A novel PCR method targeting staphostatin genes differentiates Staphylococcus aureus from Staphylococcus epidermidis in clinical isolates and nasal microbiome samples

Staphylococcus aureus is a pathogen associated with community-acquired and healthcare-associated infections, and is known for rapid acquisition of resistance to multiple antibiotics. Staphylococcus epidermidis is another staphylococcal species commonly seen as a commensal, but may cause opportunistic infections. These two closely related species may naturally co-colonize and inhabit the human skin and nose; thus it is important to develop methods to differentiate between them. Conventional differentiation methods (e.g., mannitol salt agar, coagulase test, Analytical Profile Index kits) require culturing and isolating the bacteria, while molecular methods based on polymerase chain reaction (PCR) or sequencing are culture independent, highly accurate, and sensitive. They may be objectively interpreted and can be adapted for high-throughput assays. Here, we aimed to develop a rapid and accurate method to differentiate mixtures of S. aureus and S. epidermidis in clinical isolates and microbiome samples. Using comparative genomics tools (e.g., MetaRef and PATRIC bioinformatics resource center), we compared the core and pan genomes of S. aureus and S. epidermidis to determine signature genes for each species. Among candidate genes, we chose the virulence genes, staphostatin B and staphostatin A as genetic biomarkers for S. aureus and S. epidermidis , respectively. We designed PCR primers that target specific sequences within each gene and tested them on DNA isolated from laboratory strains, clinical isolates, nasal isolates, and nasal swabs. Our developed method is simple, specific, sensitive, and can be applied to metagenomic samples as well as competition or co-evolution studies in which the two organisms are co-cultured.


Introduction
Staphylococcus aureus is a major human nasal colonizer inhabiting about one 3 third of the world's population (Yan et al. 2013). It is the causative agent of several healthcare-associated infections such as pneumonia, meningitis, endocarditis, and osteomyelitis. Healthcare-associated infections represent a global life-threatening problem, especially in developing countries. They lead to financial losses by extending the duration of hospitalization (5-29.5 days in developing countries (Allegranzi et al. 2011)), by diminishing the productivity of human resources, and by expanding the burden of treatment cost on governments. More importantly, healthcare-associated infections often lead to the emergence of multiresistant pathogens, popularly known as superbugs. The World Health Organization (WHO) reported that ten out of every 100 surgical site infections (the second dominating type among hospitalized patients) are caused by S. aureus (Allegranzi et al. 2011).
Although primary infections by S. aureus are not deadly in nature, this pathogen causes several complications, many of which are invasive, and is one of the main causes of sepsis, which leads to death if not treated properly. Thus, the emergence of multiresistant strains makes this pathogen one of the most dangerous bacteria. In fact, S. aureus is one of the so-called ESKAPE pathogens, which are declared as a global public health threat (Pendleton et al. 2013). Patients infected with methicillin-resistant S. aureus (MRSA) are 64% more likely to die than those infected with non-resistant S. aureus (World Health Organization 2018).
Staphylococcus epidermidis is another staphylococcal species that predominantly colonizes the axillae, head, and nares of humans. It is a major skin resident and constitutes about 90% of aerobic resident microbiota (Cogen et al. 2008;Sanford and Gallo 2013), but may cause opportunistic infections, especially to patients using indwelling medical devices.

1.
aureus and S. epidermidis share morphological characteristics and genotypic traits, and are practically indistinguishable on microscopic examination. The differentiation between these two closely related staphylococcal species mainly depends on culturebased methods. For example, S. aureus is able to ferment mannitol producing an acidic byproduct, which lowers the pH of the mannitol salt agar (MSA) medium and turns its color into yellow, while S. epidermidis cannot ferment mannitol.
S. epidermidis belongs to the group of coagulase-negative staphylococci (CoNS), which are distinguished from S. aureus by the absence of the coagulase enzyme. For years, selective culture media, e.g., MSA and Baird Parker agar, and biochemical tests, e.g., the coagulase test and many tests in Analytical Profile Index (API) kits have been playing an important role in their differentiation.
Although these methods give accurate results in most instances, they may fail to differentiate between the two species when present in mixed cultures or clinical specimens. For example, S. aureus strains isolated from nasal swabs were reported to produce false-negative results on MSA (dos Santos et al. 2015). The major confirmatory test for S. aureus, the coagulase test, may also provide negative results (Kateete et al. 2010). Moreover, all these tests require fresh overnight cultures, and are thus time consuming.
Consequently, there is a continuous need for cost-effective, more rapid and reliable molecular tools for identification and differentiation of these two common staphylococcal species, especially when identification is performed directly without culture. Accordingly, a few culture-independent methods were developed (Chiang et al. 2012;Ghebremedhin et al. 2008;Hirotaki et al. 2011), but none of them was used on human microbiome samples. Here, we aim to develop a novel biomarker to allow the rapid 5 molecular differentiation between S. aureus and S. epidermidis without the need for isolation or culture, notably in nasal swabs intended for microbiome analysis.
For this purpose, we used comparative genomics to find suitable markers, staphostatin A in S. epidermidis and staphostatin B in S. aureus, and developed a PCR method, which proved to be sensitive and specific in discriminating both bacterial species either from pure isolates or in mixtures.

Ethics statement
All experiments were conducted according to the national and international ethical standards. The experimental protocols, including sample collection from nasal swabs and consent forms, were revised and approved by the Faculty of Pharmacy Cairo University Safety and Health Ethics Committee (Protocol approval #MI1137 on 23/6/2014).

Bioinformatics and comparative genomics analysis
We compared the core genomes of S. aureus and S. epidermidis using MetaRef (Huang et al. 2014), a comparative genomics tool, and we used PATRIC Proteome Comparison Service (Antonopoulos et al. 2017;Wattam et al. 2014)

Collection of nasal samples
We collected duplicate nasal swabs from 12 healthy volunteers living in Cairo, Egypt, including ten nurses working at a regional hospital in Qasr El-Ainy District, Cairo, Egypt. Their ages ranged from 25 to 42 years, and they included ten females and two males.
Volunteers were asked to gently rotate sterile cotton swabs wetted with 0.9% sodium chloride solution four times in both anterior nares. All volunteers were informed of the study goals, benefits, and potential risks and provided informed consent. Bacterial isolates obtained from the nasal samples were stored at -70°C in Brain Heart Infusion broth supplemented with 20% glycerol.

DNA extraction, precipitation, and purification
From colonies: Three to five bacterial colonies, from an overnight-cultured plate, were suspended in 100 ml of deionized sterile water or in TRIS-EDTA (TE) buffer, boiled for 2 min, and finally centrifuged for 1 min. The supernatant served as a template for PCR amplifications and was aliquoted and stored at -20˚C when not in use.
From nasal swabs: DNA was extracted by previously published chemical methods (Sagar et al. 2014;Singh et al. 2013), with the following modifications: To enhance the cell lysis of Gram-positive bacteria, we used beads from the PowerSoil® DNA isolation kit 8 (MoBio laboratories Inc., Carlsbad, USA) in Eppendorff tubes containing 1 ml lysis buffer, and the beads were beaten in a MO BIO Vortex Genie® 2 Vortex (MoBio laboratories Inc., Carlsbad, USA) at 10,000 revolutions per minute for 30 seconds.
For DNA sequencing, bands from PCR products on agarose gels were excised with sterile scalpels and purified by the QIAGEN Gel Extraction kit (Qiagen, Hilden, Germany).

Primer design
The NCBI protein accession numbers of staphostatin A for S. epidermidis and staphostatin B for S. aureus were OBZ50474 and CEH26156.1, respectively. From the protein database record, we obtained the corresponding coding DNA sequence of their respective genes (sspC and ecpB), and confirmed the gene length by comparison to other open-reading frames using BlastX (Altschul et al. 1990).
Before designing the primers, we aligned the nucleotide sequences of staphostatin A to all those present in NCBI NR database to detect any variations in the nucleotide sequence of all S. epidermidis strains. We repeated the previous steps with staphostatin B in S. aureus species. Finally, we designed primers ( Table 1) for each gene using PrimerQuest © (Integrated DNA Technologies (IDT), Inc., San Diego, CA, USA) and we checked the primers' properties using OligoAnalyzer ® (IDT, San Diego, CA, USA).

Results
Limitation of microbiome analysis methods as shown by in silico community simulation The motivation of this study is the need to differentiate S. aureus from S. epidermidis in mixtures, especially in human nasal microbiome samples in which the two organisms may co-occur. This need was triggered by the reported inability of classifying most organisms to the species level, in general, and staphylococci, in particular, in nasal microbiome samples (Frank et al. 2010).
To confirm this reported observation, we aligned the full sequences of all copies of the16S rRNA gene from two representative S. aureus and two representative S. epidermidis genomes (Fig. S1). In addition, we used the 16S rRNA sequences of a third species, Staphylococcus saprophyticus, as a contro for variability in the 16S rRNA sequence within genus Staphylococcus.
The alignment demonstrated a high similarity between the 16S rRNA sequences of the three species. The few nucleotide differences were scattered around the full length of the sequence; however, there were long stretches of 100% identical sequences, for example: the first 70 nucleotides, a long stretch between nucleotides 661 and 1005, and another stretch from 1161-1265 (Fig. S1).
As common 16S amplicon-based methods for microbiome analysis use short amplicons spanning variable regions among all prokaryotes (e.g., V1-V3, V3-V4 or V4-V5), these regions are not necessarily able to differentiate staphylococcal species (Frank et al. 2010). To further investigate this limitation, we computationally generated four virtual microbiomes, each containing an unequal mixture of fragments of the 16S rRNA genes of the three staphylococcal species aligned above. Using QIIME (a standard microbiome analysis software package (Caporaso et al. 2010)), we aligned and classified the sequences in the four mixtures to the species level, and compared them to the expected proportions of the three represented staphylococcal taxa (Fig. 1). Analysis results varied based on the database used (Greengenes or RDP) and the OTU-picking protocol (open or closed reference methods). Overall, all methods successfully detected a vast majority of staphylococcal sequences; however, the software was often unable to assign the appropriate species. In many instances, the species was not identified and in fewer instances the wrong species was assigned. The S. aureus-to-S. epidermidis ratios noticeably varied (Fig. 1), emphasizing the need of genetic markers markers other than the variable regions of the 16S rRNA gene for either species.

Comparative genomics
An ideal PCR to differentiate S. aureus from S. epidermidis in mixtures should target a gene present in all known S. aureus strains and absent in all S. epidermidis strains, and vice versa. If the latter condition is not achievable, the designed primers should at least not falsely generate a product of the same size from S. epidermidis, and vice versa.
As a first step towards this goal, we compared the core genomes of S. aureus and S. epidermidis using MetaRef (Huang et al. 2014

Primer design and in silico validation
We designed a primer pair for each of the two signature genes chosen and used BlastN to test each primer pair for sensitivity and specificity. For sensitivity, we aligned each primer pair with the complete set of available genomes from the species they are supposed to detect. The degenerate primers targeting staphostatin B were aligned to all

S. aureus genomes and those targeting staphostatin A successfully matched all
S. epidermidis genomes. Specificity was also confirmed, as none of the primer pairs had any BlastN hits (at a permissive E-value threshold of 10).

PCR analysis of staphostatin A and staphostatin B in representative strains and clinical isolates of S. aureus and S. epidermidis
We initially tested both primers on ten S. aureus and two S. epidermidis strains 14 available in our laboratory collection (Fig. 2). These included reference strains for both S.
aureus and S. epidermidis as detailed in the Materials and Methods section. Additionally, we tested the primers on 13 MRSA clinical isolates and 24 isolates from nasal swabs.
Finally, to exclude the probability of false positive or false negative results because of interference, we tested each set of primers on DNA extracted from both strains, as well as a mixture of S. aureus and S. epidermidis. PCR results were 100% accurate: the staphostatin A primers were specific to all S. epidermidis strains, and the staphostatin B primers were specific to all S. aureus strains.

PCR on staphylococcal isolates from nasal swabs
After testing the assay on DNA from laboratory and clinical isolates, mostly subcultured from frozen glycerol stock, we set out to test it on freshly isolated bacteria from nasal swabs. Out of 12 nasal swabs, we obtained 24 pure individual bacterial isolates, which we identified by traditional tools as S. aureus (n= 9) or S. epidermidis (n=15). The assay was able to detect the right species every time (100% sensitivity), and there was no cross-reactivity (Table 2 and Fig. 3).
For additional confirmation of bacterial identity, we amplified a large segment spanning most of the 16S rRNA gene from representative S. aureus and S. epidermidis isolates, using universal primers, and we sequenced the amplified products, which confirmed their identity.
PCR on community DNA Having tested the PCR assay on bacterial isolates and mixtures, we moved to the main application of the assay, which is differentiating between the two staphylococcal species in microbial communities without the need of cultivating or isolating bacteria.
DNA was directly extracted from the 12 swabs that were not used in culture-based analysis and tested for amplifiability by universal primers targeting the 16S rRNA genes (James 2010). Initially, only six out of 12 DNA preparations yielded amplification products with these universal primers, while the other six failed, possibly because of some interfering chemicals. Thus, DNA was ethanol-precipitated and resuspended in nucleasefree water. All precipitated DNA preparations successfully yielded amplification products.
Both staphostatin A and staphostatin B primers gave positive PCR results with all 12 nasal samples (Fig. 4).

Sensitivity and specificity of the developed assay
The developed PCR method allowed discrimination between S. aureus and S. epidermidis strains isolated from nasal samples. Furthermore, the results of PCR performed on the isolates matched the identification results obtained by standard microbiological diagnostic tools ( Table 1).
In addition, we successfully detected the presence of S. aureus using staphostatin B primers in all the 12 nasal samples, including the samples from which we could not isolate S. aureus by culture methods (five out of 12 nasal samples). On the other hand, staphostatin A primers gave a positive PCR with two nasal samples from which we could not isolate S. epidermidis strains (Table 2 and Fig. 5).
DNA sequencing to confirm PCR specificity 16 The specificity of our designed primers for staphostatin A-and staphostatin Bencoding genes was confirmed by comparison of the PCR product sequences against NCBI GenBank.
In some instances, a non-specific band of ~400 bp. appeared in some instances with community DNA amplified by staphostatin B primers. This PCR product was sequenced, and the sequence matched the sucC gene, which encodes succinate-CoA ligase subunit beta in S. epidermidis ( Table 3). However, this band was easily distinguishable on the gels because of its different electrophoretic mobility than that of the specific bands.

Discussion
For decades, microbiologists have been developing a series of culture media and biochemical tests for systematically identifying bacterial species. These culture-based methods are valuable in identifying some bacteria to the species or even strain levels; however, in many instances it is a challenge to differentiate closely related bacteria or to identify them in a community.
Exploring new methods for accurate identification and differentiation of closely related bacterial species is a continuous and pressing need. Even when a reliable assay is established, the discovery of novel strains, or novel strain variants, may sometimes challenge the well-established methods. Additionally, there is a need for simple, rapid assays for special purposes, such as resolving commonly cohabitating groups of microbes or distinguishing species in microbial communities.
This work focused on the discrimination between the two closely related species, S. aureus and S. epidermidis, both of which colonize the human nasal cavity. The isolation of these two bacteria and their identification is pivotal for microbiological diagnosis and epidemiological monitoring, especially in a hospital context.
Surprisingly, in many instances, it was not possible in this work to distinguish S. aureus and S. epidermidis by the existing methods (e.g., culture on MSA or coagulase test), notably when the bacteria were isolated from nasal swabs. Our observation agrees with published reports. For example, coagulase-negative S. aureus strains have been isolated since 1988 (Woo et al. 2001). A recent study reported that the specificity and sensitivity of the common tests for S. aureus (culture on MSA, DNase, and tube coagulase tests) is higher when the three tests are used together (100% vs. 75% if any test is performed alone (dos Santos et al. 2015)). While commercial identification systems, e.g., API, present an appealing alternative, they are not widely or routinely used in developing countries because of their high running cost and because of resource limitations.
On the other hand, molecular methods outweigh traditional identification methods in many aspects, including their higher sensitivity and specificity, speed, and-above alltheir independence on bacterial isolation and enrichment before the actual tests are performed. Although cost is an issue with molecular methods, in general, in many cases the running cost is lower when the assays are routinely used. By eliminating intensive labor, and the need for several media and reagents, these methods cut the direct and indirect costs on the long term. The simplest and almost ubiquitously used molecular identification method at the time being is the polymerase chain reaction (PCR).
A few genes have been targeted for simultaneous detection and discrimination of S. aureus and S. epidermidis ( Chiang et al. 2012;Hirotaki et al. 2011). These gene targets 18 have either been used in combination or in the context of multi-pathogen preset real-time PCR assays.
For the differentiation between the two Staphylococcus species targeted in this work, 16S rRNA, 23S rRNA, fib, hsp60, cpn60, tuf, femA, and nuc genes have been suggested (Baron et al. 2004;Ghebremedhin et al. 2008;Goh et al. 1997;Hirotaki et al. 2011;Sunagar et al. 2013). Despite the sensitivity and universality of 16S rRNA gene sequencing in identifying many bacteria up to species level, the similarity between 16S rRNA genes of S. aureus and S. epidermidis exceeds 98% (BlastN analysis and multiple alignment shown in Supplemental Fig. S1), and therefore 16S rRNA gene surveys, which rely on the partial sequencing of variable regions within the 16S gene, cannot resolve most mixtures of the two staphylococci. Moreover, 16S rRNA analysis requires amplicon sequencing, which adds time and cost to the assays (Saruta et al. 1995;Zakrzewska-Czerwinska et al. 1992).
In particular, the V3-V4 region, most commonly used in microbiome analysis, cannot fully differentiate staphylococcal species ( Fig. 1 and S1). While this variable region may resolve S. aureus in most instances, it cannot differentiae S. epidermidis from other CoNS, and this differentiation is database dependent (Frank et al. 2010). We validated this differentiation through in silico simulation experiments, which confirmed the variability between different OTU-picking strategies and reference databases-even when the same software was used (Fig. 1).
Analyses based on partial 16S rRNA amplicons lead to both false-positive (misclassifications, i.e., assigning species that do not exist in the community) and falsenegative results (unresolved species, i.e., limiting the results to the genus level and failing to determine the exact species).
For the PCR-based method developed here, we chose two previously unused marker genes: sspC and ecpB, encoding Staphostatin B in S. aureus and Staphostatin A in S. epidermidis, respectively. Both staphostatins are cytoplasmic inhibitors of their corresponding secreted cysteine proteases (staphopain A and B) and thus protect the cytosolic proteins against the proteolytic activity of these two proteases (Kantyka et al. 2011). Staphostatin B plays other roles in the regulation of S. aureus growth, the ability to form a biofilm, and the organization of peptidoglycan layer in the cell envelope (Shaw et al. 2005). In S. epidermidis, both the cysteine protease and its inhibitor are encoded within the same operon ecpAB, while in S. aureus, both the protease and its inhibitor are encoded in the same operon sspABC (Martínez-García et al. 2018).
We tested and performed our PCR analysis on DNA extracted from isolated colonies or directly from nasal swabs, while previously developed methods were usually tested on cultured bacteria only. No cross-reactivity was observed as we tried each primer pair on either species or on mixtures of both species. The PCR results for laboratory strains and isolated colonies were confirmed by the traditional biochemical tests (which usually work properly on isolated colonies in the absence of contamination). For further validation, we sequenced the 16S rRNA gene from representative isolates and confirmed their identity as well. It is to be noted that this sequencing of a near-full length amplicon of the 16S rRNA gene is different from the methods based on 16S rRNA amplicon libraries, commonly used in microbiome analysis, as the latter targets small fragments of the 16S rRNA gene.
Even when non-specific products incidentally appeared on amplifying DNA from microbiome samples with high microbial burden, these products were of different sizes, which would not interfere with the ~200 bp amplicon size of staphostatin amplicons.

20
Finally, the PCR method described here was more sensitive in detecting S. aureus in five nasal samples, from which this species was not detected by culture-based methods.
This result can either be seen as a sign of higher sensitivity or as a possible false-positive amplification. However, preliminary microbiome analysis indicated that S. epidermidis was much more abundant than S. aureus in these five samples (manuscript in preparation).
The probability of isolating S. aureus from nasal swabs was reported to drop tenfold if S. epidermidis counts were higher than 10 5 CFU/swab (Lina et al. 2003), which might explain why S. aureus could not be isolated from these nasal swabs with higher ratios of

S. epidermidis.
Future studies will address the epidemiological value of the proposed method by conducting large-scale studies to screen patients in hospitals, health-care workers, and community members for carriage of either of the two staphylococcal species. Additionally, future experimentation may address the impact of excessive organic material on the assay, if conducted-for example-on samples contaminated with patient blood or excretions. Such details are not in the scope of this manuscript.

Conclusions
Our validated PCR-based method can be used for the differentiation between two closely related bacterial species, S. aureus and S. epidermidis, present in pure colonies or in clinical samples. Both staphostatin A and staphostatin B are good biomarkers for S. epidermidis and S. aureus, respectively. Accordingly, this method will help to overcome the difficulty of differentiating the two species in microbiome surveys, which pose a challenge to culture-based methods.