Will there ever be a universal Staphylococcus aureus vaccine?

Abstract

Developing a universal vaccine for S. aureus is a top priority but to date we have only had failures in human clinical trials. Given the plethora of bacterial virulence factors, broad range of the health of humans at-risk for infections, lack of any information regarding immune effectors mediating protection for any manifestation of S. aureus infection and overall competence of this organism as a colonizer, commensal and pathogen, we may just simply have to accept the fact that we will not get a universal vaccine. Antigenic variation is a major challenge for some vaccine targets and for many conserved targets the organism can easily decrease or even eliminate expression to avoid immune effectors without compromise to infectivity and ability to cause disease. Studies of human immune responses similarly have been unable to identify any clear mediators of immunity and data from such studies can only eliminate those found not to be associated with protection or that might serve as a marker for individuals with a higher level of resistance to infection. Animal studies are not predictive of success in humans and unlikely will be except in hindsight if and when we develop an efficacious vaccine. Successful vaccines for other bacteria based on capsular polysaccharides have not worked to date for S. aureus, and laboratory studies combining antibody to the major capsular serotypes and the other S. aureus surface polysaccharide, poly-N-acetyl glucosamine, unexpectedly showed interference not augmentation of immunity. Potential pathways toward vaccine development do exist but for the foreseeable future will be based on empiric approaches derived from laboratory-based in vitro and animal tests and not on inducing a known immune effector that predicts human resistance to infection.

Introduction

To answer the title’s question, we will have to deal with the facts that S. aureus is just too variable in its expression of vaccine target antigens,¹ is capable of infecting a wide range of animal and human tissues and thus able to survive in a wide enough varieties of niches in these hosts such that any selective pressures induced by vaccination can potentially be readily overcome by expansion of existing variants able to escape immune-selective pressures.² While some studies have identified genes commonly found among a large majority of clinical isolates^3,4 no single essential virulence factor needed for infection in most settings that can be targeted as a vaccine is known, exceptions being diseases mediated purely by toxins such as toxic-shock syndrome toxin,^5,6 exfoliative dermatitis and mediators of staphylococcal food poisoning.^7,8 Extensive genetic² and hence antigenic variability in many potential antigens precludes their use as vaccines. Variability in the level of expression leading to a highly variable surfacome^9,10 provides an easy means for bacterial escape from immune effectors by merely reducing levels of antigens to below that needed for elimination or killing of bacteria. S. aureus is also notorious for causing frequent reinfection with the same strain, indicating natural infection does not readily induce acquired immunity that can be defined and used to guide vaccine development. Further difficulties are encountered when using laboratory animals to evaluate S. aureus virulence and immunity as they are sufficiently different in their responses from those that occur in humans meaning that pre-clinical animal tests primarily function as systems of exclusion, used to judge what likely won’t work in humans but unable to predict what will work. Against these challenges we might find that, at best, we can develop vaccines to prevent specific types of S. aureus infections such as bacteremia or skin and soft tissue infections (SSTIs).

In spite of these barriers, many vaccinolgists would place the need for a highly effective vaccine against Staphylococcus aureus in the top 3–5 public health essentials. The organism is among the most frequent causes of infections in virtually all human, and many animal, tissues,^11,12 causes considerable morbidity and mortality^13-16 and community-acquired infections in otherwise healthy people continue unabated and may be increasing.^17-19 So why has it not only been so difficult to develop a vaccine but, to date, numerous trials of a variety of vaccines in humans have all failed? Unfortunately failures provide little informative insight into their basis, as they are usually multi-factorial, and hypotheses about failure are untestable. Yet the failed S. aureus human vaccine trials conducted to date were all backed up by strong pre-clinical data,^20-25 so, at best, we can conclude pre-clinical studies are insufficient to be predictive of success or failure in humans.

Recently, Proctor summarized the challenges for developing a universal S. aureus vaccine²⁶ as have Jansen et al.²⁷ and within these reviews excellent summaries of the attempts to date (Table 1 in Proctor²⁶ and Table 1 in Jansen et al.²⁷) and major challenges (Table 2 in Proctor²⁶) are provided. Therefore there is no need to repeat these except to note the subsequent publication of the results of the Merck V710 (IsdB) vaccine trial in cardiothoracic surgery patients.²⁸ The major point made in these and other reviews of S. aureus vaccine development^26,29-31 is we have insufficient insight into the basis for virulence and immunity for this organism to rationally design vaccines targeting known protective immune effectors. The multitude of genetically variable virulence factors, the undefined mechanisms of host immunity which might indicate the antigens and immune effectors needed to produce an effective vaccine, the high recurrence rates seen in humans, particularly with methicillin-resistant S. aureus (MRSA) infections^32-34 and evidence for inadequate immune responses following infection^35-38 may, in fact, be insurmountable challenges to finding a broadly-effective vaccine for S. aureus. It may be time to consider that these challenges indicate there may not be a means to come up with an effective, broad-spectrum S. aureus vaccine with currently available technologies.

How Hopeless is it?

Difficult to know. Within his reviews,^26,39 Proctor argued extensively that the pathways and mediators induced by vaccines that have been tried and failed in humans might simply elicit the wrong types of immune responses, He proposed a basis for an effective vaccine may be found in new insights regarding the role of interleukin-17 producing T-helper cells (Th17) and cell-mediated immunity in resistance to S. aureus infection. But evidence for this is fairly minimal, with some associations of human immune deficiencies and enhanced susceptibility to S. aureus skin and mucocutaneous infection^40-43 and an observation that humans with B-cell deficiencies are not particularly more susceptible to S. aureus infections than individuals with intact immune systems.^44,45 However, individuals with neutrophil deficiencies do have increased deep-seated S. aureus infections⁴⁶ indicating inefficient opsonic killing contributes to certain types of S. aureus infection. But associations among various observations can easily confound obtaining insights that can be translated into evidenced-based vaccine approaches. As T cell function, particularly T-helper function, is at the core of all acquired immunity, defects in T-helper cells do not necessarily mean increased susceptibility to infection in the setting of T-cell deficiency would exclude a role for antibody-mediated immunity. Th-17 cells are also critical for effective neutrophil-dependent host responses,⁴⁷ so defects in these cells would impair effective recruitment of this key effector of antibody-based immunity. Thus it may be that Th-17 and IL-17 responses may be necessary, but not sufficient, for effective immunity to S. aureus. More to the point, if most immunocompetent humans simply do not make a protective antibody response following either colonization or infection with S. aureus,^48,49 then there is little value in drawing conclusions about the role of antibody in protection when comparing those who can’t (B-cell deficient) and those who don’t (everyone else) make protective antibody. This point was also discussed in detail by Jansen et al.²⁷ So, if finding Th-17-inducing vaccine antigens^50,51 is not the answer, it is not clear what is.

Immune Responses In Initial and Recurrent Infections: Can We Make Some Progress Here?

Analyzing antibody levels and T-cell responses present in acute and convalescent sera is a time-honored means to identify vaccine candidates but there are few such studies for S. aureus infections and all have focused on antibody responses.^35-38,48,49 And with the high recurrence rate of S. aureus infections, many of these responses are either going to be ineffectual at preventing infection or potentially even promote infection by mechanisms such as neutralizing innate immune-activating properties of bacterial factors.^52,53 Of note, in several of these studies^35,36,48 antibody levels to many S. aureus antigens are already elevated in sera taken close to the time when a clinical sample yields a positive culture for S. aureus. This indicates that significant exposure to S. aureus that induced antibody responses had occurred prior to the time of actual microbiologic diagnosis of infection. Importantly, we must consider that the presence of elevated antibody levels early in the course of clinical infection indicates they are either too low or ineffective at protecting against infection. Although no cellular responses were measured, the common finding that antibody levels are already elevated at the onset of clinically-diagnosed infection might also suggest that cell-mediated immunity had also been stimulated but was without effect on preventing infection.

Numerous associations between patient outcomes and levels of antibodies to the S. aureus antigens have been made, but none of these associations indicate any cause and effect relationship between antibody level and protection from infection or reduction in severity of an outcome.^{35-38,48,49,54-57} Furthermore, often overlooked in these analysis is the significant overlap in antibody levels between the susceptible and resistant groups. For example, Adhikari et al.⁵⁷ reported that patients with S. aureus bacteremia who went on to develop sepsis vs. those that did not had lower overall median antibody levels against five S. aureus toxins (α and δ-hemolysins, Panton-Valentine leukocidin (PVL), staphylococcal enterotoxin C-1, and phenol-soluble modulin-α3). Fritz et al.⁵⁸ recently analyzed antibody levels to α-hemolysin and PVL in individuals with primary and recurrent SSTI and systemic infections and found no overall differences in an OD reading between those with primary vs. recurrent infections. While mean or median levels of antibodies to particular antigens might differ between groups that develop sepsis or recurrent infections vs. those that do not, many of the patients with sepsis or recurrent infections had antibody levels to toxins well above the medians in the non-septic/non-recurrent groups. This overlap indicates there is little likelihood that those antibodies were mediators of effective immunity. Also, in the study of Adhikari et al.⁵⁷ all of the patients had S. aureus bacteremia, indicating the higher antibody levels were not protective against this manifestation of infection, and the likelihood of developing a vaccine to prevent S. aureus sepsis and not bacteremia is low. In order to have a meaningful difference in antibody levels associated with resistance to S. aureus infection there needs to be little overlap in titers between those at risk who do not get infected (higher) and those that do get infected (lower). When a significant proportion of the group developing an infection has antibody titers well within the range of those that do not, it highlights the inability of such associative studies to define a mediator or even a marker of immunity. Only in settings where a protective level of antibody or other effector is defined, above which infection is rare such as has been done with the successful capsular polysaccharide vaccines to other pathogens, can meaningful conclusions be drawn about the importance of an antibody titer in relation to protective immunity and vaccine development.

Another group of interest to study are patients that have been infected or even colonized with S. aureus that are at a high risk for a recurrent infection. Among patients with MRSA infections, or even just colonized during hospitalization, re-infection rates approaching 30% occur in the 6–18 mo after discharge,^32-34 and identification of such cases markedly increases when vigorous means to detect post-discharge infections are used.^32,59 Fritz et al.⁵⁸ recently reported a recurrence rate of 62% in closely-followed individuals with primary or recurrent SSTI within 12 mo of the initial episode. Hospital-acquired, community onset infections identify patients potentially ideal for the study of acquired immunity to S. aureus. This is further substantiated by the finding that most of these recurrences are due to the same strain as the initially infecting one,^34,38,59,60 meaning that serologic variation in antigens will not be a major factor in trying to identify protective immune responses, although this could be a problem for designing a universal vaccine if immunity to a single strain is targeted to a highly serologically variable antigen. If one therefore can compare immune responses among immunologically-intact humans with a properly diagnosed S. aureus infection or colonization in the hospital who then get a recurrent or subsequent infection with those that only have a primary infection or do not progress from colonization, insights into immune responses associated with resistance to reinfection might emerge. However, as noted above, the titers associated with becoming infected have to be, for the most part, lower than and not-overlapping to any large degree with the titers in the uninfected controls.

In this context, we examined the occurrence of antibody to the phage-encoded PVL, a major cytotoxin of the highly virulent USA300 and USA400 strains,^61-63 among children with primary or recurrent skin and soft tissue infections (SSTI), the major manifestation of infection by PVL-producing MRSA. Due to its epidemiologic association with these S. aureus strains, PVL has been touted as a potential vaccine, but normal human sera and IVIgG have been found to have significant levels of neutralizing antibody to PVL.^53,64 Among infected children, we found the highest levels of antibody to PVL in those with recurrent SSTI⁵³ clearly indicating these antibodies are not protective and demonstrating there is an association between high levels of antibody to PVL and increased susceptibility to infection. Furthermore, most studies do not find patients with PVL infections do worse than those with non-PVL S. aureus infections^63,65-67 suggesting it is not having a major impact on pathogenesis of human infection or natural immunity is sufficient to defuse its toxic manifestations. Of note, neutralizing antibody to PVL was found in a mouse skin infection model to promote, not inhibit, S. aureus infection⁶⁸ due to the antibody’s ability to inhibit the immune-activating functions of PVL that initiate host innate immunity at low PVL concentrations, such as those encountered early in infection. Along the same lines, stains of S. aureus deleted for the PVL-encoding genes lukS and lukF were more, not less, virulent in a mouse model of pneumonia.^52,69

The recently published study of Fritz et al.⁵⁸ similarly looked at antibody reactivity (defined in the paper as the optical density readings obtained in an ELISA with a single 1:100 serum dilution) to α-hemolysin and PVL in various groups and overall found no differences in acute-phase titers in those with primary vs. recurrent infections and even found a decline in titers in the convalescent sera of those with SSTI but not invasive infection. However, when they examined patients with recurrent infections there were lower mean antibody levels at five time points post primary infection to α-hemolysin compared with those who did not get recurrent infections.⁵⁸ But the overall differences in the mean OD readings were small and one can infer from the reported means and standard deviations that about 80% of the individuals with recurrent infection had a titer comparable to those without recurrent infections. Additionally, paired t-tests were used to analyze the differences in titers between enrollment and convalescent/post-infection sera, but this statistical analysis is not appropriate to apply to this type of longitudinal data, wherein samples were taken from individuals at multiple time periods and all compared with the same initial, pre-infection levels. Paired t-tests assume independence of each measurement and it is clear an individual with a high OD readings at one point does not have the same chance of having a low or high OD reading at the next time point, as antibody levels decline in a predictable manner in humans. Perhaps of greatest importance, the actual toxin-neutralizing activities of the antibodies were not measured to either α-toxin or PVL. As Foletti et al.⁷⁰ showed, humans with the same ELISA titer to α-toxin had as much as a 100-fold difference in their α-hemolysin-neutralizing activity, so without functional characterization of the anti-toxin activity in the sera analyzed by Fritz et al.⁵⁸ it is difficult to accept that any association of neutralization of toxins and resistance to S. aureus infection has been demonstrated. Similarly, although Adhikari et al.⁵⁷ found significant differences in the ELISA binding titers to the individual proteins comprising PVL, LukF (p = 0.02) and LukS (p = 0.01) between patients with S. aureus sepsis and those that did not get septic, the P value for the difference in the toxin-neutralizing titer was only 0.17. Thus, in the study of Fritz et al.⁵⁸ one must also take into account the possibility that patients with recurrent infections might actually have higher neutralizing titers to α-hemolysin than those without recurrent infections, which would indicate a possible negative association of neutralizing antibody to α-hemolysin and infection risk. Given that the Merck V710 vaccine trial²⁸ revealed greater overall adverse events in the immunized group, greater rates of multiorgan failure and higher mortality rates than placebo recipients, the need for full characterization of the functional activity of any immunologic marker is paramount to avoid repetition of such unwanted outcomes from vaccine trials.

Animal Studies-Why They Really Can’t Be Improved For Vaccine Development

Because vaccine results obtained with mice, rats and rabbits do not predict human immunity. . Basically, animal and human susceptibility and the course of S. aureus infections are just too damn different from lab animals, and few other animals are routinely available to study vaccines. At most, lab animal studies serve as systems of exclusion, possibly defining what might not work. But even negative results in animals cannot predict human responses. When developing the meningococcal serogroup vaccines in the 1960s the purified capsular polysaccharides were not immunogenic in mice, rhesus monkeys, chimpanzees and only one of 5 gibbons made an antibody response following vaccination.⁷¹ But the investigators knew humans could make antibody from natural exposure.^72,73 Importantly, the purified meningococcal serogroup A and C capsular antigens were immunogenic in nearly 100% of human volunteers⁷⁴ so if one used animal results to conclude meningococcal polysaccharides were not good vaccine candidates we might not have benefited from this remarkable advance that ended the epidemic of meningococcal meningitis in military recruit camps.^75,76 We just have to admit lab animals are “furry test tubes” and at best can be used to test hypothesis about in vivo activity of immune effectors such as functionality, the need for co-factors such as white blood cells, complement, Th17 cells, etc., and determine if a vaccine target is expressed in vivo in lab animals.

A major logical reason animal models cannot be improved to guide vaccine development is that we will only know which models are associated with human vaccine efficacy when we actually have an efficacious vaccine. As all of the failed S. aureus vaccines tested to date in humans^28,77-79 have shown reductions in bacterial burdens and even protection from lethality^21,23,80-82 in many animal models, it is clear these models have no predictive capacity for effectiveness in humans. Unless an efficacious S. aureus vaccine is developed for humans that has been tested in an animal model that would also show a failure of immunity to CP, ClfA, and IsdB, then the likelihood of finding an improved animal model is remote. Furthermore, many human manifestations of S. aureus infections induce symptoms that can be identified and communicated to a physician, this is not possible in animals. For example, someone awakening with a painful hand or finger who seeks medical care that turns out to be an S. aureus joint infection identifies a potentially serious situation where the individual is at risk for a significant compromise to the functionality of their hand. This, of course, could never be modeled in an animal, we have to infect them with a sufficient level of bacteria to get frank infections in most of the animals and these levels are usually many orders of magnitude above those that initiate infections in humans. At best one can hope that significant reductions in bacterial levels or other measures of infection in immunized animals are reasonable tests for potential vaccine efficacy without relying on the results to predict outcomes in a human vaccine trial.

Additionally, if we truly evaluate the results from animal infection models we often see in many studies that there is a reduction in bacterial levels, but in reality most, if not all, of the animals are infected.^83-85 It might not matter in a human S. aureus vaccine trial if there are fewer bacteria in infected tissues of vaccinated patients, if they are still diagnosed with an infection based on a positive culture resulting from even low numbers of bacteria in the infected tissue it would be indicative of vaccine failure unless the reductions in bacterial burdens were accompanied by a significant effect on a measurable clinical outcome that had also been identified as an endpoint for the clinical trial.

Lethality studies in animals might be more predictive of efficacy in humans, and prevention of lethality would be an acceptable and readily identified endpoint for a clinical trial. But doses of S. aureus needed to induce lethality in lab animals are often quite high and lethality may not be due to bacterial burdens but rather acute toxicity from the bolus injection of these high infectious doses. It seems unlikely these challenge doses are representative of what initiates human infections with S. aureus. Furthermore, vaccines against CP antigens,²³ immunity to ClfA⁸² and vaccination against IsdB^80,81 have all shown protection from lethality in mice but these vaccines have all failed in human trials, so we can already conclude lethality models are not stringent tests for identifying vaccines likely to succeed.

Logical Assumptions and Historic Vaccine Success-Are They Indicators of a Path Toward Rationale Vaccine Design for S. aureus?

In spite of the lack of data associating any specific human immune response with protective immunity to S. aureus, some logical assumptions about potential vaccine candidates can emerge based on principles from other pathogens, notably the effectiveness of vaccines targeting capsular polysaccharides of Streptococcus pneumoniae,⁸⁶ Hemophilus influenzae type b,⁸⁷ Neisseria meningitidis⁸⁸ and Salmonella enterica serovar typhi.⁸⁹ The presence of capsule-specific immunity in resistant human populations was the basis for developing these successful vaccines, and thus one might predict analyzing antibody responses to S. aureus capsules could be highly informative. S. aureus strains can expresses either one of two capsular polysaccharides (CP), CP5 or CP8,^29,90,91 or neither, along with the poly-N-acetyl glucosamine (PNAG) surface polysaccharide antigen.^85,92,93 CP5 is produced by about 30% of strains, CP8 by about 50%^94-96 and, as best as we can tell from our own immunologically-based studies,⁹² PNAG can be detected on the surface of nearly 100% of S. aureus clinical isolates. Near universal PNAG expression among invasive clinical isolates is also supported by genetic evidence^3,4 indicating the presence of the intercellular-adhesin (ica) genes encoding the PNAG biosynthetic apparatus in almost all invasive S. aureus strains. An antigen designated 336 has been proposed as another S. aureus capsule, but it is, in fact, a cell wall teichoic acid antigen and not a true capsular polysaccharide.⁹⁷ On in vitro grown S. aureus, the CP antigens and PNAG are both co-expressed on the cell surface as intercalated molecules (Fig. 1), with significant overlap in the pixels visualized by confocal microscopy indicative of the presence of both CP and PNAG antigens in close proximity on the bacteria (Fig. 2). These results also indicate there is comparable availability of these antigens to mediators of immunity. Thus, in contrast to the description of Jansen et al.²⁷ in Table 1 of their review that PNAG is a biofilm antigen into which antibodies and immune effectors might have trouble penetrating, PNAG is, in fact, a capsular antigen like CP5 and CP8 and passive and active immunization against PNAG will most likely target infections wherein the bacteria are in the planktonic, not the biofilm, state.

Figure 1. Reactivity of S. aureus CP8 strain MN8 and CP5 strain Reynolds to antibody to either the homologous CP antigen (Anti-CP), PNAG (Anti-PNAG) or both as detected by immunofluorescence. NRS = normal rabbit serum. Binding of rabbit antibody to CP5 or CP8 to S. aureus cells detected with anti-rabbit IgG secondary antibody conjugated to AlexaFluor (AF) 588 (red). Human mAb F598 to PNAG directly conjugated to AF 488 (Green) was used to detect PNAG on the bacterial surface. Co-localization of red and green pixels in samples reacted with antibody to both CP and PNAG antigens appears as an orange-yellow to yellow color. Far right panels show fluorescence in the individual red (Anti-CP), green (Anti-PNAG) or both channels for selected bacterial cells in the micrograph co-stained with antibody to CP and PNAG antigens. White bars = 10 µm.

Figure 2. Quantification of the fluorescence intensity of the reactivity of S. aureus CP5 strain Reynolds and CP8 strain MN8 to antibody to the homologous CP antigen and PNAG. Binding of primary rabbit antibody to purified CP5 or CP8 conjugate antigens to S. aureus cells detected with anti-rabbit IgG secondary antibody conjugated to AlexaFluor (AF) 588 (red). Human mAb F598 to PNAG directly conjugated to AF 488 (Green) was used to detect PNAG on the bacterial surface. Histograms depict analysis of the co-localization of red and green pixels in samples reacted with antibody to both CP and PNAG antigens across the distances, in microns (µm), depicted on the X-axis. Arrows on photomicrographs indicate regions analyzed.

Of note, in the reports on immune responses in various populations given CP-conjugate vaccines it appeared that natural antibody to CP5 and CP8 present in the pre-immunization sera was low, (< 10 µg/ml^22,98-100) suggesting that most humans do not make much of an antibody response to CP5 or CP8 from natural exposure. In some contexts this would be encouraging, as it would indicate that an effective CP-specific vaccine could provide immunity not engendered by natural exposure. However, from the epidemiology of S. aureus infections it is clear that CP antigens are not essential virulence factors, as CP-non-expressing strains comprise 10–20% of clinical isolates, and the highly virulent USA300 clone of MRSA, increasingly found as a cause of both community- and hospital-acquired infections,¹⁰¹ does not make either CP5 or CP8.¹⁰² As non-essential virulence factors it may be that in the presence of antibody to CP antigens the infecting S. aureus strains phase-vary and stop making them without any loss of virulence. This concept is supported by the frequent recovery of isolates from human infections that have mutated the accessory-gene-response (agr) system^103-106 needed for maximal CP production.¹⁰⁷ Loss of agr-facilitated CP synthesis could allow for escape from antibody-mediated opsonic killing and such variation during infection is deserving of further investigation.

Might PNAG then serve as a conserved, broadly expressed vaccine target present of the surface as are the capsules of pathogens for which successful vaccines have been developed? Over the past 15 y PNAG has been extensively evaluated in pre-clinical settings as a S. aureus vaccine candidate.^{85,92,108-110} Overall, strong evidence for both protective immunity and function as a virulence factor in settings of experimental S. aureus infection has been obtained,^85,108-111 which, while encouraging, still represent studies comparable to those for the S. aureus vaccines that have gone before and failed in human trials. A key breakthrough in the development of PNAG vaccines was the finding that immunity elicited to the native glycoform of the antigen, in which the vast majority of the amino groups on the second carbon of the N-acetyl glucosamine monomers are acetylated, does not elicit protective antibody.¹¹² Eliminating most of the acetyl substituents results in a glycoform termed deacetylated PNAG (dPNAG) that readily elicits opsonic, protective antibody.^84,85,113 The protective antibodies effectively deposit complement opsonins onto the bacterial surface.¹¹⁰ Natural antibody to the native glycoform of PNAG found in all human sera we have analyzed are non-opsonic and non-protective in 95% of the samples,^35,36 indicating a basis for the escape of S. aureus and other PNAG-producing pathogens from natural immunity to this antigen.

Under these circumstances, wherein natural antibody to both the CP and PNAG surface antigens are either low or ineffective, it seemed quite promising to combine these into a multi-component vaccine that could potentially cover all strains of S. aureus. When we attempted to determine the additive and/or synergistic effects of having both antibody to CP and PNAG antigens in an immune serum, we were completely surprised to find they inhibited each other’s opsonic and protective activity when present in serum at specific ratios.^35,36,108 Inhibition was due to a charge-dependent idiotype-anti-idiotype binding of the variable regions of the antibodies to the negatively-charged CP and positively charged, dPNAG to each other. Inhibition was not present if antibody to either CP or PNAG antigens predominated in a serum³⁶ but in humans this could be a transient phenomenon, as not only did vaccine induced antibody to PNAG inhibit CP antibody functionality,³⁶ but the natural, non-opsonic, non-protective antibodies to PNAG present in almost all normal human sera could also inhibit vaccine-induced antibody to CP antigens.³⁵ Thus, once an initial spike in antibody to CP antigens induced by immunization declines over time, the predominant effect in human sera could be loss of functional activity of the antibody to S. aureus CP antigens.

Looking at development of antibody to CP and PNAG antigens in infected human sera provided further insight into the incompatible nature of antibody to both antigens being present in one serum. We found that among S. aureus infected humans, only those with bacteremia made opsonic antibody responses to either CP or PNAG antigens, but, in the majority of these cases, the ratios of antibodies were in the inhibitory range.^35,36 To detect the CP- or PNAG-specific opsonic killing activity, we had to remove one or the other of these antibodies from the sera by absorption, but when combined back together the opsonic killing manifest in the absorbed sera were restored to the initial non-opsonic state. Thus, the problem was not in induction of a potentially protective opsonic response to either CP or PNAG antigens in patients heavily infected with S. aureus, the problem was the inability of these two effectors to peacefully co-exist. This inhibitory phenomenon may also have contributed to the failure of the CP5/CP8 conjugate vaccine trial in hemodialysis patients⁹⁹ wherein a significant 57% (95% C.I. 10–81%) reduction in S. aureus infections among vaccinates was detected at 40 weeks post-immunization but not at 54 weeks, the pre-determined trial endpoint. From the 40–54 week period the decline in antibody levels to the CP antigens could have brought them into the inhibitory range, thus turning an efficacious signal at 40 weeks into a non-efficacious one 14 weeks later. A repeat trial of the CP5/8 conjugate vaccine in hemodialysis patients failed to show efficacy at any time point, and while no published data are available for analysis, it was suggested at meeting presentations that there was a manufacturing problem leading to less-than-expected immunogenicity of the batch of vaccine used in the repeat trial. Such a situation could also exacerbate the potential for natural antibody to PNAG to be in the inhibitory range for the immunization-induced antibody to CP antigens. Overall, another barrier to an otherwise theoretically promising approach to a S. aureus vaccine emerged and these results likely preclude developing any multi-component vaccine containing both CP and PNAG antigens.

What about inducing immunity to PNAG—would natural antibody to CP antigens be similarly inhibitory? This possibility does need to be evaluated but, as noted above, studies of antibody levels to the CP antigens in pre-immunization sera from conjugate-vaccine volunteers^22,99,100 and normal humans³⁵ indicates little natural antibody to S. aureus CP antigens arises from natural exposure. Also, we have reported that titers to CP5 and CP8 are markedly lower in most normal human sera when compared with titers to PNAG in the same sera.³⁵ Of note, in a phase I safety and pharmacokinetic evaluation of a fully human IgG1 mAb to PNAG opsonic antibody to S. aureus was detected in all recipients within hours of infusion and maintained for 50 d,¹¹⁴ indicating over this time period no interference in functional activity emerged.

As of now, the development of vaccines and passive therapies targeting PNAG are progressing but whether they will be successful or meet the same fate as S. aureus human vaccines that have gone before them won’t be known for several years at a minimum. The fully human mAb is being evaluated to determine what phase II trial would be best for obtaining a potential efficacy signal as well as additional safety and pharmacokinetic data in individuals at-risk for S. aureus infection. The oligoglucosamine-conjugate vaccine is being synthesized for phase I trials projected to commence in 2014 and potentially earlier in economically-valuable animals. Impacting the development of PNAG-targeting vaccines is the finding that not only S. aureus and S. epidermidis make PNAG but gram-negative bacteria carrying a biosynthetic genetic locus termed pga can make PNAG.^84,115-119 Surprisingly, we have recently identified a large number of major human bacterial pathogens that lack a 4-gene operon homologous to either the staphylococcal ica or gram-negative pga loci that make PNAG, and additionally found that fungal and eukaryotic parasites such as Candida albicans, Trichomonas vaginalis and Plasmodium falciparum make PNAG.¹²⁰ How this will guide and impact the development of a PNAG vaccine or potentially even affect clinical trials of CP vaccines that could be incompatible with a PNAG vaccine is not yet known, but represent important questions that will affect S. aureus vaccine development as well.

The Challenges Coming From Finding An Effective Clinical Setting For Testing a Vaccine

The question of how to identify a clinical setting or population to provide the results needed to validate the efficacy of a S. aureus vaccine is not only lacking a clear answer but it is a moving target. The CP vaccine trials in hemodialysis patients were chosen for the anticipated high infection rates experienced by this population.^78,99,121 Immunization of individuals undergoing cardiothoracic surgery for the evaluation of the V710/IsdB vaccine represented a clinical issue with a high medical need for an effective intervention as well as a setting where the at-risk individuals could be identified sufficiently ahead of time to allow for vaccination to proceed prior to surgery.²⁸ Trials of passive therapy targeting ClfA and LTA in low-birth weigh neonates also represented a setting of high infection rates and medical need.^122,123 None of these patient populations are ideal for vaccine/passive therapy evaluations but the choice is driven by numerous factors and it is unlikely there is an ideal population consisting of immunocompetent individuals that can respond to a vaccine that also have a high risk for S. aureus infection.

Further impacting clinical trial design is the changing nature of patient care that evolves as the vaccine trials are being planned and populations evaluated for basal S. aureus infection rates in order to ascertain that a sufficient signal can be garnered in a reasonable time period. Improvements leading to reductions in nosocomial infections such as better hand hygiene, surgical wound care,^124-126 infection control including contact precautions with single room isolation or cohorting¹²⁷ and introduction of interventions such as routine use of chlorhexidine baths¹²⁸ can effectively reduce infection rates with S. aureus, particularly when applied to MRSA. Thus, what might look like a good population to conduct a S. aureus vaccine trial in at the start of the trial might have marked reductions in infection rates as the trial progresses, confounding the ability to get a robust signal as to a vaccine candidate’s efficacy.

One population that emerges as a potentially robust target group are patients with a significant exposure to a health-care setting that are at high risk for community onset S. aureus infections, notably bacteremia and osteomyelitis caused by MRSA within 12 mo of the exposure.^32-34,59,60 This indicates a reasonable expectation of sufficient levels of infection such that a rationale clinical trial can be designed and implemented. Additionally, this is a heterogeneous group of individuals but they are identified during their hospitalization by an MRSA-positive culture from a superficial body site,³⁴ so enrolling volunteers is facilitated by the availability of hospital-based cultures. Furthermore, although many of these individuals have significant underlying disease leading to an immunodeficient state, or have an underlying rapidly fatal condition, they can be excluded from early vaccine trials. Hemodialysis patients represent about 20% of this at-risk group³⁴ and their ability to respond to or effectively utilize a vaccine-induced immune effector could be problematic, although in the CP vaccine trials in this poplation immunogenicity of the conjugate vaccines did not appear to be a problem in the published studies.^22,99,100 Evaluations of immune responses and testing of all effectors needed for vaccine efficacy, such as phagocyte and complement function, in the blood of hemodialysis patients should be undertaken in initial phase II trials to insure that efficacy won’t be impacted by sub-optimal function or availability of needed co-factors or rapid loss of antibody due to dialysis. Importantly, the individuals at risk for community onset MRSA infections often are not discharged with a standard of care that includes such treatments as routine antibiotic administration, so potential treatments that could reduce the occurrence of infection among controls is minimized, although encouraging optimal practices such as chlorhexidine baths in such populations might impact infection rates. Routine monitoring by visits from home-health aides and keeping close track of post-discharge medical care³² should insure maximal determination of infection rates among vaccinees in this group. Better epidemiological studies of this population might be needed to substantiate that there is a sufficient number of relatively immunologically-intact individuals that could enroll in a S. aureus vaccine trial, but if neither numbers of available volunteers or other confounding or contraindicating factors are found among this post-discharge, MRSA group they likely represent a good cohort to evaluate in any S. aureus vaccine approach.

Potential Ways Forward

With all of the barriers, failures and difficulties encountered to date on developing a universal vaccine for S. aureus we would nonetheless like to find some means to either target the more severe manifestations of infection such as bacteremia, sepsis, bone infections and pneumoniae or prevent infections that are difficult to manage such as implant-related infections. Thus, instead of a universal S. aureus vaccine we may need to settle for one targeting specific clinical manifestations. Many individuals in the field advocate a multi-valent vaccine approach,^38,129 which, while logical and could even cover different manifestations of S. aureus infection, the idea is still not predicated on any human data, only results from pre-clinical animal studies.^81,108,129 Given the cost of drug development it is unlikely a systematic approach wherein human testing of both individual and combined vaccine components will ensue, so we might end up with an effective multi-component vaccine without any real knowledge of which of the antigens are essential. PNAG remains a potential single component vaccine if it induces infection-preventing immunity in humans to S. aureus as do capsular vaccines for other bacterial pathogens, but at the moment there are no data regarding susceptibility and resistance of humans to infection based on their PNAG antibody status. One encouraging finding in regards to PNAG expression among the variety of microbial pathogens that make this antigen is that limited, but consistent, findings indicate in vivo production during human infection.¹²⁰ Of note, if immunity to PNAG can contribute to resistance to infection with S. aureus and other organisms, it is possible that a vaccine inducing immunity to this antigen could be widely used but augmentation of immunity specific to antigens from different organisms, such as α-hemolysin for S. aureus,¹²⁰ shiga toxin for organisms producing this factor^130,131 and Clostridium difficle toxins¹³² might be needed to engender maximal immunity to each pathogen. Overall, while the path to an effective human vaccine for S. aureus is so far strewn with failures and is not guided by any strong association between human immune effectors and resistance to infection, there is still a major effort and considerable investment within the biotechnology and pharmaceutical industry for pursuing active and passive therapies, engendering hope, but a cautious level of confidence, that we will ultimately be successful in this endeavor and be able to prevent at least some of the more severe or costly S. aureus infections.

Conflict of Interest

GBP is an inventor of Intellectual Properties (IP) (Human monoclonal antibody to PNAG and PNAG vaccine) that are licensed by Brigham and Women’s Hospital (BWH) to Alopexx Vaccines LLC, and Alopexx Pharmaceuticals LLC. As an inventor of the IP, he has the right to receive a share of licensing-related income (royalties, fees) through BWH from Alopexx Pharmaceuticals and Alopexx Vaccines. GBP also holds equity in these two companies. GBP’s interests were reviewed and are managed by the BWH and Partners Healthcare in accordance with their conflict of interest policies.