Identification and Antimicrobial Susceptibility Testing of Anaerobic Bacteria: Rubik’s Cube of Clinical Microbiology?

Anaerobic bacteria have pivotal roles in the microbiota of humans and they are significant infectious agents involved in many pathological processes, both in immunocompetent and immunocompromised individuals. Their isolation, cultivation and correct identification differs significantly from the workup of aerobic species, although the use of new technologies (e.g., matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, whole genome sequencing) changed anaerobic diagnostics dramatically. In the past, antimicrobial susceptibility of these microorganisms showed predictable patterns and empirical therapy could be safely administered but recently a steady and clear increase in the resistance for several important drugs (β-lactams, clindamycin) has been observed worldwide. For this reason, antimicrobial susceptibility testing of anaerobic isolates for surveillance purposes or otherwise is of paramount importance but the availability of these testing methods is usually limited. In this present review, our aim was to give an overview of the methods currently available for the identification (using phenotypic characteristics, biochemical testing, gas-liquid chromatography, MALDI-TOF MS and WGS) and antimicrobial susceptibility testing (agar dilution, broth microdilution, disk diffusion, gradient tests, automated systems, phenotypic and molecular resistance detection techniques) of anaerobes, when should these methods be used and what are the recent developments in resistance patterns of anaerobic bacteria.


Introduction
Anaerobic bacteria have been implicated in a wide range of infectious processes. As an integral part of the human microbiome, these microorganisms can be found in different anatomical sites and they can be responsible for a plethora of infections that may be serious or life-threatening [1][2][3][4]. For this reason, anaerobes are often categorized by the body site(s) where they occur, the infections they are associated with, although they are most frequently classified by their microscopic morphology (rods or cocci) and their Gram-staining (for a summary of the most important genera of anaerobic bacteria, see Table 1) [3,5,6]. Infections caused by anaerobes can also be divided according to the origin of the microorganisms. Exogenous anaerobic infections are predominantly caused by Gram-positive spore-forming bacilli (Clostridium spp.; C. difficile being an exception to this rule), these microorganisms are the causative agents of serious infections like visceral gas gangrene, lockjaw, myonecrosis and botulism. These infections are monobacterial in almost all cases and the effects of bacterial toxins are When it comes to the cultivation of anaerobes in a clinical microbiology laboratory, accepting only the samples appropriate for such purposes is of utmost importance. Specimen collection should be done using suitable equipment, while making sure that the sample is not contaminated with the resident flora of the body site in question. If sample processing is taking place somewhere else (e.g., is sent to a reference laboratory) adequate precautions need to be met for the use of anaerobic transport media and storage. The following specimen types should be considered for cultivation: samples taken from normally sterile body sites (e.g., blood cultures, cerebrospinal fluid samples, synovial fluid, samples from the chest and abdominal cavity), surgical discharge (or sample taken from the surgical site), samples from deep wounds (e.g., animal bites), abscesses of different anatomical localisation (liver, brain, lungs), perioral and gingival samples, samples from diabetic foot ulcers, respiratory secretions taken with a double-lumen device, urine samples (taken only with suprapubic aspiration) and faecal samples (in the suspected case of C. difficile infection). The best specimens for anaerobic cultivation are those taken using a needle and a syringe. Any other sample type received should not be accepted for anaerobic processing since they do not meet the requirements (risk of oxygen exposure: false negative result; contaminants: false positive results and/or cultivation of microorganisms with no clinical significance) [5,6,49,[53][54][55].
Adequate laboratory conditions are necessary for the cultivation and diagnosis of anaerobic bacteria, the most important of which is a suitable method to anaerobiosis (stable anaerobic environment for cultivation) and the availability of pre-reduced anaerobically sterilised (PRAS) media. Some physical (McIntosh-Filde's jar), chemical (e.g., basic pyrogallol to bind atmospheric oxygen) and biological (Fortner's method; co-cultivation with a strong aerobic metabolism in a split Petri-dish, e.g., Serratia marcescens) methods are now only of historical significance. Depending on the size of the clinical laboratory and the number of incoming isolates, GasPak sachets, anoxomat systems or anaerobic chambers (glove boxes) are used, although the availability of some of these is limited to anaerobic reference laboratories [5,6,49,53,56].
Anaerobic bacteria are fastidious microorganisms, culturing them requires nutrition-rich media and some of these include components that allow for the selective growth of these bacteria. Examples include liquid enrichment media like Holman-broth (chopped meat and carbohydrates) and thioglycolate-broth and a variety of pre-reduced anaerobically sterilised solid media (anaerobic blood agar, Brucella blood agar supplemented with 5% lysed horse blood, 5 [5,49,57]. A variety of methods can be used for the identification of anaerobic isolates, from presumptive identification (based on growth characteristics, colony morphology, susceptibility to given antibiotics, presence or absence of fluorescence, classical biochemical tests and spot-tests) to commercial kits (e.g., Vitek 2 ANI card, API 20A, RapidID 32A: bioMérieux, Fr., MicroScan, BBLCrystal Anaerobe ID: Beckton Dickinson, UK) [49,58]. When it comes to the biochemical ID, it is important to be aware of the enzymes we target in an assay. Reactions involving constitutive enzymes can be included in rapid identification kits, while in the case of inducible enzymes, the presence of the substrate and additional time for the reaction is required. An additional method, specific for these bacteria, is available: using gas-liquid chromatography (GLC), a rapid, accurate and reliable identification method of anaerobes is possible, based on the alcohols, organic acids and short-chain fatty acids they produce [59]. This method can also be used to differentiate between aerobic and anaerobic infections (e.g., from a haemoculture), many laboratories however, do not possess the instruments required to carry out this assay [60,61]. For further reading on the phenotypic methods concerning anaerobic bacteria, see the Wadsworth-KTL Anaerobic Bacteriology Manual, which is considered as a reference in the phenotypic identification of these microorganisms [5].
The introduction of molecular methods, array-based systems and novel technological advancements in clinical microbiology, such as whole-genome sequencing (WGS) and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (which utilizes the spectra of conserved ribosomal proteins for identification) has revolutionised the detection and correct identification of anaerobic bacteria. Although the development of databases containing bacterial spectra, required for identification is still underway (with projects like the 'European Network for Rapid Identification of Anaerobes' [ENRIA] leading the charge), MALDI-TOF MS changed the face of diagnostic bacteriology in the last decade, both in the case of aerobes and anaerobes [62][63][64][65][66][67][68][69][70][71]. Whole-genome sequencing is a novel technology used for culture-independent diagnostic testing (CIDT). WGS is considered the "golden standard" of bacterial species identification, based on the complete sequencing of their genetic material and then, comparing the data to reference genomic libraries based on bioinformational methods. Nevertheless, while there are some commercial MALDI-TOF MS systems available, specifically for all purposes (including anaerobes and fungi) of routine clinical microbiology (e.g., VITEK MS: bioMérieux, Fr., MALDI Biotyper or Microflex LT of Bruker-Daltonics, Gr.) [72,73], sequencing technologies are-in most cases-still in their developmental/experimental phase (and not necessarily well characterized for anaerobes) [29,62,[74][75][76]. Besides the growing interest in the research related to anaerobic bacteria, the description and verification of new species has also been facilitated by the introduction of these new technologies-although such research is limited to reference laboratories with the appropriate facilities [20,77,78].
Compared to the processing time of aerobic bacteria or facultative anaerobes the workup of anaerobic isolates usually takes much longer. For this reason, continuous communication between the clinicians and the diagnostic lab is crucial. The laboratory should supply any clinically relevant information in a precise and timely manner, while the feedback of the physicians is also important, since based on the information about the symptoms and the clinical picture, the lab can narrow down (or include more specific tests) the list of the possible causative agents. If anaerobic bacteria were present at the infection site and we failed to detect them, it can often lead to inappropriate therapeutic choices and clinical failure [8].

Members of the Bacteroides Fragilis Group
Anaerobic Gram-negatives include some of the most important human pathogens among the group of anaerobic bacteria. The group is included in the phylum Bacteroidetes (together with pathogens like Capnocytophaga canimorsus, Porphyromonas gingivalis, Prevotella melaninogenica and Tannerella forsythia) [6,50]. Together with the taxonomy of other anaerobic species, the classification of Bacteroidetes has undergone major changes due to findings of studies related to molecular, sequencing and other related metagenomic studies. The most profound changes can be seen in the genus Bacteroides, in which mostly Gram-negative, bile-resistant bacilli remained ( Table 2). Species that were saccharolytic, bile-sensitive (pigmented [formerly known as B. melaninogenicus] or nonpigmented) were relocated to the genus Prevotella, while the asaccharolytic, pigmented species to the genus Porphyromonas [24]. The previous classification of the group was as an order of subspecies (e.g., B. thetaiotaomicron was previously known as B. fragilis ssp. thetaiotaomicron) but later was reclassified based on studies related to the DNA homology of the respective species. The members of the Bacteroides spp. are important constituents of the faecal bacteria flora of adult humans, they have an active role in the metabolism of bile salts, nitrogen-containing substrates and various carbohydrates and polysaccharides (e.g., arabinogalactan, starch, pectin, xylan) [6,23,50].
The antibiotic resistance levels among anaerobes is the best characterized in Bacteroides isolates and according to the relevant literature, the species of the genus have the highest levels of antibiotic resistance and the most numerous number of antibiotic resistance mechanisms among all human pathogenic anaerobic bacteria (described in detail in Section 5) [79,80]. It is important to note that resistance levels of these bacteria are in inverse relationship with their incidence and clinical significance: P. distasonis and B. thetaiotaomicron (13-23% of Bacteroides isolates) exhibit the highest levels of resistance, while the strains of B. fragilis is generally the most susceptible species in the B. fragilis group.
B. fragilis is a pleomorphic (with varying sizes of 1.5-6 µm), non-pigmented, non-motile, encapsulated, Gram-negative rod. The species is catalase-positive (which is not common for anaerobe bacteria), indole-negative and the clinical isolates are growing well in the presence of 20% bile (bile resistant). They do not produce fluorescence under UV-light and they are resistant to vancomycin, kanamycin and colistin antibiotic disks, which can be used for their presumptive identification. B. fragilis is the most frequently isolated member of this group and although they only represent 2-5% of the species isolated from the fecal flora, it accounts for >50% of the clinical isolates of the Bacteroides genus and is responsible for around 80% of infections caused by the genus [50].
Certain strains termed enterotoxigenic B. fragilis (ETBF) produce a metalloprotease enterotoxin, which is coded by the bft genes on the B. fragilis pathogenicity island (BfPAI). The disease was first described in new-born lambs and calves, nowadays it is acknowledged as a significant cause of a self-limiting diarrheal disease in adults. Certain reports drew the conclusions that colonisation with enterotoxigenic B. fragilis is also associated with higher risk for colorectal cancer and irritable bowel syndrome (IBS), however there is still debate over the conclusions of these results. The proper diagnosis of enterotoxigenic B. fragilis infection is laborious, detection of the toxin genes (bft1-3) by polymerase chain reaction or the determination of the biological activity of the BFT protein in a cell culture assay (for this purpose, only those cell lines can be used, which have the ability to polarize in vitro, e.g., HT-29, Caco-2, HCT-8, MDCK, etc.) is needed. Animal models have also been described to verify the presence of the enterotoxin of B. fragilis, however the use of these models is only appropriate in a research setting [50,81,82]. A very important virulence factor of these bacteria is the presence of the capsule (consisting of two, well characterised polysaccharides), the significance of which in abscess formation was proven earlier in animal studies [83]. Additional virulence determinants of this species include the presence of various fimbriae and adhesions, which help the bacteria to adhere to matrix proteins of the host [50]. The most significant role of B. fragilis as an infectious agent is thought to be in anaerobic bacteraemia, where the associated mortality is estimated to be around 19% (60% if left untreated) [84,85].
B. fragilis is frequently termed as the "anaerobic Escherichia coli" due to the many parallels in the characteristics of these microbes: (i) both are bile-resistant Gram-negative rods; (ii) both colonise the colon and have integral roles as a part of the microbiota; (iii) both have the proclivity to become multidrug resistant pathogens through a variety of resistance mechanisms (iv) some strains produce enterotoxin (enterotoxigenic B. fragilis vs. enterotoxigenic E. coli) and capsule (in the case of extra-intestinal pathogenic E. coli strains). The presence of this species was associated with pathogen synergy, it has been described that the presence of B. fragilis and E. coli in mixed intra-abdominal infections enhanced the anti-complement environment of the infection site through bacterial virulence factors [86].

Members of the Clostridium Genus
Anaerobic bacteria were traditionally classified into the Clostridium spp. based on the following characteristics: (i) their staining result was Gram-positive (ii) being strict anaerobic (iii) formation of characteristic endospores (their position is useful in species identification; C. tetani has round, terminal spores, C. botulinum and C. difficile forms oval, subterminal spores, C. perfringens has central spores with oval shape, i.e., in the appropriate environment) (iv) inability to reduce sulphates (SO 4 2− ) into sulphites (SO 3 2− ) [6,7]. Nevertheless, the classification of these bacteria based on these phenotypic criteria has become difficult, since some species do not form spores or only do in specific conditions (e.g., C. perfringens, C. ramosum), some are uncharacteristically tolerant to atmospheric oxygen (e.g., C. tertium, C. histolyticum), while others consistently present as Gram-negative after staining (C. ramosum, C. clostridiforme). Due to these discrepancies, it's not surprising that the reclassification of multiple species to various other genera (e.g., Butyrivibrio, Dendrosporobacter, Enterocloster, Eubacterium, Faecalicatena, Lacriformis, Sedimentibacter) has been proposed, based on 16S rRNA gene sequencing data [87,88]. Species of the Clostridium spp. are transient or permanent members of the normal flora of the gastro-intestinal tract and skin of animals and humans alike. Their spores naturally occur in the soil, their presence is facilitated with the use of manure as fertilizer. It is important to point out that many clinical specimens could contain Clostridium spp. as accidental contaminants, not involved in the disease process. It is up to the clinician to decide, whether the presence of these bacteria has any clinical significance, based on the symptoms of the patient, the presence of other microorganisms of pathogenic potential, frequency of isolation of the species and local epidemiology. The adequate management of clostridial infections involves administering antibiotic therapy (e.g., penicillin, vancomycin, metronidazole), with tissue debridement, where it is necessary. In some cases (like tetanus) toxoid immunization and antitoxin therapy are also important tools for clinicians [1,2,6,7].
The pre-requisite for clostridial wound infections is trauma to host tissue, to lower the oxidation-reduction potential and provide an environment suitable for anaerobic growth, followed by accidental contamination of the wound. Clostridial wound infections usually are polymicrobial in nature, because the sources of wound contamination (faeces, soil) is polymicrobial, involving a variety of other bacterial isolates (e.g., Bacillus spp., Bacteroides spp., Escherichia spp., Proteus spp., Staphylococcus spp.) [6,7]. All clostridial species involved in human diseases (see Table 3.) produce some kind of protein exotoxin (e.g., tetanus toxin, botulinum toxin) which plays an important role in the corresponding pathologies. Based on the pathogenesis of the disease caused by these microorganisms, they can be divided into the group of histotoxic clostridia (or gas-gangrene clostridia: C. perfringens, C. novyi, C. septicum, C. histolyticum) and neurotoxic clostridia (lockjaw/tetanus: C. tetani; botulism: C. botulinum, C. baratii, C. butyricum). Toxin-producing C. difficile strains represent an additional group, as the causative agents of antibiotic-associated diarrhoea (AAD), pseudomembranous colitis and toxic megacolon [6,7].
Various strains of C. perfringens produce potent necrotizing and haemolytic toxins and enzymes, based on this quality, these strains can be classified into types A-E. All types produce the α-toxin (the principal virulence factor of this species), which is a lecithinase that destroys the membranes of red and white blood cells and of surrounding tissue cells. The production of α-toxin can be detected by the Nagler-reaction, using egg-yolk agar (EYA). Other toxins of C. perfringens have a haemolytic, necrotizing and cardiotoxic effect. Among its other enzymes, the most important are collagenase, hyaluronidase and deoxyribonuclease. Some of the type A strains produce a heat-labile enterotoxin during sporulation (often occurring in the small intestine, at a relatively high pH), binds to the membrane of the epithelial cells of the small intestine, causing diarrhoea. The activity of this toxin is greatly increased by digestion with trypsin [6,7,89]. Tetanospasmine (or tetanus toxin, which is a typical AB-type exotoxin), the most important virulence factor of C. tetani, is released by dissolution the microbial cell. The toxin can reach the spinal cord, partially through the blood stream and the lymphatic system but mainly and typically by retrograde axonal transport through the peripheral neurons. The active A subunit, which is a zinc-endopeptidase, inhibits the release of inhibitory neurotransmitters (glycine and γ-aminobutyric acid) from inhibitory neurons. Thus, the proliferation of acetylcholine generates a continuous activation resulting in spastic, convulsive paralysis (causing the characteristic symptoms of the disease) [6,7].
The botulinum toxin (or botulotoxin) is the exotoxin of C. botulinum. It is produced as a consequence of lysogenic conversion, seven different (A-G) structurally different variants are known, of which the A, B, E and F-type toxins have been reported in human diseases. For any symptom to occur, whether or not it is caused by preformed toxins (classical food poisoning) or the effect of toxins produced by the microorganism in the body (infant and wound-botulism), the botulinum toxins is responsible. Botulinum toxin is a very powerful poison and in contrast to other toxins, protein-digesting enzymes of the intestinal tract do not hydrolyse it but it is heat-labile, meaning that the proper treatment of the food (at 80 • C for at least 10 min) inactivates it. In many respects, it is similar to the tetanus toxin (it is also an AB-type zinc-endopeptidase). However, unlike tetanus toxin, the botulinum toxin does not migrate retrograde to the neuronal cell, after entering the neuromuscular junctions through endocytosis. The toxin cleaves the protein required for acetylcholine release in the presynaptic neuron, the release of acetylcholine is thus omitted and in the absence of the neurotransmitter, the distinct symptoms of flaccid paralysis can be observed [6,7,90].
Similarly to other members of the genus, exotoxins are responsible for the symptoms of C. difficile infection as well. The A-toxin is an enterotoxin, which causes injury to intestinal cell-cell connections with subsequent increase in the permeability of the intestinal epithelium and fluid secretion. It also enhances the migration white blood cells, inducing a local inflammatory reaction. B-toxin is a cytotoxin that damages the cytoskeleton of the cells by disorganizing intracellular actin filaments. The two toxins have a synergistic effect but the A-toxin negative strains seem to be just as capable of causing illness and even causing hospital epidemics. The third toxin is the so-called binary toxin (or CDT), the exact mode of action of which is unknown, although it has been recognized that strains producing binary toxin are associated with higher mortality and produce A and B toxins in much larger quantities than CDI-negative strains [6,17,18].

Opportunistic infections
The bacteria in bold represent the most prevalent species in the group.

General Considerations
According to the international guidelines, the susceptibility testing of anaerobic bacteria is very expensive, time-consuming and requires experienced laboratory staff, testing every patient's isolate received in a routine laboratory is not warranted. There are some diagnostic institutions (e.g., in low-income countries, with no technical capabilities) where even the cultivation of these microorganisms is a challenge. In these cases, isolates are usually sent to higher-tier facilities or a national anaerobic reference laboratory. Based on current recommendations, susceptibility testing should be performed in the following cases: (i) infections of a serious and life-threatening nature (e.g., in endocarditis, bacteraemia, abscesses involving the brain) [2,10,84]; (ii) relapsing infections or infections that did not respond to empirical therapy at all; (iii) when the antibiotic therapy needs to take place for an extended time period (e.g., infections involving bones, joints, implanted devices or grafts) [12,13]; (iv) there is limited or no available data on the susceptibilities of the given organism; (v) there is a known pattern of resistance against the antimicrobial agent by the microorganism [79]; (vi) the causative agent is a particularly virulent anaerobe (see Table 4) with unpredictable resistance [8,23]; (vii) the microorganism was isolated in pure culture and/or from a normally sterile body site (see Section 2) [8,23]. Besides the abovementioned points, susceptibility testing should be performed wherever it is possible, for epidemiological purposes and to guide the choice of the therapeutic agents [8,91,92]. This aspect of surveillance (both locally and globally) needs to be reiterated, since recommendations on first-line agents of therapy is usually based on similar data (Table 5). For example, due to the high level of resistance, cefoxitin and cefotetan are not advised as first-line drugs and clindamycin has been completely removed from such recommendations (although it can still be used in various oral infections and aspiration pneumonia), with beta-lactam/beta-lactamase inhibitor combinations emerging as first line drugs next to metronidazole [1,93]. It is also important to realise that different susceptibility testing methods are appropriate for various needs and end points, e.g., whether it is performed in a diagnostic laboratory of a hospital, for epidemiological/surveillance purposes in a reference laboratory or in a research institution [94]. Table 4. List of anaerobic bacteria where routine susceptibility testing is recommended [23].

Bacteroides fragilis group
Bilophila wadsworthia Clostridium innocuum Clostridium perfringens Clostridium ramosum Fusobacterium spp. Prevotella spp. Sutterella wadsworthensis novel clinically important anaerobes (where surveillance data is not available) Table 5. Antimicrobial agents that should be included in the routine susceptibility testing on anaerobic isolates [11,23,92].  [96], the British Society for Antimicrobial Chemotherapy (BSAC; which has since harmonized its criteria with EUCAST standards [97,98]) and the Deutsches Institut für Normung (DIN; German Institute for Standardization) [99]. The ESCMID Study Group for Anaerobic Infections (ESGAI) [100] and the Anaerobe Society of Americas (ASA) [101] play an equally important role in the facilitation of proper practice in the case of these bacteria.
Since there is no specified testing method for anaerobes by EUCAST, the standard procedures established by CLSI (currently M11-A8) are most frequently used [102]. According to this document, broth microdilution and agar dilution are the reference methods for this group of bacteria, although different methods are also accepted if their uniformity is confirmed with the reference method. CLSI also recommends that at least 100, randomly selected isolates of a given genera should be tested for their resistance patterns, to acquire adequate information on the regional (hospital) patterns of resistance [102]. Irrespective of the method used for the susceptibility testing on anaerobic bacteria, these all share the following: (i) without the proper (stable) level of anaerobiosis during the incubation period, the results cannot be interpreted [103]; (ii) since the generation time of these bacteria is longer than their aerobic counterparts, results of the susceptibility testing can only be interpreted after 48-72 h (or more); (iii) extensive quality control and staff experience is required [94]; (iv) breakpoints determined by CLSI and EUCAST do not always match, making interpretation of the susceptibility results difficult [102,103].

Disk Diffusion Method (Kirby-Bauer Method)
Disk diffusion susceptibility testing is an easy-to-perform and cost-effective technique developed by EUCAST, their standardized methods and clinical breakpoints were embraced by clinical microbiology laboratories in Europe as well as in other parts of the globe [104]. Due to the popularity of this method, several studies (by EUCAST or otherwise) aiming at the evaluation of the suitability of this technique for the use in routine anaerobic bacteriology (mainly aiming at B. fragilis group isolates and "fast-growing" anaerobic strains) were performed [105,106]. Although some of these studies showed promising results in reproducibility (the inhibition zone diameters correlated well with the MICs in the case of metronidazole, imipenem, moxifloxacin and tigecycline), in other cases intermediate (I) and susceptible (S) isolates had overlaps in their inhibition zone diameters (in the case of beta-lactam/beta-lactamase inhibitor combinations and cefoxitin) [107]. Since the criteria for the use of this method has only been determined on aerobic bacteria so far, antimicrobial susceptibility testing of anaerobes using the disk diffusion method is not recommended in the routine lab yet and it is only used in a research setting or for preliminary screening for the antibacterial activity of different bioactive compounds [105]. However, "identification" disks containing specific antibiotics (usually colistin, kanamycin, metronidazole, penicillin and vancomycin are used) and sodium polyanethole sulfonate (SPS) are routinely used in the clinical microbiology laboratories for the presumptive identification of these bacteria [5].

Broth Microdilution Method
Broth microdilution assay is usually carried out using 96-well microtiter plates, where two-fold serial dilutions of the antibiotics are made. Commercial and in-house preparations are both available and the latter is more appropriate for the particular needs of the laboratory (various drugs can be tested in different concentrations), using only small volumes of reagents and media (Brucella broth supplemented with 5% lysed horse blood, 5 µg/mL hemin and 1 µg/mL Vitamin K1; according to CLSI recommendations) in the process. The results can be interpreted visually or with the help of a photometer. The results of the broth microdilution method are reported in Minimum Inhibitory Concentration (MIC) values, or the lowest concentration of antibiotics that stopped bacterial expansion. The commercially available standard broth microdilution panels are the breakpoint panels, in which only one or a few concentrations of each antimicrobial agent are tested in a single panel. Unfortunately, this method is not an all-round solution for anaerobic bacteriology, since apart from Bacteroides spp. the growth of other species is not consistent in broth (homogenous growth cannot be attained, which leads to reporting false susceptibility, spore-forming bacteria with no visible growth can also be falsely reported as susceptible). Another significant drawback of this method is that it can only be performed in an anaerobic glove box (for the appropriate level of anaerobiosis to be maintained), therefore its use in the clinical microbiology lab is limited [108].

Agar Dilution Method
Agar-dilution method is currently the gold standard for the antimicrobial susceptibility testing of anaerobic bacteria. During this procedure, nutrient agar plates (Brucella Agar supplemented with 5% laked sheep blood, 5 µg/mL hemin and 1 µg/mL Vitamin K1; according to CLSI recommendations) are made with various antibiotics incorporated in the media in different concentrations, following with the inoculation of the plates with a standardized number (inocula of 0.5 McFarland standard from a pure culture incubated for 48 h) of bacterial cells, usually with the help of a Steers-Foltz replicator device. After incubation of the plates for 48-72 h in anaerobic atmosphere, the plates are read and compared visually. Based on this method, the lowest concentration of antibiotic that inhibited the growth of the given bacterial strain is the MIC. Agar-dilution is an labour-intensive method recommended for anaerobic reference laboratories [23,102].

Gradient Tests (E-Test, Spiral Gradient Test)
Gradient tests such as the E-test are the most frequently used in hospital laboratories for performing anaerobic susceptibility testing. During the procedure, a plastic/paper strip (containing a pre-determined amount of antibiotic in a gradient fashion and a corresponding scale for MIC determination) is placed on the plate, after inoculation with the sample. It is recommended to dry the agar plates before applying the strips, to prevent excess moisture to change the characteristic concentration gradient. The zone of inhibition will present itself in an elliptical shape (hence the 'E' in the name of the test), the MIC should be read at the point where the zone of inhibition intersects the strip. The MIC reading of the E-test is independent of the number of colony forming units (CFU) of the inoculated sample, therefore the results should be considered reliable, and there is good correlation with the reference method. E-test can be used for the detection of heteroresistance (e.g., in the case of imipenem resistance of B. fragilis or in the case of metronidazole resistance of C. difficile and B. fragilis) as well [109,110]. It is easy to perform and it is especially useful for the testing of a small number of isolates or only one antibiotic in laboratories without a steady flow of numerous anaerobic isolates.
The spiral gradient endpoint (SGE) technique is a special susceptibility testing method, which entails the use of a special agar plate, on which an apparatus deposits a specific amount of antibiotic stock solution in a spiral pattern [111]. By doing so, the concentration of the antibiotic is decreasing from the centre of the plate, creating a concentration gradient. After the inoculation of the plate with the relevant bacteria in a radial manner (thus, a single plate can be inoculated with multiple isolates and their susceptibilities to a single antibiotic can be determined simultaneously), the extent of growth is marked and the distance of the colonies from the centre of the plates is recorded. This data is then entered into a software (usually supplied by the manufacturer) which takes into account the physico-chemical characteristics of the antibiotic when determining the MIC related to the isolate in question. The emergence of resistant mutants can also be detected with this technique (colonies growing beyond the endpoint can be observed), while clumping of the bacteria was described (making interpretation difficult) when high-density inocula was used [112]. This method is becoming more and more popular because it's easy-to-use and it showed favourable results, when compared with the results of agar dilution, however it is similarly high-priced to the reference method [111,113]. The lack of commercially available designs based on this method and the additional need to calculate the MIC from the observed results are some of the other drawbacks of this method [23].

Examples of Resistance Detection Using Phenotypic and Genotypic Methods
Some rapid tests are available for the detection of β-lactamase enzyme production of bacteria. Such methods, e.g., nitrocefin disks are practical due to the ease-of-use and the express results they provide. These colorimetric assays (a positive result is read if the disk changes to an orange or red colour in 5-60 min) are particularly useful if penicillin or ampicillin is the desired therapy [114]. During the interpretation of the results, it should be kept in mind that a negative result does not rule out resistance due to other mechanisms (e.g., changes in permeability due to poring loss or efflux pumps, modifications of PBPs) [115]. Performing this test is unnecessary on B. fragilis isolates, as the overwhelming majority produces β-lactamase enzymes [50]. Novel methods such as chromogenic media also present practical solutions for the use in the routine laboratory. An attempt for the development of such media was the Bacteroides chromogenic agar (BCA), in which 3,4-cyclohexenoesculetin-β-D-glucoside was included in the media instead of esculin to target the β-glucosidase activity of B. fragilis. In a precipitation reaction between the hydrolysed substrate with iron salts added to the media, an insoluble black precipitate appears, which allows for the differentiation of B. fragilis colonies from a polymicrobial isolate. Additionally, supplementation of BCA media with the appropriate concentrations of meropenem or metronidazole allows for the rapid detection of resistant strains of B. fragilis (e.g., from faeces), although this method is not yet routinely used in diagnostic laboratories [116].
Detection of resistance genes using polymerase chain reaction is a reliable method of identifying various antibiotic resistance determinants (for example: nim genes responsible for metronidazole resistance, erm genes conferring resistance to the macrolide-lincosamide-streptogramin (MLS) group of antimicrobials, the cfiA gene of carbapenem resistance, gyr and parC implicated in quinolone resistance, cat gene of chloramphenicol resistance, various tet genes making the microorganism resistant to tetracycline etc.) [23,117]. The end point would be the development of a multiplex PCR or a similar complex system, which would hopefully be able to identify multiple genes of resistance at once [117]. However, the development of similar methods is hindered by the fact that the presence of the resistance determinant genes in the genome does not automatically confer resistance to the antibiotics in question. This phenomenon has been well defined in the case of B. fragilis group isolates, harbouring "silent" cfiA or cepA genes (requiring specific insertion sequence (IS) elements to be activated), thus their genotypic resistance did not reflect their phenotypic resistance (i.e., their in vitro antimicrobial susceptibility patterns) [23,117,118].
Various experimental studies were conducted to test the potential future applications of matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry in the detection of various resistance determinants in anaerobes. In these studies, B. fragilis was predominantly used as a test organism for the detection of carbapenem resistance (as one of the most worrying resistance trends) from standard laboratory strains and later from clinical samples, such as blood cultures [118][119][120]. There is great interest in MALDI-TOF MS-based resistance determination, due to the superior speed of communicating the results, not to mention that these tests could be implemented on the mass spectrometry devices already purchased (with the addition of the appropriate database for detection). The detection of beta-lactamase production is also possible with the use of this technology. The hydrolysis of beta-lactam antibiotics by these enzymes results in a molecular mass shift, which can be detected by a MALDI-TOF MS-based assay, although due to its time-consuming nature, this method is not yet widely used [121][122][123].
Whole-genome sequencing for the prediction of antimicrobial resistance is a novel and emerging research area, the development of which is further facilitated by the emergence of reasonably-priced WGS systems [124][125][126][127]. Detection of known antibiotic resistant determinant genes from the bacterial strain isolated from the patients allows for a culture-independent method to predict the susceptibility pattern of microorganisms. This makes WGS a useful tool for surveillance purposes, even in the case of drugs that are usually not included in classical (phenotypic) susceptibility testing routines (e.g., in drug development). [74,128,129]. Additionally, resistant phenotypes caused by dissimilar genetic backgrounds can be differentiated [124,125]. According to recent studies on selected species groups, the concordance rate between the presence of antibiotic resistance genes and the actual antibiogram (phenotype) of the microorganism is between 72-99%, although the use of this method is still in its experimental stages, due to the lack of properly developed reference libraries of bacterial resistance determinants (seeing that only the resistance genes included in the web-based databases will be recognized-a common disadvantage of molecular diagnostic methods) [129,130].
Some data available on the resistance trends of anaerobes (apart from hospital-level or regional surveillance studies conducted where resources were available) is restricted to the studies published by anaerobic reference laboratories or major trans-national collaborations [23,[132][133][134][135][136]. Although the variations between different geographical regions are notable, common tendencies can be observed.
While three decades ago the antibiotic susceptibility pattern of anaerobic bacteria was straightforward, nowadays we cannot so easily predict the efficiency of the chosen empirical therapy [91]. Clinicians can no longer "expect" certain drugs to work in anaerobic infections because they showed potent activity before [137]. Though not at the same speed as some other microorganisms (e.g., carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa or members of the Enterobacteriaceae [138]) resistance has steadily increased among anaerobes during the last 30 years.
With the use of broad-spectrum antimicrobials in addition to the suitable surgical measures, the issue of emerging resistance of anaerobic bacteria was maybe less obvious [93,131]. But if we observe the data from the (inter)national surveillance reports from Europe (Table 6) [139][140][141][142][143][144][145][146][147] and the United States (Table 7) [95, 135,[144][145][146][147], the same trends can be discovered: steadily growing resistance to all classes of antimicrobial agents, with some of them rendered completely useless. Several publications report the emergence of multidrug resistant anaerobic Gram-negative bacteria (especially within the B. fragilis group isolates), harbouring multiple resistance genes or with a combination of intrinsic and acquired resistance mechanisms [80,85,[147][148][149][150][151]. In these cases, the bacteria were usually termed multidrug resistant if they were resistant to 3-4 antibiotic classes besides metronidazole (due to nim nitroimidazole resistance genes) and the carbapenems (a metallo-β-lactamase encoded by cfiA/ccrA genes) [79]. Mobile genetic elements (plasmids containing resistance determinants, insertion sequence elements, transposons) have a significant role in the spread of the multidrug resistant phenotype in anaerobes.
The significance of the abovementioned tendencies is further underlined by the fact, that treatment failure has been described in empirical treatment in cases of anaerobic bacteraemia, as result of a multidrug resistant B. fragilis infection [9,84,[152][153][154][155]. What makes this problem even more insidious is the fact that the correlation between the presence of multidrug resistant anaerobic bacterial strains and clinical failure is hard to prove. Mixed infections that include anaerobic bacteria often improve by drainage or surgical debridement [2]. The well-being of the patient (chronic diseases, immunosuppression) and the effect of the administered antibiotics on the aerobic component of the infection could further influence clinical outcome. Lastly, if the specimen collection and processing of the anaerobic isolate was not appropriate, the suspected microorganism cannot be cultivated, even if there is strong clinical suspicion of its presence [63].  A summary of some of the important resistance mechanisms (and their genetic determinants) is presented in Table 8. While the growing problem of resistance stems from the fact that resistance is gradually developing against drugs that once were efficacious in treatment (analogous to the trends observed in other bacteria, e.g., Enterobacteriaceae) it is important to be aware that anaerobes-owing to their biological nature-are inherently resistant to some antimicrobials. For example, all anaerobic bacteria are resistant to the aminoglycosides, since the antibiotic cannot reach its target molecule (30S subunit of the ribosome). The uptake of an aminoglycoside drug by a bacterial cell is a two-step process, requiring the presence of oxygen-or nitrogen-dependent electron transport chains, a mechanism that all anaerobes lack, therefore making the impact of this antibiotic class limited. Anaerobes exhibit similar innate resistance towards fosfomycin, trimethoprim, aztreonam and all 1st and 2nd generation quinolones through a variety of processes, therefore using these drugs in therapy would be ill-advised [156][157][158]. Other instances of intrinsic resistance are species dependent, like in the case of metronidazole resistance of various Gram-positive anaerobes (Actinomyces spp., Lactobacillus spp., Bifidobacterium spp., Propionibacterium spp.), the macrolide and rifampin resistance in Fusobacterium nucleatum and F. mortiferum and the resistance against cephalosporins in C. difficile.
Clindamycin was considered the gold standard for the treatment of anaerobic infections some 40-50 years ago but with the emergence of high levels of resistance among C. difficile (~70%), B. fragilis group (30-40%), Prevotella spp. (10-40%), other related anaerobic Gram-negative bacteria (~10%) and Peptostreptococcus spp (~10%), this drug lost its significance as a first-line drug. Clindamycin resistance of chromosomal origin is linked to tetracycline resistance determinants, although their presence in plasmids and conjugative transposons was also described. Several studies showed that macrolides (azithromycin, clarithromycin) are potent agents against some anaerobic bacteria, although their bacteriostatic effects hinder their usefulness in serious infections. Resistance to the macrolide-lincosamide-streptogramine-group of antimicrobials manifest uniformly (which can be attributed to various erm genes): due to the methylation of two specific adenine residues of the 23S rRNA, the effective binding of the drugs to ribosome is prevented [159][160][161][162][163].
Chloramphenicol was a drug-of-choice for the treatment of serious anaerobic infections (especially if the central nervous system was involved), nowadays is an infrequently used drug, due to its considerable toxicity and serious reversible (haemolytic anaemia, optic neuritis, bone marrow suppression) and irreversible (fatal aplastic anaemia) side effects. Resistance against chloramphenicol is mostly plasmid-mediated, by inactivation of the active drug through acetylation or nitro-reduction [164,165].
Quinolones were generally considered ineffective against anaerobic bacteria due to their bacteriostatic effects and the poor drug penetration to the target sites, together with the low affinity for the target enzymes, owing to point mutations in the gyrA-B (Topoisomerase II) and parC genes (Topoisomerase IV). Additionally, quionolone resistance determining regions (QRDR; like that found in E. coli) were observed in several anaerobic species, with an increased expression of efflux proteins. These quinolone resistance mechanisms were mostly observed in B. fragilis group, C. perfringens and C. difficile isolates [166][167][168][169]. Moxifloxacin is a new drug in this group of antimicrobials, with FDA approval for complicated skin and skin structure infections, with promising anti-anaerobe activity.
Resistance against β-lactam antibiotics is very common among anaerobes (predominantly due to the production of β-lactamase enzymes), although there are significant variations to the levels of resistance between these bacteria. Almost 100% of B. fragilis isolates are resistant to penicillin G, since they constitutively produce a chromosomally-encoded penicillinase enzyme; this is around 50% in the case of Prevotella spp. and between 8-17% for Porphyromonas spp. While Cl. perfringens isolates are~100% susceptible to penicillin G (ampicillin), resistance has emerged in non-perfringens clostridia (e.g., Cl. clostidiforme, Cl. butyricum, Cl. ramosum) [170]. The production of a class 2e cephalosporinase (encoded by the cepA and cfxA genes; B. fragilis) confers resistance against cefoxitin and cefotetan; this enzyme is inhibited by β-lactamase enzyme inhibitors (clavulanic acid, sulbactam, tazobactam). Cephamycins such as cefoxitin (with a 7α-methoxyl side chain) and cefotetan (with an N-methylthiotetrazole side chain) can still be considered appropriate therapy for anaerobic infections (in contrast to cephalosporins) but consulting regional epidemiological data about resistance levels is recommended. An emerging issue is the appearance of a Zn 2+ -metallo-β-lactamase enzyme (essentially a carbapenemase, encoded by the cfiA/ccrA genes; it is inhibited by EDTA) in B. fragilis. These isolates are resistant against the carbapenem antibiotics, which are usually preserved as last-resort drugs in serious, life threatening infections. It has been described that more Bacteroides isolates harbour these resistance genes than the number of isolates with a phenotypic resistance. The probable cause of this phenomenon is that these genes are not expressed in levels ("silent genes"), where they could exhibit decreased susceptibility carbapenems. However, with a one-step mutation, or with the insertion of an IS element, the strain can easily mutate to become resistant. Reduced permeability of the drugs through the outer membrane in some Fusobacterium spp., Porphyromonas spp. and Bacteroides spp. can also lead to decreased susceptibility. PBP alterations (leading to reduced binding affinity of these drugs) were described in several Gram-positive anaerobic cocci. It can be said that, for now, β-lactam antibiotics (especially beta-lactam/beta-lactamase inhibitor combinations and carbapenems) still represent the frontline of treatment in mixed aerobic-anaerobic infections [23,93]. Penicillin G is still a useful drug in our disposal and its use is appropriate if the strain in question is susceptible. Beta-lactam-beta-lactamase inhibitor combinations (with resistance levels of 0.5-10%) and carbapenems (0-2%) should be considered safe choices for empiric therapy [151,[171][172][173][174][175][176][177][178][179][180][181]. The importance of surveillance is extremely important related to this group of drugs, with the number of resistant strains steadily growing and with the emergence of multidrug resistant anaerobes (note: some carbapenem and metronidazole resistance elements share mobile genetic elements) [79,149]. Such a pathogen of increasing importance is Sutterella wadsworthiensis, with increasing resistance rates to piperacillin/tazobactam, cefoxitin, clindamycin and metronidazole [182,183].
Metronidazole resistance is common in various Gram-positive anaerobic rods (Actinomyces spp., Propionibacterium spp., Lactobacillus spp.), while the prevalence of resistant Gram-positive cocci and Gram-negatives is usually very low (<1%). Metronidazole (and other related drugs) are a part of the 5-nitroimidazole group drugs and are one of the most important agents for the treatment of anaerobic infections. Metronidazole is a pro-drug: for the desired antimicrobial activity, it must be reduced for the release of toxic nitroso-residues (PFOR; pyruvate: ferredoxin oxidoreductase and the nitroreductase enzymes) that are extremely reactive (able to damage DNA, proteins and membranes of bacteria). Some consider metronidazole an example for the "ideal antibiotic model" [184]. Resistance to this drug is conferred by the nim genes (nimA-J) coding for a nitroimidazole-reductase: this transforms the pro-drug to a non-toxic amino-imidazole, with no antimicrobial activity. These nim genes share around 70% homology (B. fragilis group), apart from nimI (Prevotella spp.). nimB has been associated with moderate to high level metronidazole resistance. Other mechanisms of reduced susceptibility include lower levels of the PFOR enzyme (a significant decrease of the reducing power of the microorganisms) and the induction of lactate dehydrogenase (LDH) activity. These resistance determinants are predominantly found in the chromosome but their presence has been described in transferable plasmids and IS elements as well, with some experts fearing the rapid dissemination of these genes [149,150,[185][186][187][188][189][190][191][192][193][194].
Tetracycline resistance is very common in anaerobic isolates (>95% of Bacteroides spp., >50% of Prevotella spp., Fusobacterium spp. and Clostridium spp.) to the point that the use of these drugs is not advised. Among the resistance mechanisms, drug efflux (tetA-E, tetK-L), ribosomal protection (tetM, tetQ, tetW) and oxidation (tetX) can be found. The tet genes are induced by sub-inhibitory tetracycine exposure, stimulating the spread of resistance trough conjugative transposons [160,170,[195][196][197]. Tigecycline (the tert-butylglycamido-derivative of minocycline) is a novel drug in this antibiotic class, having FDA approval for soft tissue and intra-abdominal infections.
Efflux pumps are important determinants of antimicrobial resistance both in aerobic and anaerobic bacteria, their over-expression can often be associated with therapeutic failure. These transmembrane proteins can bind and expel various substrates and xenobiotics (such as antibiotics) before they could reach their target site or protein. Several efflux pumps have been characterized among anaerobic bacteria: the BexA of B. thetaiotaomicron (a transporter of the multi-antimicrobial extrusion protein [MATE] superfamily) the bmeABC1-16 (members of the resistance-nodulation-cell division [RND] superfamily, analogous to the AcrAB-TolC tripartite efflux systems of E. coli) and TetA-E efflux transporters (members of the major facilitator superfamily [MFS] superfamily) of B. fragilis, bcrABD (an ATP-binding cassette [ABC] transporter) of Cl. perfringens and XepCAB (member of the RND superfamily) of P. gingivalis [198][199][200][201]. Experimental studies have shown that knocking out one or more of these efflux proteins has a remarkable effect on the MICs of several antibiotics, although due to the redundancy of these proteins in the bacterial genome, other transporter mechanisms usually compensate for the loss of the selected transporter. An emerging therapeutic strategy is the use of efflux pump inhibitors (EPI) as adjuvant compounds, although these compounds were only tested extensively in laboratory conditions so far [198,202].
(Since this is not the main topic of our review, the information on the various modes of antibiotic resistance among anaerobes presented above is a brief summary of the available literature. For further reading on this topic, see [11,23,93,200,203]). Table 8. Examples for antimicrobial resistance mechanisms exhibited by anaerobes [11,23,93,200,203].

Conclusions
Anaerobic microorganisms are now widely accepted as significant pathogens in human diseases, as such, the proper diagnosis and treatment of these infections are important healthcare priorities. The emergence of antimicrobial resistance in anaerobic bacteria is a discernible phenomenon, which deserved the attention of people working in diagnostic microbiology and infectious disease treatment. Microbiology laboratories should perform antimicrobial susceptibility testing to the best of their abilities, taking into account the number of incoming anaerobic isolates, as well as monetary considerations of the given institution. This incentive is important to broaden the information on regional, national and global patterns of resistance of anaerobes. Based on the plethora of literature available, the slow but steady increase of metronidazole and carbapenem resistance isolates is a cause for concern, especially if we observe what happened to the efficiency of other agents (like clindamycin) in the previous 20-30 years.
Author Contributions: M.G. did the literature review and formulated the initial draft of the manuscript. M.G., G.S. and E.U. discussed the initial draft and produced together the final version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.