that enhances model performance.

MALDI-TOF MS biomarker models often highlight discriminatory peaks between groups of spectra that are different enough in their m/z value that they can be discerned even without the use of computer models. For instance, within M. bovoculi a strong peak at 9057 m/z is specific for genotype 1 whereas a peak at 6550 m/z is specific for genotype 2 (Hille et al., 2020). Having strong discriminatory peaks such as these allows those without ClinProTools 3.0 software to differentiate groups by manually examining spectra. This was not the case for the current study and highlights the need for biomarker models and their algorithmic approach to analysis when spectra peak differences may be subtle between groups.

The SVM model developed in this study using bovine blood agar and CaCl₂ provides an efficient method of RTX phenotyping for M. bovoculi without the need for PCR or genomic sequencing. A negative predictive value of 91.7% means that RTX – strains can be classified accordingly with acceptable accuracy. While the importance of RTX toxins in the pathogenesis of IBK is not fully known, they are regarded as likely important virulence factors and RTX + strains are overrepresented in cases of IBK (Angelos et al., 2007a; Dickey et al., 2018). The ability to classify RTX – strains in this manner may prove beneficial in the formulation of autogenous vaccines for IBK as this will allow vaccine manufacturers to eliminate any RTX – strains from consideration and include only strains that are more likely to represent disease associated, RTX + strains within the vaccine formulation. With hemolysis shown to be RTX mediated in M. bovis, an RTX phenotyping MALDI-TOF MS biomarker model for M. bovis analogous to the one described here for M. bovoculi, would provide the same utility for vaccine formulations that choose to include M. bovis. Developing an M. bovis model would benefit from a library of sequenced M. bovis isolates whose RTX status is defined, unless hemolytic activity alone was used to assume RTX status. Here, we focused solely on M. bovoculi given our in-house library of previously sequenced isolates in order to examine the biomarker phenotyping proof-of-concept since the known presence of RTX components allowed us to avoid using hemolysis alone to classify RTX presence. In addition to its relevance to IBK, this study also serves as a blueprint of methods and proof-of-concept for utilizing MALDI-TOF MS spectra and biomarker models to distinguish strains of bacteria based on their ability to produce secreted exotoxins. Such methods could prove beneficial in differentiating other bacterial pathogens of both humans and animals that possess secreted exotoxins as virulence factors. Beyond being a proof-of-concept, the methods described here also reduce the time required to classify a M. bovoculi isolate by RTX status based solely on hemolysis. The gross appearance of hemolysis is often not readily apparent until 48 hours of growth while the MALDI-TOF MS biomarkers models we developed here use colonies that were grown for only 24 hours.

A major limitation for this study is the inability to assure only a single variable, in this case RTX presence or absence, differentiates the groups of isolates. It is possible that some of the peaks incorporated into the models developed here represent isolate components unrelated to RTX itself. We have mitigated the likelihood of this limitation affecting the overall study interpretation by: 1) utilizing only genotype 1 strains for the entire study, and 2) incorporating a large number of isolates and spectra within each group, and 3) utilizing geographically diverse populations of isolates for both RTX status groups.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

MH performed MALDI-TOF analysis, developed the biomarker models, wrote the original manuscript draft and edited the manuscript. MC and AD performed genomic sequencing analysis on the bacterial strains and edited the manuscript. JHL performed MALDI-TOF analysis. JDL conceptualized the study, supervised the project, and edited the manuscript. All authors contributed to the article and approved the submitted version.

This project was funded by the Nebraska Experiment Station with funds from the Animal Health and Disease Research (section 1433) capacity funding program (accession 1017646) through the USDA National Institute of Food and Agriculture. The use of product and company names is necessary to accurately report the methods and results; however, the United States Department of Agriculture (USDA) neither guarantees nor warrants the standard of the products, and the use of names by the USDA implies no approval of the product to the exclusion of others that may also be suitable. The USDA is an equal opportunity provider and employer.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.