1.1. Taxonomy and the general features
of the Mollicutes

Mycoplasmas belong to the class Mollicutes (meaning soft skinned). Mollicute or mycoplasma is often used in a trivial way to refer to any species in the class. From a human perspective, the order Mycoplasmatales with a single family Mycoplasmataceae , is the most important. Mycoplasmataceae contains two genera, namely Mycoplasma and Ureaplasma . The genus Mycoplasma contains more than 100 species, of which 13 at present are considered a part of the human flora. Two Ureaplasma species U. urealyticum and U. parvum (previously known as U. urealyticum , parvo biovar) are commonly found in the human urogenital tract.

For many years, mycoplasmas were believed to be viruses, because they could pass through filters considered to block the passage of bacteria, i.e. 0.45μ or even 0.22μ pore size. In the 1930s, the concept of viruses became clearer, and mycoplasmas were now believed to be bacteria. During the fifties, some mycoplasma species were grouped with the bacterial L-forms, which are classical bacteria deficient in cell-wall formation. Finally, in the early sixties, the issue was settled, and the mycoplasmas found their current place in the taxonomic system. The filterability and lack of formation of a cell wall after incubation in medium without antibiotics, have now become one of the key properties defining new species as mycoplasmas (17; 18).

The mycoplasmas are the smallest prokaryotes capable of self-replication. They are believed to have evolved from Gram-positive bacteria most closely related to the clostridia by degenerative evolution leading to genomic reduction (19). Mycoplasmas, particularly species with the smallest genomes, have high mutation rates, suggesting that they are in a state of rapid evolution (20). M. genitalium is the species with the smallest genome of all Mollicutes studied thus far with a genome of only 580 kb.

In 1995 the Haemophilus infuenzae genome was published as the first complete genome sequence from a free-living organism (21). Later the same year, the M. genitalium genome sequence was published (22), making it the second bacterium to have its genome fully sequenced. After the larger 816 kb. M. pneumoniae genome was fully sequenced (23), it was shown that M. genitalium contains a subset of the M. pneumoniae genome complement (24), with all coding M. genitalium genes having their counterpart in the larger M. pneumoniae genome. Based on comparison of the M. genitalium 16S rRNA gene sequence, M. genitalium belongs to the M. pneumoniae cluster ( Figure 1 ) with the closest relatives being M. pneumoniae, M. alvi, M. gallisepticum, M. imitans, M. pirum, and M. testudinis . A newly isolated glucose fermenting mycoplasma species from the human respiratory tract (25) also belongs to this cluster. This species is probably going to be named M. amphoriforme .

Most mycoplasma genomes have a low guanine plus cytosine (G+C) content, within the range of 24-33% G+C. The M. genitalium genome has a comparatively high G+C content of 32% (22), although some of its genes have a significantly higher G+C content i.e. the rRNA genes with a G+C content of 44%.

Mycoplasmas are difficult to study by classical genetic tools, both because of their fastidious growth requirements, and as a consequence of the absence of selectable markers. Furthermore, in contrast to other bacteria but similar to mitochondria, Mycoplasma use the UGA codon to code for tryptophan instead of the common STOP signal. This means that expression of mycoplasma genes in common bacterial systems is complicated due to the synthesis of truncated proteins. M. genitalium is a motile mycoplasma species (26), and like most of the other motile species, M. genitalium is flask shaped and uses its terminal tip-like structure to attach to surfaces and provide a gliding motility ( Figure 2 ). The speed averages about 0.1 μ/s which is slower than that recorded for M. pneumoniae but faster than that of the avian species M. gallisepticum (26; 27). Attachment is mediated mostly by the main adhesin protein MgPa, the features of which is described in chapter 2.

The adhesin proteins may also have important functions during cell division. In M. pneumoniae , the P1 adhesin protein (the counterpart of the MgPa protein in M. genitalium ) is located at one cell pole in all adhesive cells. By immunostaining and DNA content determination, it has been shown, that cells with a single P1 focus at one cell pole had a lower DNA content than cells with two foci. Those with one focus at each cell pole had the highest DNA content, suggesting that the nascent attachment organelle is formed next to the old one and migrates to the opposite cell pole before cell division (26).

2.1. Adhesion

Adhesion of Mollicutes to host cells is a prerequisite for colonisation and subsequent infection. The lack of a cell wall and cell wall associated structures such as fimbriae often responsible for adhesion of other bacteria indicates that this process is mediated by membrane bound components, which have been termed adhesins. Adhesins have been intensely studied both in M. pneumoniae and in M. genitalium . Both species bind erythrocytes from a variety of species (38) with their adhesin molecules clustered at the tip structure of the polar cell (39). Both species also attach to eucaryotic cells such as Vero cells (28), as shown already in the initial description of the species (10), but more importantly, M. genitalium attaches to the epithelial cells of human fallopian tube (37) ( Figure 3 )

A variety of techniques have been applied in the identification of the adhesins. SDS PAGE of mutant strains incapable of haemadsoption has been most widely used in the characterisation of proteins involved in adhesion (40). Monoclonal or polyclonal, monospecific antibodies to missing proteins have then been used to document the importance by blocking the adhesion (41-43). In addition, the similarity between M. pneumoniae and M. genitalium has been used to look for cross-reactive antibodies or cross-hybridising DNA probes from already identified M. pneumoniae adhesins (44). Based on characteristic lipoprotein-specific features, 21 putative lipoprotein genes have been identified in the M. genitalium genome (24), but only a minority have been characterised as adhesins (22). Recently, adhesion has been studied by transposon mutagenesis using Tn 4001 to produce knock-out mutants deficient in adhesion (45). Using this method, some of the accessory proteins involved in clustering of the adhesins have been characterised, but the disadvantage of this approach is the non-random insertion of the transposon in specific hot spots in the genome. Therefore, the development of methods for targeted disruption of genes based on homologous recombination, which has been demonstrated for M. genitalium gene MG218 appears to be an important tool in the further investigations (46).

The main adhesin P1 of M. pneumoniae closely resemble the M. genitalium counterpart MgPa (ORF-2 or mgpB of the MgPa operon). The genes for both adhesins are organised in operons with three genes (47; 48). The M. genitalium MgPa operon consists of ORF-1 (29 kDa protein), ORF-2 (MgPa), and ORF-3 (MG192; 114 kDa protein). The function of the ORF-1 gene and the equivalent ORF-4 in M. pneumoniae is not known. The protein from the M. pneumoniae ORF-4 gene has been expressed in E.coli and used to immunise mice, but the antibodies were unable to recognise any antigen in immunoblots of M. pneumoniae proteins (49). In M. pneumoniae , the ORF-6 gene product appears to be cleaved in a 40- and a 90 kDa component (49). In contrast, the M. genitalium ORF-3 gene is present in full length in the protein profiles (50). The MgPa and the ORF-3 gene product are immunogenic both in immunised animals and in humans (50) and both have repeated sequences in the chromosome.

Another adherence component of the tip is designated P30 in M. pneumoniae (51) and P32 in M. genitalium (44). The genes encoding P30 and P32 are located in operons located distant from the MgPa and P1 operons, and are expressed together with adherence accessory proteins designated HMW3 (for high molecular weight) and P69, respectively (44). These accessory proteins, together with the other HMW proteins in M. pneumoniae and their analogues in M. genitalium , are important for clustering of the adhesin at the tip, and for maintaining the shape of the cell, acting like a cytoskeleton (39).

An additional membrane associated and surface exposed protein in M. pneumoniae has been designated P116, and is highly immunogenic. P116 may also be adherence associated, as antibodies directed against it abolish adhesion (43). This gene has its counterpart in MG075 (22), but protein expressed from this gene has not been detected by 2D PAGE (52), or among immunoreactive M. genitalium proteins identified by mass spectroscopy (50). Interestingly, the MG075 gene appears to be variable among strains as demonstrated by amplified-fragment length polymorphism fingerprinting (AFLP) (53). The variability has been shown widespread and not only limited to the restriction enzyme recognition site (Branko Kokotovic, personal communication, 2003).

The receptor for M. pneumoniae adhesion is believed to be long-chain sialo-oligosaccharides on the host cell (54), but other structures such as a glycoprotein isolated from human lung fibroblasts by Geary et al . (55; 56) may also be involved, since neuraminidase treatment does not completely abolish adherence. The same purified fibroblast glycoprotein was also capable of blocking the adhesion of M. genitalium to MRC-5 fibroblasts (56). Recently, surfactant protein D, considered being a part of the innate immune system and present in the lung alveoli, was shown to bind M. pneumoniae , presumably by a non-protein interaction mechanism (57). Thus, it appears that M. pneumoniae lipids may act as adhesins and that it is able to use the host's defence mechanisms for pathogenesis. However, whether M. genitalium uses similar receptors in the urogenital tract is still an open question. Recently, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a key component of the glycolytic pathway, was shown to be one of the ligands involved in binding M. genitalium to human vaginal and cervical mucin (58). GAPDH exhibits its main function in the cytoplasm as a housekeeping component. Finding that it may also serve as an adhesin may be yet another way to rationalise the limited genomic information available in M. genitalium .

In conclusion, the presence of several adhesins in M. genitalium is well documented, and information about their interaction with host cells is accumulating. However, the interaction with host cell receptors and the relative importance of the different types of adhesin/receptor interactions remains to be determined.

2.2. Evasion of the host immune response by antigenic variation

Mycoplasmas are dependant on an intimate contact with the host cells. In order to evade the immune response of the host, and to adapt to changing environments, they have a need for phenotypic plasticity, such as variation in the antigenic composition of the surface components. Other mechanisms, such as mimicry of host antigens, and survival within professional phagocytes, may help the parasite to survive (reviewed in (59)). The antigenic variation can be brought about by two basically different mechanisms. Either, the pathogen may regulate the expression of virulence factors in response to changes in the environment by signal transduction pathways, or the microbial population as a whole may spontaneously and randomly generate distinct new phenotypes that will survive the host response (59). Since mycoplasmas have few regulatory genes that could serve as sensors to environmental stimuli, and a few genes encoding transcriptional factors, the most important means for phenotypic plasticity in mycoplasmas is antigenic variation caused by molecular switching events (36).

Surface exposed membrane proteins are the major antigenic determinants of mollicutes, and a variety of mechanisms have evolved to generate high frequency intragenomic changes in nucleotide sequence or DNA conformation at selected chromosomal loci (60; 61). The functional consequence is, that these alterations in genotype can rapidly promote phenotypic heterogeneity, even in small, clonal populations of bacteria, such as the limiting inoculum that initiates an infection (62). The phenotypic changes do not occur more often when they are useful than when they are not (63). The capability to generate genetic variation has apparently evolved at particular loci to promote random phenotypic variation as a result of a constantly changing host environment. Thereby, the adaptive potential has been maximised without compromising household functions. Interestingly, host organisms generate random diversity in B lymphocytes by similar mechanisms (64).

Multiple copies of partial gene sequences have been reported for many pathogenic bacteria. Considering the very limited genetic information that mycoplasmas contain, the number of mycoplasmal genes involved in diversifying the surface exposed antigens is remarkably high (60).

2.2.1. Sequence variation of the MgPa gene:

In M. genitalium , nine repetitive elements, which are composites of non-contiguous regions of the MgPa operon, have been identified (3; 22). The sequence identity between the repeats and the operon sequences ranges from 78-90% and together with the sequence of the MgPa operon, they represent 4.7% of the total genomic sequence (22). Considering the evolutionary pressure towards a minimal genome (65), it would seem reasonable to expect such repeats to have an important function. The repeated MgPa sequences were identified by probing M. genitalium chromosomal DNA with a probe prepared from fragments of the MgPa operon (66). It was shown, that approximately half of the gene was present as multiple copies. The restriction fragments used for probe preparation were designated A-J ( Figure 4 ), and although this division did not correlate well with boundaries between repeated and single copy parts of the MgPa gene, this nomenclature has been used and expanded in subsequent work (3).

During the development of the MgPa based PCR (1), sequence variability in the amplified sequence was evident when clinical specimens were examined in contrast to the sequence conservation observed when the cultured stains available at that time was examined. When PCR amplified fragments of the MgPa gene from the four new Danish isolates (4) were analysed, it was found, that they were different from each other and different from the previously isolated strains.

At the IOM meeting in 1994, Camella Bailey presented the initial analysis of some of the MgPa related genomic repeat sequences and at the same meeting, our preliminary data showing the variability of the MgPa genes from the four Danish M. genitalium strains was presented. By a joint effort, we were able to show, that the B-region of the MgPa gene had five genomic repeats (later seven repeats were detected) and that the same region in the MgPa gene of the Danish strains was hypervariable (3). Furthermore, a difference in the B-region between the early passage strain of M. genitalium used in our laboratory (designated G37-DK), compared to the G37-US strain identical to the strain subsequently deposited in the ATCC was found. However, a sequence matching exactly the variant sequence in G37-DK was found in one of the repeats of the G37-US strain, indicating that the repeat sequences provide a source for sequence variation in the operon (3).

After the genome sequence of M. genitalium was published (22), the general structure of the genomic repeats was shown to comprise sequence similar to the MgPa operon segments B, E and F, followed by a stretch with an unusual high A+T percentage (around 80%). Then, a rather long stretch of the ORF-3 gene K and L regions, was in some cases interrupted by a G-region sequence from the MgPa gene, followed again by an ORF-3 sequence from the border between the L and M regions ( Figure 5 ) (67).

The complete MgPa sequence and part of the ORF-3 sequence of the two Danish strains M2288 and M2300 were determined, and aligned with the sequence of the G37 (US) strain (67). When the number of base differences per 30 bases was calculated and related to the sequence position, it was shown ( Figure 6 ), that relatively conserved areas were interrupted by hypervariable regions. Most often, these regions were confined to areas where MgPa genome repeats were present. However, in a few positions, spikes with a high variability were detected. By inspection of the sequence data, it was found, that these hypervariable regions were caused by repeats of AGT base triplets coding for serine. Three of these serine repeats were found, two in the MgPa gene, and one in the ORF-3 gene with repeat lengths varying from strain to strain from five to 11 serine codons. The function of the serine repeats is not clear, but it could be speculated, that they are important in the three-dimensional structure of the molecule, functioning as spacers between the adherence mediating epitopes (67). Interestingly, Opitz and Jacobs (42) found, that most of the adherence mediating epitopes of M. genitalium G37 are found in the hypervariable regions, primarily in the B and EF regions. They used overlapping octapeptides covering the MgPa protein to map the epitopes of adherence inhibiting monoclonal antibodies. In contrast, out of three polyclonal monospecific rabbit antisera produced by immunisation with recombinant MgPa fragments generated from the N- terminal, middle, and C-terminal parts of MgPa, only the antibody against the C-terminal part was able to block adhesion (43). The reason for this discrepancy is not clear, but obviously, the experimental set-ups are different, with the latter system being closer to the in vivo situation. The observation, that the C-terminal part is more immunogenic is also substantiated by the observation that human antibodies, primarily recognise the C-terminal fragment when the same fragments were reacted with human sera (50).

Ongoing work indicates that antigenic variation occurs also in vivo . A different pattern in restriction enzyme analysis (REA) of the MgPa gene was found between two isolates obtained from a patient with recurrent non-chlamydial non-gonococcal urethritis (NCNGU) with an 11-week interval. A different REA type was demonstrated in the 3' end of the MgPa gene (the G-region) and in the ORF-3 gene (the KLM region) (67). Furthermore consecutive isolates of M. genitalium obtained from male chimpanzees inoculated intraurethrally with M. genitalium during the early animal experiments (68) accumulated an increasing amount of mutations after five weeks post inoculation (Chimp A52 used as an example in Figure 7 ). However, all mutations were located within the hypervariable regions, and all the mutations could be accounted for by the sequences present in the MgPa-reps (69). The conserved sequence found during the first five weeks of infection correlates closely with the absence of antibodies in the chimpanzees during the first five weeks of infection (68). This finding strongly supports the notion, that the selection of M. genitalium strains expressing different epitopes within the adherence mediating domains is driven by selective pressure from the immune system (69).

2.2.2. Other variable antigens

As demonstrated by AFLP fingerprinting, the MG075 gene appears to be variable among strains (53). However, since the gene product is unknown, the effect of the variation is not known.

Recently, Ma and Martin (70) amplified a fragment of the putative lipoprotein gene MG309 from 42 specimens from 31 patients and demonstrated that this gene had a variable number of tandem repeats. They found 11 different types of these repeats varying in number from 7 to 17. In five patients, a mixture of 2 or 3 types were detected, suggesting that the M. genitalium population was changing from one type to another. Lipoproteins are usually membrane associated, but the specific function and location of the MG309 gene product is not known. The gene is expressed, however, as shown by reverse transcriptase PCR of broth grown M. genitalium cells (David H Martin, personal communication).

In conclusion, antigenic variation occurs in vivo . This has been shown to take place in the MgPa operon as demonstrated by the recombination between the MgPa-reps and the operon. Recombination with the MgPa-reps may generate almost unlimited variation in these epitopes allowing unique opportunities to evade the immune response and to adapt to various cell surfaces. Other variable antigens are currently under study and may prove to offer additional variation.

2.3. Intracellular location of M. genitalium .

From the early days of mycoplasmology, surface versus intracellular location has been disputed. During the 70's and 80's most evidence favoured an extracellular location. The studies of Lo et al . on the virus-like infectious agent (71), which was later shown to be a strain of M. fermentans (72) demonstrated quite convincing, that structures resembling mycoplasmas could be observed intracellularly by EM. Later, the same group discovered a new mycoplasma species isolated from patients with AIDS (73) which appeared to be capable of invading eucaryotic cells by means of a specialised arrow-shaped tip structure justifying its naming as M. penetrans (74). Most importantly, this phenomenon was also demonstrated in urothelial cells from infected patients, providing evidence that it occurred also in vivo (75).

Documenting intracellular persistence of mycoplasmas could explain the severe difficulties experienced in eradicating them with antibiotics, which has been an enigma for years. Both in cell cultures and in humans, mycoplasmas may be isolated after antibiotic treatment with relevant doses of antibiotics to which the strain is highly susceptible in vitro (76).

Using electron microscopy (EM), the morphology of the Danish M. genitalium strain M2300 was studied during its isolation by co-culture with Vero cells (28). Like M. penetrans , it appears that M. genitalium enters the cell through its specialised tip structure. Intracellularly, they were located in membrane bound vacuoles, often near the nucleus ( Figure 8 ). Only 10% of the Vero cells revealed invasion of mycoplasmas; whether this reflects the age of the cell or just an uneven distribution within the cell causing the section to appear empty is not clear (28). EM studies are difficult to interpret because two-dimensional sections of a three dimensional structure is studied. It could be argued, that what was interpreted as mycoplasma cells in vacuoles were actually the bottoms of crypts formed by invagination of the membrane. This issue could be addressed by staining the cells with ruthenium red, which stains the mucopolysaccharide surface coats of the cells. Taylor-Robinson et al . successfully used this approach to demonstrate, that the intracellular vacuoles with M. hominis and M. fermentans seen in their studies did not communicate with the surface of the cells (77; 78). However, in the M. genitalium study, this approach was not successful, possibly because the Vero cells were grown in a medium without foetal calf serum (28).

Mernaugh et al . used the G37 type-strain and human lung fibroblasts to study the kinetics and molecular mechanism involved in invasion. The adhesion to the cell surfaces occurred almost instantaneously, as the cells were covered with mycoplasmas already at "zero" time. After one hour, mycoplasma cells were seen invaginating the cell membrane with a morphology similar to clathrin-coated pits (40). Clathrin-coated pits are important in the signal modulation in cells being one of the main pathways for inactivation of membrane receptors by endocytocis (79). It has also been suggested that C. trachomatis enters the cell by means of these structures (80). After 12 hours, a significant number of M. genitalium cells were observed in vacuoles, and at 96 hours, lysis of the infected cells was almost complete (40).

Baseman et al . labelled M. genitalium , M. pneumoniae , and M. penetrans with DiIC 18 , a fluorescently labelled cationic lipophilic tracer that can be used to label the mycoplasma membrane yet retaining mycoplasma viability (81). Labelled mycoplasmas were grown with cell cultures, and by laser confocal microscopy, thin sections of the cells could be visualised, showing clusters of fluorescently stained mycoplasmas without contact with the eucaryotic cell membrane. Another interesting finding was the existence of M. pneumoniae mutants capable of cytadherence, but incapable of invasion. Unfortunately, these mutants have not been further characterized. Subsequent work by the same group showed, that M. genitalium , M. pneumoniae , and M. penetrans were able to persist intracellularly for extended periods of time (82). Intracellular survival was documented for M. pneumoniae and M. genitalium by culturing HEp-2 cells in a medium containing a high concentration of gentamycin for up to six months (82). Since both mycoplasma species are susceptible to gentamycin, and since gentamycin does not cross the HEp-2 cell membrane, only those mycoplasmas persisting inside the cells would survive. Survival and even multiplication of M. pneumoniae was documented by weekly cultures of fractionated HEp-2 cells showing an increasing ratio of mycoplasma cfu per cell. For M. pneumoniae the increase was more pronounced in the cytoplasmic fraction than in the nuclear fraction. Interestingly, M. genitalium could not be subcultured from its intracellular state, but persistence was documented by PCR, which remained positive for all 6 months. Furthermore, DNA replication of both M. pneumoniae and M. genitalium was shown by incubating the cells with medium containing radiolabelled thymidine in the presence of an inhibitor of eucaryotic DNA polymerase. After 48 hours, total DNA extracts were subjected to restriction enzyme digestion with E co RI, and autoradiographed to demonstrate the mycoplasma specific restriction enzyme pattern (82). M. penetrans was also capable of prolonged persistence, but since this species is resistant to gentamycin, the experimental design regarding this species was less convincing. However, an interesting observation was made during attempts to eradicate M. penetrans from the cell culture with ciprofloxacin: Two weeks after the inclusion of the antibiotic to the medium, cultures turned negative, but even though the treatment was maintained for one month, viable M. penetrans cells were recovered after one week without antibiotic. Surprisingly, the mycoplasmas had maintained their ciprofloxacin susceptible phenotype (82).

In conclusion, it appears that M. genitalium behave like a facultative intracellular pathogen, i.e. it is capable of retaining viability both extracellularly and intracellularly in vitro . Whether the same occurs in vivo has not been proven. However, from clinical experience with patients receiving multiple courses of highly active antibiotic treatment with temporary improvement but subsequent relapse, it seems likely that such mechanisms may act in at least in a subset of patients.

3.1. Culture

M. genitalium was first isolated in 1980 from two of 13 men with NGU using the SP4 medium. Evidence of growth occurred extremely slowly; initially more than 50 days were needed for a colour-change to occur (10). The two new strains were designated G-37 and M-30. G-37 was later selected as the type strain of M. genitalium (16).

Even though strains G-37 and M-30 were re-isolated from the original specimens after they had been refrozen and transported back to the UK, later attempts using culture of urogenital specimens have not been very successful. Taylor-Robinson et al . failed to isolate M. genitalium although they reported "presumptive isolation" based on colour-change in the mycoplasma broth from men with NGU (83). However, a colour change in the mycoplasma broth could also be caused by other bacteria, including unknown mycoplasma species, or by human cells, which may grow in the rich SP4 medium.

Samra et al . (84) studied a variety of clinical specimens using SP4 and a modified medium with horse serum. They also observed acidification of the medium in about 15% of the specimens, but were unable to subculture any microbes. Among the 513 specimens studied, 33 were urethral specimens from men with NGU, and 83 were cervical swab specimens from women suspected of an STI. However, no isolates of M. genitalium were obtained. Although not clearly described, the swab specimens appear to have been transported in Stuart's transport medium and inoculated mostly within three hours. Although the viability of M. genitalium in Stuart's medium was found to be five days, this contrasts with experience from our lab. We observed a poor recovery of M. pneumoniae from clinical specimens transported in this medium (unpublished observation), and ascribed it to the presence of methylene blue, which inhibits the growth of M. pneumoniae and some other species. Whether M. genitalium is sensitive to growth inhibition by methylene blue is at present not clear.

In 1988, Joel Baseman reported on the isolation of four strains of M. genitalium found among 16 cultures of M. pneumoniae isolated from recruits during a vaccine trial for an M. pneumoniae vaccine (33). This publication made mycoplasmologists rethink the dogma that M. genitalium was restricted to the urogenital tract. At approximately the same time, and in the same laboratory, an isolate of M. pneumoniae isolated from the synovial fluid of a patient with pneumonia and polyarthritis was investigated actively for the presence of a mixed infection with M. genitalium by epi-immunofluorescence on agar-grown colonies (85; 86). Approximately 1-2% of the colonies were found to react more strongly with the M. genitalium polyclonal antibody than with the polyclonal antibody against M. pneumoniae . Subsequently, the M. genitalium strain was recovered by selection of colonies within the inhibition zone of M. pneumoniae in a disc growth inhibition test (a principle closely resembling selection of resistant colonies from the zone around an antibiotic containing disc) (85; 86). Thus, five M. genitalium strains have been recovered from what was believed to be cultures of M. pneumoniae by the same group of persons within a short period of time. Since then, a lot of work has been spent on the importance of M. genitalium in the respiratory tract, but no other reports about respiratory tract isolates have been reported.

Eight strains of M. genitalium were reportedly isolated by culture in SP4 medium from urogenital tract specimens from STD clinic patients in China (87). Unfortunately, these strains have not been made available outside of China despite several requests.

One of the aims of the current thesis was to develop diagnostic tools to study the role of M. genitalium in human disease, and since culture is the mainstay of infectious disease diagnostics, an improved culture method was developed (4). Like many other laboratories with interest in human mycoplasma infections, attempts to isolate M. genitalium from urogenital specimens had also been made at SSI, however, without much success. During the first clinical study (2), specimens were collected in SP4 medium with the specific purpose of culture. However, despite incubation for 12-14 weeks and the use of three different medium formulations with quality-controlled ingredients, no isolates were obtained. Since it had been documented by PCR, that M. genitalium was present in some of the specimens, the development of a culture method was facilitated. Furthermore, growth of the bacterium could be followed by subjecting dilutions of the growth medium to PCR. Considering the susceptibility of cultured cells to mycoplasma infection with a variety of reported species, and also with the knowledge of propagation of fastidious strains of M. hyorhinis in cell cultures (88), attempts were made to grow M. genitalium in cell cultures (4). This approach proved to be very efficient but extremely time-consuming. Four new strains were recovered and cloned to ensure the purity of the strains. For most strains, at least six months were required from the cell-culture was inoculated, until the strain was ready for cloning (4). Furthermore, some strains did not grow on solid media after filtration and therefore, they were difficult to purify. Since the publication of the manuscript, three additional strains have been recovered (8); two of them were from the same patient attending an STD clinic in Bordeaux, France at several occasions with a recurrent M. genitalium urethritis. From this patient, three PCR positive specimens were kindly provided by Dr. de Barbeyrac, Bordeaux, and all three specimens grew in the cell culture system, but only two have been adapted to cell-free growth. An early cell-culture passage of one of the strains (M 2282) from the isolation study (4) was later successfully grown in cell-free medium and cloned. In Table 1 , all currently available strains of M. genitalium have been listed.

The Ultroser HY serum substitute was shown to be superior to foetal calf serum, when isolation was attempted in Vero cells (4). This product was discontinued from the distributor, and only recently has it again become available from another producer. However, cell-assisted growth of M. genitalium has also been reported with standard cell-culture medium containing foetal calf serum (George Kenny, personal communication and (89)) although it has not led to new isolates.

Recently, Baseman et al . again reported isolation of M. genitalium (90). Low-income minority women were enrolled in an intervention trial to reduce the recurrence of STDs. M. genitalium was isolated from cervical and vaginal swab specimens from 31 of 838 women using SP4 medium with 0.25 mg/ml of ciprofloxacin. The latter was stated to inhibit growth of contaminating flora, in particular M. hominis . However, since the MIC 50 for M. hominis is 0.5 mg/ml (91), more than half of the strains would not be inhibited at all. The effect on wild-type M. genitalium strains is unpredictable as the MIC ranges from 0.25-10 mg/ml (92), meaning that at least some strains would be inhibited. Furthermore, the specimens were passed through 0.45μ filters known to reduce the viable count > 10-fold. Evidence of growth of M. genitalium was observed already from 14 days and up to 20 days. This is a very rapid growth compared to the > 50 days required for isolation of the original G37 strain. It was notable, that all isolations occurred over a period of 6 months. It is not described how many specimens were positive for each of the 31 women, however, if the same proportion of the 838 women were examined each month during the 22 month study period, 229 women had been examined in the 6-months successful period, resulting in a positivity rate of 13,5%. Another surprising finding was that only 12 (39%) of the 31 culture positive women were PCR positive at the time of culture. It should be noted, however, that the DNA extraction method applied on the clinical specimens included a washing step where the pellet from the centrifuged specimens was resuspended in sterile water and incubated on ice for 20 min. before a repeat centrifugation step. Bearing in mind that M. genitalium is a wall-less prokaryote, the washing step is a very efficient method of inducing osmotic lysis leading to false negative PCR results.

3.2. Serology

The cross-reactions between M. pneumoniae and M. genitalium have significantly hampered the use of specific serology for diagnosis and epidemiological studies. Over the years, several techniques have been applied, but in most studies, the diagnostic performance has not been validated.

Already in 1982, Klaus Lind demonstrated the significant serological cross-reactions between M. pneumoniae and M. genitalium (93) in both complement fixation tests, metabolism inhibition, indirect haemagglutination, and disc growth inhibition. The cross-reactions were found both with antisera from rabbits immunised with M. pneumoniae and M. genitalium , and with human sera from patients with M. pneumoniae infection. Only the metabolism inhibition test with sera from patients with M. pneumoniae infection appeared to discriminate between infections with the two pathogens in the three sera examined. In a more extensive study, where also microimmunofluorescence, crossed immunoelectrophoresis and heamadsorption inhibition tests were included, these findings were to some extent confirmed (34). Interestingly, cross-reactions with rabbit antisera have been reported by others to be less pronounced (94), particularly in the metabolism inhibition test. This test, involving the growth of M. genitalium in the presence of human or animal serum, however, is not very practical for large-scale studies and has only been used in a few human studies. Tully et al . (95) studied sera from HIV-positive and HIV-negative blood donors. They found that the M. genitalium titre was significantly higher for patients with AIDS than for HIV-negative blood donors, but the importance of this finding is not clear (95).

Taylor-Robinson et al . observed a four-fold or greater rise in antibodies to M. genitalium in men with NGU using a microimmunofluorescence assay but the correlation to presumptive isolations was poor (96).

As a part of a prevalence study (2), acute phase serum specimens were collected, and IgG, IgM, and IgA were measured in an ELISA. A method for antigen preparation, which had previously been shown to detect clinically relevant serological responses to M. pneumoniae (97) as well as immunoblotting methods were used, however, no correlation between the PCR status and the antibody response could be shown. It should be noted, however, that only acute phase sera were investigated, and since it has been shown in a chimpanzee model (68; 98), that antibodies are slow to develop after urogenital inoculation, these findings should be interpreted with caution.

Wang et al . (99) studied the seroprevalence of M. genitalium antibodies in different populations with Triton X-114 extracted Lipid-Associated Membrane Proteins (LAMP). The assay was validated with serum and urine specimens from 104 patients; 40 (38%) were LAMP ELISA positive. 15 (38%) of the LAMP positive patients were urine PCR positive for M. genitalium , and all PCR positive patients were detected. The LAMP positive, PCR negative could have been caused by previous exposure to M. genitalium , cross-reactions with M. pneumoniae or other species, or lack of sensitivity of the PCR test. The LAMP ELISA thus appears to be one the most promising of the serological assays applied on male patients.

Immunoblotting was also applied in a Danish study (100) of serum from different groups of infertile women, where it was shown, that women with tubal factor infertility reacted with the M. genitalium MgPa, the 114 kDa protein, and a recombinant truncated MgPa fragment more often than women with infertility due to a male factor or unexplained infertility. Even more importantly, this relation was observed also after controlling for the C. trachomatis serological status (100). The recombinant MgPa fragment has later been applied in an ELISA format, which provides a means for rapid screening of a large number of patients (50). However, the significant variability of the MgPa protein should be taken into account, before the test can be used more widely. Among the Danish strains, a significant number of mutations are present in the 5' part of the cloned sequence. If the sequence variations in the MgPa gene (3) are biologically meaningful, it would be expected that they involved exposed and antigenically relevant parts of the expressed protein.

Recently, women enrolled in the above-mentioned study by Baseman et al . (90) were followed with sequential serology over a 22-months follow-up period using a modification of the LAMP ELISA described by Wang et al . (99). Using an ELISA OD cut-off determined by testing serum specimens from pregnant women with a low risk of STDs, 83% of the 29 M. genitalium culture positive women had positive ELISA results at some time during the 22 month observation period. However, the time of seroconversion correlated poorly with the time of isolation. Furthermore, the OD values appeared to be rather low and fluctuated over time, and no results were presented for culture negative high-risk women (90).

3.3. DNA probes and Polymerase Chain Reaction

Before the development of PCR methods, now documented to be of tremendous importance in studies on the role of M. genitalium in human infections, a few attempts were made using DNA probes. The design of radiolabelled oligonucleotide probes targeting the 16S rRNA was reported to have a detection limit of approximately 1000 organisms (101) and could be tailored to detect only M. genitalium . Clinical studies using such probes have not been reported, but actually, the sequence information from this study was used in one of the first publications on PCR for M. pneumoniae (102). Risi et al . used a cloned 256 bp. fragment of unknown function, and used it to probe simulated female genital tract specimen and found a detection limit of approximately 10,000 genome copies (103). Results from clinical studies with this probe have not been published. In contrast, a whole-genomic, nick-translated probe with a reported detection limit of 10 4 to 10 5 genome copies was used to study urethral specimens from 203 men (104). No evidence suggesting an important role of M. genitalium in acute NGU was found, but men with recurrent NGU were more often positive. Considering that a high proportion of the patients have a very low DNA load in urogenital specimens (8), it is surprising, that Hooton et al . found an overall rate of 15% M. genitalium positive in their study. Thus, data obtained with the TaqMan quantitative PCR assay on urethral and FVU specimens found that 28% of the positive urethral swab specimens contained less than 10 genome equivalents (geq) per assay corresponding to less than 12,600 geq/swab. Because of the centrifugation step involved in specimen treatment of FVU specimens, approximately 280 geq/ml of urine would correspond to the 10 geq per assay (8). Thus, highly sensitive assays are needed for detection of M. genitalium with a high clinical sensitivity. At present, only nucleic acid amplification tests offers the sensitivity needed and only PCR based assays have been published thus far.

The first PCR based test was published in 1991 (1) and very shortly after, Helen Palmer from David Taylor-Robinson's group published their method (105). Both of the assays were based on the MgPa DNA sequence. At that time, it was believed, that an adhesin gene would be rather conserved, since it has an important role in the pathogenesis of the infection. It was therefore surprising, that the region of MgPa flanked by the MgPa-1 and MgPa-3 primers (1) appeared to be variable. An E co RI site present in the G-37 strain, from which the MgPa sequence was determined, was present also in the M-30 strain and in the four respiratory tract isolates, but in none of the amplicons from PCR positive Danish patients. Other sequence variants were also found by restriction enzyme analysis, and later by sequencing (8). This observation led to further studies into the genetic variability of M. genitalium (3) and to the acknowledgement of the need for alternative PCR assays (2; 6). The assay developed by Palmer et al . was a hemi-nested PCR (105), and usually this type of assay is extremely sensitive, but very prone to contamination due to carry-over. In an assay targeting a variable gene sequence, though, there is also a risk of false negative reactions, if a mutation is present in one of the primer binding sites. In one of the few head to head comparisons of M. genitalium PCR assays, however, a slight modification of the original hemi-nested assay using a new outer forward primer, was compared with our MgPa-1/MgPa-3 assay (106) and the two assays were found to be in perfect agreement. Fortunately, both assays have primers located in rather conserved parts of the MgPa gene.

A modification of the MgPa-1/MgPa-3 assay was published by Totten et al . and used for their studies on male urethritis (89), mucopurulent cervicitis (107), and endometritis (108). Recently, a modification of the assay adapting the detection format to microtiterplate hybridisation and incorporating an internal processing control was published (109). This assay modification facilitated high-throughput detection of M. genitalium and has the advantage of using the same procedures as the manual Amplicor® C. trachomatis PCR assay (Roche)

The MgPa-1/MgPa-3 assay was validated as a confirmatory assay as part of the development of a 16S rRNA gene PCR (6) where it was found to detect M. genitalium in 40 of 41 specimens positive in the 16S assay. In a recent head-to-head comparison of 64 FVU specimens with the MgPa-1/MgPa-3 assay and another 16S assay (110), it was found, that the 16S based assay detected nine positive specimens, and the MgPa-1/MgPa-3 assay detected eight of them. The modified hemi-nested PCR (105; 106) confirmed all nine, but since the original MgPa-1/MgPa-3 assay was changed, using only one tenth of the primer concentration that was recommend in the original publication, the sensitivity may have been impaired. As can be seen in Table 5 , the MgPa-1/MgPa-3 assay together with the Palmer et al . hemi-nested PCR have been used most widely in the clinical studies published thus far. In Table 2 are listed the different PCR assays that have been published and used for diagnostic purposes.

Using the MgPa gene as a target for diagnostic PCR could be challenging because of the genetic variation in this gene. This was clearly demonstrated in our MgPa-476/MgPa-903 assay used for the first clinical study (2). At the time of the study, detection of M. genitalium by PCR was a new method, and a confirmatory assay was needed to substantiate the findings with the MgPa-1/MgPa-3 assay. Using purified DNA from the M. genitalium G-37 strain and the six other strains available, the MgPa-476/MgPa-903 primer-set provided excellent results with an apparent low limit of detection. However, during the study, it became clear, that although a complete concordance was found between the results with the two primer-sets, the intensity of the amplicon produced by the MgPa-476/MgPa-903 primer-set was very low for some of the specimens clearly positive in the old assay (2). Furthermore, a remarkable variation in the restriction enzyme analysis of amplicons from the clinical specimens was found, and with some of the amplicons, an internal probe did not hybridise (2). The reason for this variation was later explained by the observation that the MgPa-903 primer was located in the B-region of the gene (described in more detail in chapter 2). This region is hypervariable due to the presence of repeat sequences scattered throughout the genome (3). As can be seen in Figure 9 (reproduced and modified from (3)), only strain M 2321 of the four Danish strains, matched the primer perfectly, and even the Danish G-37 strain had a mutation in the 5' end of the primer.

Placing a primer in the hypervariable B-region probably also explains an observation by de Barbeyrac et al . (111). They found, that whereas their MGS-1/MGS-4 primerset (located outside the B-region) could detect M. genitalium in 8 (17%) of male urethral swab specimens, the MGS-1 primer in combination with the MGS-2 primer (placed in the B-region) was unable to detect any positives. Interestingly, the amplicons produced by the MGS-1/MGS-4 primer-set flanked the same E co RI site as the MgPa-1/MgPa-3 primers, and all urethral specimens from French patients lacked this restriction site.

The experience with the MgPa-476/MgPa-903 primer-set provided further impetus to develop an assay based on the 16S rRNA gene (6). This gene sequence was not publicly available when the development of the assay began, and based on the experience with the MgPa gene, it was decided to sequence the gene from the M. genitalium G-37 strain, as well as from the four Danish isolates. As compared with the G-37 strain, the Danish isolates had a single base substitution at position 1430, included in the reverse primer of the Eastick et al . 16S assay (110) recently published. The M. genitalium 16S sequence was highly similar to that of M. pneumoniae , and, furthermore, 16S rRNA gene sequences are notorious for their secondary structures, so the task of finding efficient, specific primer-combinations was rather complicated. However, the MG16-45F/MG16-447R (Table 2) provided both a low limit of detection of M. genitalium DNA, and a good specificity towards other mycoplasma species, in particular to M. pneumoniae . An internal process control was developed, and this design has been used in most of the other assays developed in our laboratory since then. It was concluded, that screening with the 16S assay and confirming all positive results with the MgPa-1/MgPa-3 assay provided a good sensitivity and specificity (6), whereas the MgPa-476/MgPa-903 primer-set was more or less useless, at least with the modifications introduced between its development and the 16S validation study. Since 1994, the 16S PCR has been used as the primary assay in all clinical studies performed in our laboratory (4; 5; 7; 8; 112-120), and the protocol has been adapted in several other laboratories (89; 107; 121).

Another 16S rRNA gene based assay was described by Sasaki et al . (122). This assay, however, has only been used on simulated clinical specimens, and the reported limit of detection was about 1000 M. genitalium cells, which is insufficient for clinical use.

One of the more recent developments in diagnostic PCR is the homogeneous real-time PCR. The major advantage of this type of assay is the closed format, where amplification is detected without opening the PCR tube. Thereby, the risk of amplicon contamination is drastically reduced. Furthermore, real-time assays can be performed quantitatively with a wide dynamic range; the probe-based assays provide additional specificity to the assay, and the subjective reading of an ethidiumbromide stained agarose gel is no longer needed.

The first real-time PCR assay for M. genitalium was published by Yoshida et al . in 2002 (123). The assay was based on detection of the 16S rRNA gene, but the forward primer My-ins hybridised to almost all species of mycoplasma ( > 50 hits with 100% homology in a GenBank Blast search). The reverse primer MGSO-2 is a modification of the mollicutes specific primer described by van Kuppeveld et al . (124), which is used for detection of mycoplasma infection in cell cultures in many laboratories. The modified primer has 100% homology with all members of the M. pneumoniae 16S cluster. Consequently, the specificity resides only in the probe, which has only two mismatches with the M. pneumoniae sequence. It is, therefore, likely, that the primers will amplify 16S sequences from a wide variety of mollicutes including ureaplasmas, and even though the probe is specific for M. genitalium , the competition for primers and other reaction components is likely to influence accuracy of the quantitation of M. genitalium DNA. Questions could also be raised against the sensitivity of the assay, since it was shown that all five FVU specimens from men with a simultaneous M. genitalium and gonococcal infection were negative in the TaqMan assay, and the same applied to one of two specimens from men with asymptomatic M. genitalium infection. These specimens were previously found positive in an assay based on sequencing of PCR amplicons obtained with mollicutes 16S specific primers (125).

Due to the homology between the M. genitalium and M. pneumoniae 16S rRNA genes, and the predominance of secondary structures, the design of specific and sensitive primers and probes for a TaqMan assay was found to be difficult. From previous experience, it was known that results obtained with the MgPa-1/MgPa-3 primer-set (1) correlated very well with those of the16S rRNA gene based assay (6). Since partial MgPa sequence information from four Danish strains was available (3), primers and probes from conserved regions of the MgPa gene were selected (8). An internal processing control was designed using the same principles as used in the 16S assay. The specificity of the new assay was validated against all human mycoplasmas and a range of phylogenetically related species. Sensitivity was validated by comparing the results of the 16S rRNA gene PCR, by analysing the standard curve in the presence of different clinical specimen types, and by adding known amounts of M. genitalium to clinical specimens to control for the matrix effect.

Using this assay it was shown, that men with urethritis had higher DNA loads in the FVU than men without urethritis. This relation was less obvious when urethral swab specimens from the same patients were tested, but if data for grey-zone and high-grade urethritis were combined, it could be shown that patients with urethritis had higher DNA loads than controls. The TaqMan assay appeared to be slightly more sensitive than the 16S PCR. This was shown by finding two TaqMan positive of 10 16S PCR negative urethral swab specimens from patients with a confirmed infection as documented by a positive FVU specimen. Surprisingly, the sequence of the forward primer was not completely conserved when more MgPa-1/MgPa-3 amplicons from patients were sequenced. However, comparing detection limits with the 16S assay and diluted specimens proved that this variation did not influence sensitivity. The TaqMan MgPa assay has been adapted to a SmartCycler® instrument in Karlstad, Sweden with very good results (Thomas Ahlquist, personal communication) and in Glasgow the RotorGene® instrument is being used also with good results (Jill Shepherd, personal communication).

In order to compare the quantitative results found with the TaqMan assay, 83 positive urethral swab specimens were tested in a dual-probe LightCycler® assay targeting the housekeeping gene gap (126). An excellent correlation between the two methods was obtained ( Figure 10 ), and presently a larger head-to-head study is in progress.

Recently, a LightCycler assay detecting the P115 gene (MG299) with LNA (locked nucleic acid) probes was applied for detection of M. genitalium and for monitoring the response to treatment in nine infected men (127). Thus, important new insight in M. genitalium disease can be gained by the use of quantitative techniques.

Using the best PCR methods available is not sufficient, if the specimen is inadequate. In an attempt to determine the optimal type of diagnostic specimen both for M. genitalium and for C. trachomatis , the relative performance of FVU, urethral and cervical swab specimens was determined (9). Using a Chelex extraction method, FVU proved to be the optimal specimen type for men detecting > 97% of the M. genitalium and C. trachomatis infections. In women, however, FVU detected only 71% of the M. genitalium infections and 90% of the C. trachomatis infections. Combining the detection from FVU with a cervical swab specimen increased these figures to 96% and 99%, respectively. Consequently, FVU specimens from women should always be supplemented with a cervical swab specimen for optimal sensitivity, whereas the urethral swab specimen, which is considered painful by most women could be avoided. Another important finding was that female FVU specimens appeared to be unstable at -20°C since 27% of the specimens subjected to sample preparation after storage were found to be false negative, whereas those extracted initially after receipt in the laboratory remained positive (9).

In conclusion, diagnosis of M. genitalium infections in the future should be based on real-time PCRs since these are highly sensitive and specific. The quantitative aspects may provide important information in relation to a variety of scientific questions, but appears to be of less importance from a diagnostic point of view. It is encouraging, that the TaqMan MgPa assay has proven so easy to adapt to other instrument platforms and until a commercially available kit is at hand, it may be one of the preferred methods for new laboratories inexperienced in "home-brew" PCR.

4.1. Modified Henle-Koch postulates

The original Henle-Koch postulates are a series of requirements that must be met before an organism can be considered causative for a given disease (130). The original postulates are listed in Table 3 .

These postulates were established in 1840 by Jakob Henle and further developed by his pupil Robert Koch based amongst others on his observations of anthrax.

One of the important strengths of the Henle-Koch postulates is the fact, that if a given microorganism is established as the causative agent of a specific disease, it is implied that the disease can be treated and/or controlled by elimination of the organism or by preventing contact with the organism.

However, fulfilling the postulates is a very stringent requirement; for certain diseases and organisms, only some of the postulates can be fulfilled; e.g. the causative agent for leprosy, Mycobacterium leprae , has not been isolated in pure culture and subsequently used to induce experimental disease. According to a rigorous interpretation, viruses could never fulfil the postulates since they cannot be propagated without cells. Furthermore, the postulates leave no room for factors other than the implicated microorganisms playing critical roles in the causation of disease, such as host susceptibility or interaction with other organisms. Moreover, the model is not suited to manage diseases that may be due to several different causative agents like pneumonia and urethritis.

As a consequence of these shortcomings of the Henle-Koch postulates, many revisions of the postulates have been proposed (reviewed in (131)). David Taylor-Robinson suggested modifications of the postulates (132), that should be fulfilled before regarding a mycoplasma species as a cause of disease ( Table 4 ).

The degree to which M. genitalium fulfils these criteria in the case of male NGU and NCNGU will be discussed in the following.

4.2. Epidemiology

Due to the difficulties in culturing M. genitalium , epidemiological studies have been few in the years after the discovery of the organism. In a study using culture and serology, Taylor-Robinson et al . (83) failed to isolate M. genitalium although they reported "presumptive isolation" based on colour-change in the mycoplasma broth from 7 (32%) of 22 men with NGU, from 5 (42%) of 13 men with NCNGU, and from 2 (10%) of 22 men without urethritis. M genitalium was also presumptively isolated from rectal swabs from both homo- and heterosexual men. Using a microimmunofluorescence (MIF) antibody test, a significant rise in antibodies was detected in 4 (29%) of 14 patients with NGU and in 2 (12%) of 17 without urethritis. Disappointingly, only one of the four NGU patients responding serologically had presumptive isolations, and none of the two men without urethritis, who seroconverted, had presumptive isolations (83). Using colour change in the mycoplasma broth without further identification as evidence of M. genitalium growth may have a low specificity since it could be caused by other bacteria, including unknown mycoplasma species. Several years later, some of the specimens remaining from their study were examined with PCR, and the presumptive isolations from the urethral specimens were to some extent corroborated, since M. genitalium was detected in 9 (50%) of 18 NGU specimens, and 1 (14%) of 7 specimens from men without urethritis. No details were given, however, on the precise correlation between culture and PCR results (133; 134). Hooton et al . (104), using a nick-translated probe prepared from genomic DNA of M. genitalium , studied 203 men attending a sexually transmitted disease clinic, and detected M. genitalium in the same proportion of men with acute gonococcal urethritis (14%; 3/21), acute C. trachomatis positive NGU (10%; 3/30), acute NCNGU (13%; 4/31); and in 12% (10/84) of patients without urethritis, but in 27% (10/37) of men with persistent or recurrent NGU. M. genitalium was found more often in homosexual than in heterosexual men, which led the authors to suggest that M. genitalium may reside in the gastrointestinal tract (104).

Further clinical studies were made possible due to the development of PCR assays (1; 105).

In the first clinical study applying the PCR technique, urethral, rectal, and throat samples from 99 male STD clinic attendees were studied and M. genitalium DNA was found in 17 (17%) of the urethral swabs, but in none of the rectal and throat swabs indicating that the urogenital tract is the primary site of infection (2). Significantly more men with NGU (27%; 13/48) were positive for M. genitalium than were those without urethritis (9%; 4/47). In men with urethritis, M. genitalium was found more often in C. trachomatis negative NGU (35%; 12/34) than in those with chlamydial NGU (7%; 1/14) indicating that the two microbes may act as separate causes of urethritis (2). Almost simultaneously with the publication of the Danish data, Horner et al . from David Taylor-Robinson's group published a study with the same conclusions (135). The Danish study was criticised for not having examined the asymptomatic men for microscopic signs of urethritis (136). At the time of the study, asymptomatic men attending the STD clinic in Copenhagen did not routinely have urethral smears examined for microscopic signs, so the chosen inclusion criteria would be more true to the situation in everyday clinical life (137). Furthermore, the conclusions regarding the role of M. genitalium would be biased only in direction of decreasing the importance of M. genitalium . Later studies (5; 138; 139) have actually demonstrated that M. genitalium is more closely associated with symptomatic than with asymptomatic urethritis. Only one (10%) of 10 men with asymptomatic urethritis was M. genitalium positive (5) and Horner and Taylor-Robinson found only two (8%) of 25 men with asymptomatic urethritis positive for M. genitalium , a prevalence closely matching that found in male STD clinic attendees without urethritis (138). Horner et al . recently presented further evidence that M. genitalium is associated with discharge using data from the original study (135) and applying multivariate analysis to control for confounding factors (139) and Taylor-Robinson et al . (140) studied 52 men with persistent or recurrent NGU and found 11 (21%) to be M. genitalium positive.

Since the two original studies (2; 135), several PCR based studies have been presented (5; 89; 110; 117; 121; 125; 127; 129; 141-153). A total of 5455 patients have been examined in 23 published studies including three studies without controls (summarised in Table 5) showing an overall M. genitalium prevalence of 20.8% (470/2261) among men with NGU and 5.9% (124/2107) among the control patients. In 20 studies, a case-control design was used. Although there were important differences between the studies in terms of design, clinical criteria used for diagnosis of urethritis, and different PCR methods used, the heterogeneity between studies was not found to be statistically significant, thus allowing for calculation of a pooled odds ratio. However, bearing the differences in mind, the pooled odds ratio should be interpreted with caution. Of the 2143 patients with NGU, 452 (21.1%) were M. genitalium positive as compared to 122 (6,4%) of the 1920 patients without NGU resulting in a DerSimonian-Laird random effects analysis pooled odds-ratio of 4,2 (95% CI 3.1-5.5; p < 0.0001). In all studies, M. genitalium has been detected more often in men with NGU than in those without (Summarised in Figure 11 ).

In 16 of the clinical studies, it is possible to calculate the prevalence of M. genitalium in men with NCNGU (2; 5; 89; 117; 121; 125; 127; 129; 135; 141; 142; 145-148; 151) (summarised in Figure 12 ). In these studies, the M. genitalium prevalence in the NGU group was 19.3% (345/1786) compared to a C. trachomatis prevalence of 27.7% 496/1786). Except for the studies of Gambini et al . (148) and Mena et al . (129), the M. genitalium prevalence has invariably been higher in the NCNGU group than in the total NGU group. Among the NCNGU patients, 21.9% (283/1290) were M. genitalium positive compared 6.0% (93/647) of the patients without urethritis resulting in a DerSimonian-Laird pooled odds-ratio of 5.8 (95% CI 3.9-8.7; p < 0.0001). M. genitalium was found less often in chlamydial NGU (12.5%; 62/496) than in the NCNGU group (p < 0.0001, Fisher's exact test).

Thus, M. genitalium behaves largely independent of C. trachomatis and with a prevalence of about two thirds that of C. trachomatis among men with urethritis. It is interesting to note, that in the studies performed on black African men (117; 150; 151), the odds ratios for being M. genitalium positive both in NGU and in NCNGU tend to be somewhat lower. This could partly be explained by the syndromic definition of urethritis used in these studies where microscopy of urethral smears was not performed, and absence of symptoms was used to define the control group. However, the higher prevalence in men without symptoms may also reflect true variation in the relative importance of M. genitalium in populations with a very high burden of STIs.

Finding M. genitalium in higher numbers in patients with NGU than in those without was another of the revised Henle-Koch postulates (Table 4). Recently, data supporting the presence of a higher number of M. genitalium DNA copies in FVU from patients with NCNGU than from asymptomatic men was presented (123). These researchers applied a quantitative TaqMan 5'nuclease assay based on 16S rRNA gene sequences to M. genitalium positive specimens. Both the sensitivity and the specificity of the assay could, however, be questioned (discussed in more detail in chapter 3.3). Only two asymptomatic men were examined and one of them was actually negative in the TaqMan assay.

The TaqMan quantitative PCR assay (8) (described in more detail in chapter 3.3) was applied to test FVU and urethral swab specimens and showed that there was a significantly higher number of M. genitalium genome equivalents in the FVU from men with urethritis than from those without ( Figure 13 ).

In conclusion, the first of the modified Henle-Koch postulates has been fulfilled with a high degree of certainty both for NGU and for NCNGU.

4.3. Antibody response

As described in chapter 3.2, the cross-reactions between M. pneumoniae and M. genitalium have significantly hampered the use of serology for diagnosis and epidemiological studies. Using a MIF assay (96), Taylor-Robinson et al . observed a four-fold or greater rise in antibodies to M. genitalium in 29% of patients with NGU and in 12% of those without. However, the correlation between antibody response and presumptive isolation of M. genitalium from the urethra was not very convincing (83; 133). As a part of the first Danish prevalence study (2), acute phase serum specimens were collected. IgG, IgM, and IgA was measured by ELISA and immunoblotting, but a clear correlation between the PCR status and the antibody response could not be shown (chapter 3.2). Using Lipid-Associated Membrane Proteins (LAMP), Wang et al . (99) studied the seroprevalence of M. genitalium antibodies in different populations. They collected serum and urine specimens from 104 patients (68 HIV infected asymptomatic individuals and 36 intravenous drug users), and found that 40 (38%) were LAMP ELISA positive. 15 (38%) of these were urine PCR positive for M. genitalium , indicating that antibodies from previous exposure were persisting. On the other hand, none of 64 LAMP ELISA negative were PCR positive demonstrating an apparent satisfactory sensitivity. The correlation between a positive M. genitalium PCR test and detection of M. genitalium antibodies was highly statistically significant. Whether the PCR positive patients had signs or symptoms of urethritis was not mentioned. Considering the apparent performance of the LAMP assay, it is surprising, that it has not been studied more thoroughly.

Although further studies are needed, particularly with collection of follow-up serum specimens from PCR positive patients, the second of the modified Henle-Koch postulates appears to be at least partly fulfilled.

4.4. Effect of treatment

Considering the amount of evidence linking M. genitalium to NGU, and in particular to NCNGU, it is surprising, that randomised controlled clinical trials aimed at determining the optimal antibiotic treatment regimen have not been conducted. The relevance of in vitro studies can be questioned, since the number of strains is small. Furthermore, in view of the fact that seven of the strains circulating seem to have a clonal origin (53), the conclusions that can be drawn from most of the published studies are very limited. The only study including the four strains isolated from Danish patients (92) is, therefore, the most relevant one to consider. Furthermore, this study gives both the usual initial minimal inhibitory concentration (MIC), defined as the MIC at the time the control broth culture shows colour change, and a final MIC read one week later. From a clinical point of view, the final MIC would seem more appropriate, since it reflects the presence of M. genitalium cells capable of surviving the antibiotic. The surviving cells could be either a resistant subpopulation or an effect of the slow growth of M. genitalium , allowing resting cells to survive the effect of bacteriostatic antibiotics. The main conclusions from Hannan's study (92) were, that the M. genitalium strains tested were highly susceptible to the macrolide group of antibiotics with erythromycin showing a final MIC of max. 0.1 mg/L, and azithromycin with a final MIC of max. 0.005 mg/L. The only other group of antibiotics capable of suppressing the growth of all M. genitalium strains was the newer quinolones with extended Gram positive spectrum exemplified by Sparfloxacin with a max. final MIC of 1 mg/L. Most importantly, however, the final MIC for the tetracylines reached 10 mg/L for the least susceptible strains (92). This is a very important observation, since tetracylines are the drugs of choice for treatment of NGU in most countries due to the low cost and high clinical efficacy towards C. trachomatis infections. Furthermore, since the G37 type-strain and the clonal derivatives of this strain frequently used in the evaluation of new antibiotics is susceptible to tetracyclines, the erroneous conception that M. genitalium is susceptible in vitro to tetracyclines is widespread (155)

Several studies have indicated that treatment for C. trachomatis urethritis with tetracyclines or quinolones is not always sufficient to eradicate M. genitalium (summarised in Table 6 ).

In all studies, the concordance between persistent positive M. genitalium PCR and clinical treatment failure has been a common feature. Horner and colleagues (135; 156) found that seven of seven men who had M. genitalium detected after doxycycline treatment had urethritis at follow-up. Likewise, Johannisson et al . (147) found that eight of 13 men (62%) still harboured M. genitalium after treatment with tetracyclines for 10 days and that six of them had urethritis.

In a study of Swedish STD clinic patients (7; 113), 10 of 17 men (59%) treated with tetracyclines were M. genitalium positive at follow-up. The patient's symptoms responded to treatment, but the urethritis remained. In a subgroup of M. genitalium positive men with urethritis, 16 received treatment with tetracyclines and six received 1.5 g of azithromycin distributed as a 500 mg loading dose and 250 mg in the following four days. Whereas all patients treated with azithromycin were PCR negative at follow-up, 10 (63%) of the 16 patients treated with tetracyclines were M. genitalium positive at follow-up (p=0.015)(7; 113).

In contrast, Gambini et al . treated 35 M. genitalium positive patients with doxycycline (200 mg x 7) and 17 positve patients with azithromycin (1g stat) and found that only two (6%) of those treated with doxycycline and three (18%) of those treated with azithromycin had "partial clinical response" and remained M. genitalium positive one week after end of therapy. Those patients responding microbiologically, were also clinically cured (148). Whether the higher cure rate after doxycycline resulted from use of a 200 mg daily dose for seven days, as compared to the other studies using doxycycline in a lower dosage remains to be determined but appears unlikely based on the clinical experience in Scandinavia. The early follow-up could also be questioned, but no systematic studies regarding clearance of M. genitalium DNA after treatment have been conducted.

The effect of quinolones could also be questioned. Maeda et al . examined 12 M. genitalium PCR positive men treated for NGU with levofloxacin (the l -isomer of ofloxacin) of whom eight (66%) remained PCR positive after treatment; one had persistent NGU, and of the seven that appeared to be cured, two did not return and five returned with recurrent M. genitalium PCR positive NGU. Interestingly, in a subsequent study by the same group (158), the M. genitalium load was measured by quantitative TaqMan PCR. The M. genitalium DNA load increased from below the detection limit one week after initiation of therapy to > 10 4 DNA copies/ml in four of the patients at the time of recurrence.

Considering the diversity in the dosage of tetracyclines and azithromycin that has been used in the published studies, a randomised controlled trial should include at least two arms comparing doxycycline 100 mg bd for seven days and azithromycin 1 g stat as recommended in the clinical effectiveness guidelines (159; 160). However, considering the slow growth of M. genitalium , which may call for a longer duration of treatment, the azithromycin regimen comprising 500 mg in a single dose on day one, followed by 250 mg od for four days, which has provided excellent cure rates, should also be considered. That a longer dosage of azithromycin would be beneficial, was actually provided in a much-overlooked study from 1991 (161). Here it was shown, that azithromycin 500 mg in a single dose on day one followed by 250 mg od for two days was superior both to azithromycin 1 g stat and doxycycline 200 mg daily for seven days for treatment of male urethritis.

In conclusion, although the published studies are small and could be confounded by reinfection from untreated partners and by the persistence of DNA from non-viable M. genitalium cells, it seems as if there is a strong correlation between eradication of M. genitalium and clinical cure. Those patients with microbiological treatment failure appear to suffer from persistent or recurrent disease. It would therefore be reasonable to accept that the third demand of the modified Henle-Koch postulates is fulfilled.

4.5. Transmissibility and animal studies

The transmissibility of M. genitalium has been studied both in animal experiments, and recently, data regarding transmission between sexual contacts has begun to accrue.

4.5.1. Animal models

The use of an animal model for male urethritis has severe limitations. Many microbes have narrow host specificities and furthermore, it is impossible to control the urination of the animal, which may disturb the interpretation of urethral smears for detection of inflammatory cells. The best animal studies of experimental male urethritis caused by M. genitalium have been performed in chimpanzees. Two such studies have been published (68; 98). In the first study, four animals were inoculated intra-urethrally with the G37 type-strain. Two of the animals became persistently infected for 13 weeks; they developed a measurable antibody response after 5 weeks, and a urethral inflammatory response was noted. The animals were treated with a tetracycline drug resulting in microbiological eradication of M. genitalium and normalisation of the urethral cellular response. Six months after successful treatment, one of the animals was re-challenged with an isolate from the other animal, but whereas a naïve animal was infected with an identical inoculum, the re-challenged animal remained uninfected. Although a very limited basis for conclusions, this finding may indicate some degree of protective immunity (68). In a second experiment (98), six male chimpanzees were inoculated with strains re-isolated from the two animals in the first experiment. All animals became infected, but two of them cleared the infection spontaneously after one and five weeks, respectively. The remaining four were colonized for up to 18 weeks before treatment was commenced. Five of the six animals showed an inflammatory response in the urethral secretions. There was no correlation between the presence of a urethral PMNL response and the development of a detectable antibody response, since one animal without urethritis had a systemic antibody response, whereas the single animal without seroconversion had a clear urethral PMNL response and high titres of M. genitalium in the urethra. Two animals had M. genitalium recovered from blood cultures (98). The isolations from the bloodstream demonstrate the invasive potential of M. genitalium but may also give rise to speculations regarding haematogenous spread to other organs. Such a spread might explain some cases of sexually acquired arthritis, and the detection of M. genitalium DNA in the synovial fluid of patients with arthritis (162).

Lower primates may have a decreased susceptibility to infection. This was clearly demonstrated in a study using both of the human strains G37 and M30 (10) and the same two strains after re-isolation from intra-urethrally inoculated chimpanzees. The strains isolated directly from humans were not capable of establishing an infection in any of six rhesus monkeys ( Macacca mulatta ) and only two of four cynomolgus monkeys ( Macacca fascicularis ) were colonised and shed small numbers of M. genitalium organisms until they spontaneously cleared the organisms after 9 and 28 days, respectively. No antibody response was noted. Challenge with the chimpanzee passaged strains resulted in colonisation/infection with an inflammatory response. However, only a poor antibody response was recorded (98).

4.5.2. Sexual transmission

Only limited information about the important question as to whether M. genitalium is sexually transmitted is available. Keane et al . (146) studied 39 men with NGU (index patients) and their female partners. C. trachomatis was detected in six (43%) female partners of 14 C. trachomatis positive men. M. genitalium was detected in seven (58%) female partners of 12 M. genitalium positive men. Thus, the concordance rate for M. genitalium seems to be as high as that for C. trachomatis . M. genitalium was also detected in two (9%) of 22 female partners to M. genitalium negative men; the corresponding figure for C. trachomatis was four (11%) of 37° C. trachomatis negative men (146). In a recent study from Sweden, Anagrius and Loré (121) attempted partner tracing in 27 male and 26 female index patients infected with M. genitalium . Out of 38 reported male partners to female index patients, 26 were examined for M. genitalium and 10 (38%) were positive. Corresponding figures for 22 of the 30 reported female partners to male index patients were 10 (45%) infected (121). In another Swedish study, 17 female partners of 18 M. genitalium positive men were traced and 12 (71%) of them were M. genitalium positive (one of the women was given as the steady partner for two men) (Lars Falk, personal communication). Transmission rates of C. trachomatis have recently been established with the use of PCR, and rates of transmission have been in the order of 26% to 68% in large series (163; 164). Thus, M. genitalium appears to be transmitted at least as efficient.

Although further studies regarding the sexual transmission of M. genitalium are needed, particularly on the transmission from female index patients to men, the chimpanzee studies have documented that the fourth criterion of the modified Henle-Koch postulates is fulfilled.

4.6. Conclusions

Using the modified Henle-Koch postulates for establishing that M. genitalium is a causative agent in urogenital tract disease, it seems reasonable to conclude, that M. genitalium can cause NGU. M. genitalium is sexually transmissible with transmission rates similar to those of C. trachomatis . Compared to the number of patients with urethritis caused by C. trachomatis , the share of patients with urethritis caused by M. genitalium is probably smaller, but not insignificant. In a range of studies, M. genitalium has been found in 10-45% of men with NCNGU, and considering the poor microbiological cure rate after the conventional treatment with tetracyclines, a significant number of patients may develop recurrent or persistent urethritis. Randomised controlled clinical trials aiming at determining the optimal treatment of the infection, are urgently needed.

5.1. Epidemiology

5.1.1. Prevalence studies without controls

As part of the validation of the MgPa1-3 PCR assay (1), M. genitalium was detected in 5 (7%) of 74 women having specimens submitted for C. trachomatis culture. Palmer et al . found M. genitalium in 10 (18%) of 57 women attending an STD clinic using their heminested PCR (165) and Tsunoe et al . from Japan found, that among 174 female commercial sex workers, M. genitalium was detected in 22 (13%) as compared to 1 (1%) of 90 pregnant women (166). This prevalence should be judged against a C. trachomatis prevalence of 19% and a prevalence of N. gonorrhoeae of 33% in this group. In contrast, de Barbeyrac et al . did not find M. genitalium in any of 55 female urogenital tract specimens from STD clinic patients (111). In an ongoing prospective study from Örebro, Sweden, an M. genitalium prevalence of 6% was found among 464 women attending the STD clinic. There was no significant difference between patients with M. genitalium and patients with C. trachomatis in regards to signs and symptoms. However, C. trachomatis was found more frequently, being detected in 10% of the women (Falk et al . submitted).

5.1.2. Cervicitis

A few studies have addressed the correlation between M. genitalium infection and cervicitis. One of the major problems in interpreting these studies has been the varying definitions of cervicitis. Some studies have considered the presence of 10 PMNLs in a cervical smear significant, whereas others demand 30 PMNLs/hpf to define cervicitis. The finding of more PMNLs than epithelial cells in a vaginal wet smear is commonly used and the specificity of these signs has not been thoroughly validated. Less objective measures such as the presence of yellow mucopus from the cervical canal collected on a white cotton swab have also been included. One problem is the lack of exact definitions of what is normal and abnormal. In a study of 59 women between 22 and 26 years of age who participated in a Papanicolau smear screening, none of the women were M. genitalium PCR positive, and only one (2%) was C. trachomatis positive. The standard definition of < 4 PMNL/hpf was used for urethritis and cervicitis was defined as more PMNL than vaginal epithelial cells in the vaginal wet smear. Surprisingly, 22% of the women had signs of urethritis and/or cervicitis demonstrating a low specificity of these signs (167).

One early study by Uno et al . defined cervicitis by endocervical discharge and ≥ 20 PMNL/hpf and detected M. genitalium in 5 (8%) of 64 women as compared to none of 80 asymptomatic pregnant women (168).

In a recent US study using archived cervical secretions from 719 women collected in 1984-86 on filter paper, Manhart et al . found that M. genitalium was strongly associated with cervicitis being detected in 7% of all the women examined but in 11% of women with cervicitis (≥ 30 PMNLs/hpf) and in 5% of women without cervicitis (p=0.004). In a multivariate logistic regression analysis correcting for other factors found to be associated with cervicitis, and excluding the 172 women infected with N. gonorrohoeae and/or C. trachomatis , M. genitalium remained strongly associated with cervicitis (OR 3.1; 95% CI 1.5-6.8), further supporting an independent role for M. genitalium as a cause of cervicitis. The attributable risk percent (estimated by OR-1/OR) was 70%, suggesting that among women with cervicitis and M. genitalium , 70% of cervicitis can be attributed to M. genitalium (107).

In a study from Sweden (121) including 445 women, an overall prevalence of M. genitalium of 6% was found. M. genitalium was significantly associated with signs of urethritis or cervicitis as defined by ≥ 5 and ≥ 30 PMNLs/hpf, respectively, since 21 (10%) of 217 women with cervicitis/urethritis were M. genitalium PCR positive as compared to 6 (3%) of 218 without (p=0.003) (121). In contrast, Casin et al . were unable to demonstrate any correlation between the presence of M. genitalium and clinical, demographic or microbiologic data in 170 consecutive women with genital symptoms attending an STD clinic in Paris (169). However, only 25 asymptomatic women attending a private gynaecological practice were included and 4 (16%) were M. genitalium positive. Surprisingly, 65 (38%) of the symptomatic women were positive in the MgPa1-3 PCR assay, which is remarkable when compared with a C. trachomatis prevalence of 8%. This raises the suspicion that the specificity has not been optimal which may obscure the relationship between symptoms and the infection.

5.1.3. Pelvic inflammatory disease

Pelvic inflammatory disease (PID) is the clinical syndrome caused by the spread of microorganisms from the lower to the upper genital tract. A range of bacteria including those found in bacterial vaginosis, C. trachomatis , and N. gonorrohoeae can cause PID. Establishing a connection between M. genitalium and upper genital tract infection is of major importance in determining the significance of the infection. PID is more likely to lead to severe disease with loss of working days and to long-term sequelae such as tubal factor infertility, ectopic pregnancy, and chronic pelvic pain. The fundamental problem in PID research is concerned with case definition and diagnostic accuracy. Laparoscopy is considered the "gold standard", but is not a realistic possibility in daily clinical practice (170). Studies using laparoscopy for case definition are likely to be biased towards a more severe spectrum of the disease and, therefore, other invasive methods such as endometrial biopsies for detection of plasma cell endometritis may play a role in research, but is not applicable as a standard diagnostic procedure. In a study of 115 women presenting to an STD clinic in Kenya with acute pelvic pain (108), plasma cell endometritis was found in 58 and M. genitalium was detected in cervical swabs or endometrial biopsies in 16% compared to only 2% of women without endometritis (p=0.02). As many as 33% of the women studied were HIV positive, but M. genitalium was not detected significantly more often in the HIV infected women. In another study of women with clinically suspected PID (171), M. genitalium was detected more often in HIV positive women (19% vs 5%), but the data presented does not allow an evaluation of the importance of this finding. Whether the results of the two studies could be generalised to HIV-negative Caucasian populations is not clear.

In a recent case-control study, Simms and colleagues (172) examined 45 women with clinically diagnosed PID and found M. genitalium DNA by PCR in endocervical swabs from nine (16%) as compared to none of 37 control patients. It should be noted, however, that the control population consisted of women attending for bilateral tubal ligation. These women were significantly older, and would be expected to have considerably less risk taking behaviour regarding STIs.

Earlier studies relying on serology have been controversial. Møller et al . (173) studied 166 women with clinically suspected acute PID admitted to hospital in Århus, Denmark. Serum was collected at 0, 1, and 4 weeks after admission; 31 women with negative serology for C. trachomatis and M. hominis as determined by MIF and indirect heamagglutination (IHA) tests, respectively, were examined for antibodies to M. genitalium by MIF. Using a cut-off titre at ≥ 16, 12 (39%) showed a ≥ 4 fold change in titre between acute and convalescence sera. However, no information regarding the changes in M. genitalium titre among women with C. trachomatis PID or without PID was given, and therefore, the significance of the findings is difficult to interpret. In contrast, Lind et al . (174) examined 95 patients seen in Odense, Denmark with clinically suspected PID. After laparoscopy, 39 women were considered not to have salpinigitis. From these women seven had paired serum specimens available, and none showed a change in titre. Likewise, none of 21 patients with gonococcal or chlamydial salpingitis seroconverted, and only one (4.8%) of 21 patients with salpingitis but without evidence of gonococcal or chlamydial aetiology responded with a ≥ 4-fold rise in an IHA test for M. genitalium . Since the difference in test principle could be important, sera from 10 of the patients from whom three or more sera were available were examined in a MIF test comparable to that used by Møller et al . (173). However, only in one was a fall in titre from 80 to 10 against both the M. genitalium and the M. pneumoniae control antigen observed (174). The reason for the discrepancy between the two studies is difficult to determine exactly. However, the number of sera from Odense studied with the MIF test was small, and the difference not statistically significant.

5.1.4. Bacterial vaginosis

Bacterial vaginosis (BV) is a very common clinical syndrome with characteristic clinical, biochemical, and microbiologic features. BV is usually diagnosed according to the Amsell criteria (175) demanding at least two out of four criteria fulfilled: 1. Thin, malodorous homogeneous, grey, nonpurulent vaginal discharge; 2. pH of the discharge > 4.5; 3. Presence of "clue cells" on wet mount examination, and 4. "Fishy" odor when mixed with 10% potassium hydroxide. Another definition is provided in the Nugent criteria by which Gram stained vaginal secretions are scored according to the microbial flora (176) or in modifications of the criteria described by Spiegel (177). The Gram scoring systems are based on the relative amount of the Lactobacillus morphotype (large Gram-positive rods) and of the Gardnerella morphotype (small Gram-variable rods). M. hominis is found more often in women with BV with isolation rates up to 75%, and remains significantly associated even after adjustment for co-infection with other BV-associated flora (178; 179). To a lesser extent such a correlation is also found for U. urealyticum (179) with isolation rates as high as 95%, but the frequent isolation in women without BV (around 40%) obscures the connection. In one of the early reports on the M. genitalium PCR assay developed by Palmer et al . (165) M. genitalium was found in 4 (16%) of 25 unselected women and in 3 of 10 women with BV. More recently, Keane et al . (180) studied 15 women with BV, none of whom were M. genitalium positive, as were only 2 (12%) of 17 women without BV attending an STD clinic. This lack of association between BV and M. genitalium was also reported as part of the cervicitis study mentioned previously (107). Thus, also in this respect, M. genitalium behaves very much like C. trachomatis .

5.1.5. Adverse pregnancy outcome and infertility

There is little information as to whether M. genitalium can cause adverse pregnancy outcome, either as preterm labour, abortion or stillbirth. One study has been published where M. genitalium was detected in only 4% of midtrimester vaginal swabs from 124 women delivering preterm (181). This is probably slightly higher than the prevalence in a normal population of pregnant women as estimated from a recent study (182) where only 0.7% (6/915) of women pregnant in the first trimester were M. genitalium positive. However, even with a significant difference in the two prevalences it is not going to be an important factor. M. genitalium was more common in women aged < 20, those with concurrent bacterial vaginosis, women of Afro-Caribbean or Black African ethnic origin, women in social classes 3-5, and single women. Only one woman with M. genitalium infection miscarried, and none of those followed until term had a preterm birth but numbers were small (182)

These findings were corroborated in a recent case-control study of 1014 women who gave birth or aborted at the national reference hospital of Guinea-Bissau (183) of whom 6% were M. genitalium positive, but it was not shown to have any effect on the outcome of pregnancy.

Regarding infertility, only indirect evidence has been presented. In a Danish study (100), sera from 308 infertile women were investigated for antibodies to M. genitalium by immunoblotting against whole cell proteins of M. genitalium and M. pneumoniae as well as against a recombinant MgPa antigen. Reactions in patients' sera were evaluated against the MgPa band in the whole cell antigen preparation, and reactions with the cloned MgPa fragment was used for confirmation. Among the women with tubal factor infertility, 22% (29/132) were seropositive as compared to 7% (11/176) of women with either male factor infertility or unexplained infertility with normal tubes. The recombinant MgPa fragment has later been used as antigen in an ELISA test with promising results (Helle Friis Svenstrup, personal communication). Such an ELISA test would facilitate seroepidemiological studies and be a very valuable tool if it proves to be sufficiently sensitive and specific.

In conclusion, the first of the modified Henle-Koch postulates seems to be fulfilled with some degree of certainty for cervicitis and PID. Only indirect evidence exists for infertility, whereas BV and adverse outcome of pregnancy does not appear to be associated with M. genitalium . The clinical studies of M. genitalium in different diseases are summarised in Table 7 .

5.2. Antibody response

The two studies of salpingitis (173; 174) and the Danish study of patients with infertility (100), have been discussed above. Only recently has one study been published (90) looking at the development of antibodies in women with acute infections documented by PCR. However, as discussed in chapter 3.2 severe criticism could be raised against the methodological approach of the study. More studies are therefore needed in order to fulfil the second of the modified Henle-Koch postulates. Particularly in PID studies, the simultaneous detection of M. genitalium in the lower genital tract and the development of an antibody response could be very informative.

5.3. Effect of treatment

No studies specifically aimed at determining the treatment efficacy in women have been published. However, during a prospective study of the prevalence of M. genitalium in Swedish STD patients (7), 26 M. genitalium PCR positive women were identified. Among 14 women treated with tretracyclines in standard dosages, 10 (71%) were still PCR positive at follow-up, and without improvement in symptoms and signs. In contrast, out of 12 women treated primarily with azithromycin (500 mg day 1 and 250 mg the following 4 days) due to a known M. genitalium positive partner all of 10 women returning for follow-up were PCR negative. The 10 patients with treatment failure after treatment with tetracyclines were subsequently treated with azithromycin and all nine who returned for follow-up were PCR negative. Obviously, this study was not designed as a treatment study, but the results emphasise that a randomised controlled trial is needed in order to document the fulfilment of the third of the modified Henle-Koch postulates.

5.4. Transmissibility and animal studies

5.4.1. Animal models

To some extent, animal studies of female M. genitalium infections have been more rewarding than those performed in male animals. Female squirrel monkeys became heavily colonised shedding more than 10 5 ccu/ml 2-3 weeks following intravaginal inoculation. A serum antibody response was detected as late as 10-11 weeks after the inoculation (98). Six female marmosets were inoculated intravaginally. Four of the animals were infected as indicated by repeated recovery of the organisms on vaginal swabbing, and the infection persisted for up to 149 days. In addition, the infected marmosets exhibited a serum antibody response detectable by MIF, and a persistent vaginal PMNL response not seen in two uninfected (inoculated with a 100-fold dilution of the inoculum received by the other animals) and in two uninoculated animals (184).

All four female chimpanzees inoculated intravaginally with M. genitalium reisolated from male chimpanzees shed organisms for 12 to 15 weeks, at which time-point they were treated with tetracycline. In contrast to the male chimpanzees, none of the animals had M. genitalium recovered from the blood. A vaginal PMNL response was documented, but overt discharge was not observed. All inoculated chimpanzees developed a clear-cut antibody response, in some detected after only 3 weeks.

In experiments where M. genitalium was inoculated directly into the oviducts, marmosets and grivet monkeys developed a moderate to severe endosalpingitis with infiltration of inflammatory cells into the tubal epithelium and luminal exudates. Adhesions very similar to that seen after inoculation of C. trachomatis were also observed. Antibody responses were observed 25 to 32 days after infection of the marmosets, but the titres did not appear to be maximal before 8 weeks after the infection. A remarkable finding was the failure to recover M. genitalium from the fallopian tubes. This may be indicative of problems when studying human material (185).

5.4.2. Sexual transmission

The demonstration that M. genitalium is sexually transmissible between partners has been described in chapter 4.1.4.

In conclusion, the results of animal experiments offer substantial evidence for the pathogenicity of M. genitalium for the urogenital tract of subhuman primates and the agent is sexually transmitted between partners. Thus, the fourth of the modified Henle-Koch postulates would appear to be fulfilled.

5.5. Conclusions

Using the modified Henle-Koch postulates, it seems reasonable to conclude, that M. genitalium can cause mucopurulent cervicitis although the demonstration of an antibody response is lacking. M. genitalium is sexually transmissible with transmission rates similar to those of C. trachomatis . The role of M. genitalium in women is less well established than in men, and further studies are needed to address the relation to PID and late sequelae such as infertility. Only indirect evidence exists for infertility, and BV and adverse outcome of pregnancy does not appear to be associated with M. genitalium infection. As in men, randomised controlled clinical trials aiming at determining the optimal treatment of the infection are needed.

6.1. Respiratory tract infections.

As mentioned in previous chapters, respiratory tract infection with M. genitalium has been the subject of much controversy. The best evidence stems from the four M. genitalium isolates obtained as mixed cultures from M. pneumoniae isolates (33). In the early years of PCR, de Barbeyrac et al . studied 75 BAL specimens from adult patients with pulmonary infiltrates of unknown aetiology (111). Four (5%) were PCR positive, and three of the patients were at the same time positive in a M. pneumoniae PCR. Most interestingly, all four respiratory tract amplicons contained the E co RI restriction enzyme recognition site found in the G37 type strain, whereas none of eight amplicons from urogenital specimens possessed this site. It would seem highly unlikely, that all respiratory tract strains had one restriction enzyme type as opposed to the urogenital tract strains having another. Unfortunately, no attempts to confirm the findings were reported, whereas all urogenital tract specimens were confirmed with another primer-set (111).

We have used PCR methods to study respiratory tract specimens from different patient populations (114); 50 recruits with pneumonia of whom 14 had evidence of M. pneumoniae infection by serology and/or PCR; 13 M. pneumoniae culture positive patients, 29 patients with respiratory tract infection and a positive M. pneumoniae complement fixation test, but PCR and culture negative. However, no M. genitalium PCR positive patients were identified.

Considering the risk of having mother-to-child transmission of C. trachomatis during birth causing neonatal pneumonia, very little is known about transmission of M. genitalium . In a study of 47 high-risk pregnant women (186), 8 neonates were studied for transmission of urogenital mycoplasmas including M. genitalium . Transmission of M. genitalium from mother to the newborn child was documented in one premature child with acute respiratory distress. However, the data presented did not include control patients and it is not clear how the infants were selected. Neonatal pneumonia is definitely a disease entity that should be studied in the near future.

6.2. Arthritis

Infection with M. genitalium may give rise to symptoms from the joints either because of a direct invasion causing, in principle, a septic arthritis, or it may cause a reactive arthritis, where the pathogen may not necessarily need to be present in the joint. Because of the similarities in disease manifestations with C. trachomatis , it has been speculated that M. genitalium may give rise to sexually acquired reactive arthritis (SARA).

Indirect evidence for M. genitalium being a cause of seronegative arthritis was provided by serology, where it was shown, that five of 10 women had antibodies to M. genitalium as compared to none of the controls (187). Later, two of 13 synovial fluid specimens were found to be M. genitalium PCR positive; in one patient C. trachomatis negative urethritis was documented whereas the other had a chronic polyarthritis (162). The isolation of M. genitalium from a synovial fluid isolate of M. pneumoniae gave further support to a possible association between M. genitalium and arthritis (85; 86). Using PCR, M. genitalium was detected in nine (35%) of 26 specimens from the temporomandibular joint (188). In the same study, C. trachomatis was detected in 42%, but no control tissues were analysed, thus raising questions about the specificity of the laboratory methods used.

6.3. Ocular infection

SARA is to some extent overlapping with Reiter's syndrome. One of the symptoms of Reiter's syndrome is conjunctivitis, and since one of the patients having a positive M. genitalium PCR from synovial fluid had Reiter's syndrome (162), it would have been interesting to test a conjuctival specimen for M. genitalium DNA. Recently a male patient with NCNGU and unilateral conjunctivitis was found to be positive for M. genitalium in both an FVU specimen and a conjunctival specimen (189). Conjunctival symptoms had been ongoing for five months and no relief from local treatment with anti-allergic eye-drops was obtained. He reported intermittent dysuria for several months and was treated with doxycycline 100 mg bid for 10 days and improved. The sequence of the MgPa1-MgPa3 amplicon from both sites were identical suggesting that the eye infection could be a result of self-inoculation

Neonatal conjunctivitis should also be studied further. Considering the low prevalence of M. genitalium among pregnant women, mycoplasmal conjunctivitis may not be numerically important, but since M. genitalium is resistant to chloramphenicol (final MIC 10-25 mg/l) (92) it could have therapeutic implications.