The relationship between mucosal immunity, nasopharyngeal carriage, asymptomatic transmission and the resurgence of Bordetella pertussis

Mucosal immunity and its link to herd immunity against respiratory pathogens

For the purposes of this manuscript, we define “mucosal immunity” as any process that leads to a protective immune response on a mucosal surface, even if the origin of that response may not reside locally within the mucosa itself. This definition therefore includes secreted antibody (IgA or IgG) that may derive from non-mucosal B-lymphocytes elsewhere in the body but which serve to interdict mucosal pathogens remotely.

Protein–polysaccharide conjugate vaccines targeting the trio of encapsulated respiratory pathogens HiB, Streptococcus pneumoniae, and Neisseria meningitidis have provided important insights into the mechanisms of vaccine-derived herd immunity^17–23. In each case, the vaccines yielded steep reductions in disease among vaccinated and unvaccinated individuals, i.e. indirect herd effects, and, for each, the critical mechanism yielding herd effects proved to be a reduction in NP colonization rates, in turn mediated by sterilizing mucosal immunity^23–27. To note, modeling studies suggest that at least half of the overall benefits of these vaccines rest on indirect, rather than direct, protection²⁸. In other words, herd immunity is not an incidental bonus feature of these vaccines; it is why they are so effective.

Obviously, B. pertussis disease differs substantially from disease caused by HiB, pneumococcus, or meningococcus, whose pathogenesis is mediated by inflammation and/or the interaction between the pathogen and biological systems within the host (e.g. complement and coagulation cascades and inflammatory pathways). By contrast, pertussis deals death at a distance through the elaboration of multiple toxins whose contributions to the pathogenesis of classical pertussis disease are complex and remain only partially understood^29–31. Nonetheless, it is reasonable to postulate that pertussis' ability to adhere to, infect, and persist on the human respiratory mucosa could be an important factor in the overall epidemiology of pertussis. This theory seems supported by recent experimental mouse model data showing that B. pertussis can form biofilms that allow it to adhere to abiotic surfaces as well as to murine nasal and tracheal mucosa³². Several cross-sectional studies in humans have also presented data compatible with the hypothesis of asymptomatic infection. But, while suggestive, these are not definitive proof for the simple fact that pertussis, as with all infectious diseases, has an incubation period. Without longitudinal follow up, there is no way to know whether an asymptomatic infected individual sampled “today” was not destined to have become symptomatic in coming days^33,34.

Failure of traditional explanations to account for the resurgence of pertussis

Compared to the pre-vaccine era, wP vaccines exerted a profound (~99%) reduction in pertussis incidence (Figure 1)³⁵. But the newer-generation aP vaccines have been disappointing given that pertussis incidence has increased in many of the countries (though not all) where they replaced wP vaccines, including the US, Canada, Australia, Spain, Ireland, and the United Kingdom^36–40. The key question is why.

Several hypotheses have been raised, of which the first three can be discounted:

1.  Detection bias: traditional culture-based methods for diagnosing B. pertussis are insensitive because of the fastidious nature of the organism and because exposure to antibiotics can yield false negative results^41,42. Such limitations could be overcome by using serologic markers of pertussis infection (such as a rise in anti-pertussis toxin [PTx] antibodies) or through the use of PCR, which may remain positive for weeks after the start of effective antibiotic therapy⁴³. The argument that PCR could result in an artefactual increase in incidence is intuitively appealing and could also explain the increased detection of milder/atypical presentations of pertussis^44,45. Serologic markers are more difficult to interpret, since rises in titer may result from boosting after natural exposure absent disease or exposure to cross-reacting antigens.

Regardless, improved ascertainment does not adequately explain the increases in severe pertussis and deaths reported among infants. It has been argued that severe or fatal disease in young infants should be relatively robust against detection bias, being inherently unlikely to evade diagnosis and reporting. Furthermore, the timing of the rising incidence is difficult to reconcile. For example, in the US, infant pertussis rates had been slowly increasing since the 1970s, preceding the switch to enhanced diagnostics¹¹. Conversely, the UK’s Health Protection Agency shifted to PCR and serologic diagnostics in 2006 (Figure 3)^46,47. In the years 2007–2009, the total number of cases reported increased modestly from ~100–300/year to 600–800/year, coincident and possibly explained by the phase in enhanced diagnostics. But that pattern shifted abruptly in 2011–2012 with over 1,000 and 9,000 cases in each year, respectively. The degree and rapidity of that increase is not readily explained by a phased adoption of new diagnostic techniques. Rather, it is more suggestive of a multi-year nationwide outbreak of pertussis that peaked in 2012. Enhanced diagnostics, and particularly PCR, may explain some increase in detection and very likely explains some of the shift towards the detection of milder cases, but the increase in pertussis incidence is real, not artefactual.

Figure 3. Pertussis incidence in the United Kingdom, by age group, 1998–2015.

This figure presents data from the United Kingdom of pertussis incidence disaggregated by age categories over a 17-year period. The introduction of enhanced pertussis diagnostics and of the transition from whole cell to acellular (aP) pertussis vaccines are noted in annotations. To note, the scale representing incidence differs by age category (≥6 months versus <6 months of age). While there is year to year fluctuation in the incidence of pertussis, during the period 2011–2013, incidence rates spike. This abrupt increase cannot be easily explained as an artefact of enhanced diagnostics that were being phased in since 2006. A more parsimonious explanation is that this represents a nationwide epidemic of pertussis occurring after some delay from the transition to aP vaccines in 2005. PCR, polymerase chain reaction.

2.   Poor vaccine coverage: aP vaccines were first introduced in the US in 1991 at limited scale for the fourth or fifth doses of an otherwise wP vaccine series, switching to the fully aP series in 1996. According to 2015 state by state statistics from the CDC, among US toddlers aged 19–35 months between 91 and 99% received three or more doses of DTaP and between 76 and 92% had received four or more doses⁴⁸; the introduction of the Tdap boosters to adolescents, older adults, and most recently pregnant women has further expanded coverage⁴⁹. The rise in pertussis in the US occurred despite historically high uptake of pertussis vaccines.

3.   Vaccine refusals: as with measles, outbreaks of pertussis in unvaccinated or under-vaccinated populations have been well documented⁵⁰. Moreover, pertussis incidence increased among communities with high vaccine refusal rates, even among those who had been fully vaccinated, suggesting that the unvaccinated/under-vaccinated individuals pose a risk to the fully vaccinated members of those communities⁵¹. Yet, given that US pertussis vaccine uptake is so high on average, the impact of vaccine-refusing communities must, by definition, be small and largely limited to those communities. It is difficult to see how this could significantly affect the overall increases seen across the general population in the US. Under-vaccination is an important factor in localized outbreaks but is insufficient to explain pertussis’ general and dispersed resurgence.

Evidence for two other hypotheses for the pertussis resurgence are more plausible:

4.   Poor persistence of anti-pertussis antibodies following aP vaccination: many studies have documented rapid declines in pertussis antibodies, often to baseline concentrations, within as few as 2–3 years of the last aP vaccination (DTaP or Tdap), and these declines coincide with falling clinical efficacy^52–56. Klein et al. estimated a 42% increase in the odds of acquiring clinical pertussis disease per year since the fifth dose of DTaP⁵⁷. Similarly, a meta-analysis by McGirr and Fisman estimated that the odds of pertussis disease increased by 33% for each year after the last aP vaccination, such that only 10% of fully vaccinated infants would remain protected after a median of 8.5 years⁵⁸.

Such data argue that poor immunologic persistence is an important factor in the resurgence of pertussis. However, it seems unlikely that all of this can be explained by that factor. Logically, with waning immunity, one would expect to see the rise proportional to increasing age groups, with the rise highest in adolescents, lower in school-aged children and toddlers, and lowest of all among vaccinated infants, which is what the UK data show (Figure 3)^46,47. The increases are bimodal, with the largest increases in incidence occurring among individuals aged 10 years and over and among infants aged <3 months. The rise among adolescents is well explained by poor immunological persistence after vaccination. By contrast, the increased incidence among infants too young to have been vaccinated is better explained by an overall rise in pertussis transmission, which may emanate from those older age groups. Immunologic persistence of protection elicited by aP vaccines has proven poorer than anticipated and very likely contributes to the resurgence of pertussis among older age groups. It is also likely that these older children act as a reservoir of infection that contributes to pertussis disease among very young infants.

5.   Leaky efficacy due to evolutionary shifts: in contrast to wP vaccines, which expose the human immune system to a wide complement of B. pertussis antigens, aP vaccines include only a few or even one (i.e. pertussis toxoid-only vaccines) highly standardized antigen/s. While far less reactogenic, a potential tradeoff is that the vaccine is less adaptable to the plasticity of the B. pertussis genome.

Bart and colleagues recently published a series of phylogenetic analyses among global pertussis isolates collected spanning the pre-vaccine era, the wP vaccine era, and the aP vaccine era^59,60. The introduction of aP vaccines, and to a lesser degree wP vaccines, appeared to coincide with shifts away from the specific allelic isoforms of the genes coding for most of the aP vaccine antigens. This included Fim2 and Fim3, PRN, and PTx. In each case, there was a shift to allelic isoforms genetically distinct from those contained in the aP vaccine (e.g. Fim 3-1 to Fim 3-2 and 3-3; PRN1 to PRN2; and PTxA2 to PTxA1 [the aP vaccine alleles are shown in bold font]). In addition, the predominant promoter gene for PTx shifted from PtxP1 to PtxP3, notable because PTxP3 hyper-expresses PTx relative to PTxP1, a further strategy for immune evasion simply by increasing the concentration of the antigenic target. Lastly, multiple studies have identified disease-causing pertussis isolates that ceased expressing one or more of the aP antigen genes entirely, a definitive strategy for evading antibodies targeting these proteins. This includes, singly or in combination, PRN, FHA, and, more recently, PTx^61–64. Strains expressing PRN have virtually disappeared from the US^65,66. While rare and dissociated from the profound lymphocytosis often seen in infant pertussis and mediated by PTx specifically⁶¹, PTx-minus strains are surprising given prior assumptions that PTx was an obligatory virulence factor⁶⁷.

But, while provocative, the degree to which these shifts contributed to the pertussis resurgence is unclear. This uncertainty is justified by the continued effectiveness of aP vaccines in a number of countries where non-vaccine alleles have become quite common. For example, Sweden’s pertussis toxoid-only vaccine has retained efficacy for over 20 years^68,69. It is also true that pertussis rates were slowly trending up in the US since the 1970s preceding the switch from wP to aP vaccines, though this has accelerated dramatically in the aP vaccine era (Figure 1).

Another problem with this explanation is that the rise in population incidence observed in multiple countries seems to occur roughly 5–10 years after switching to aP vaccines and that the increases in incidence are often quite abrupt (see Figure 2). Presumably, the accumulation of non-vaccine antigen alleles among circulating strains of B. pertussis would be a gradual process, resulting in an incremental loss of efficacy with time. These abrupt step-wise transitions conflict with that expectation and do not account for why these seem to occur after a stereotypical delay from the wP to aP switch.

A good example of this leverages the natural experiment afforded by the US’ switch from wP to aP vaccines between 1991 and 1996. Klein et al. examined the incidence of PCR-confirmed pertussis illnesses as a function of age, which provided the natural experiment: anyone under the age of 11 years would mainly have received aP vaccines, those over the age of 15 could have received only wP vaccines, and those in between would have received a blended schedule of wP and aP vaccines. As shown in Figure 4, the incidence of pertussis increased with advancing age through 10 years, a trend well explained by waning immunity. But then incidence drops precipitously among the older cohorts who had received partly or fully wP vaccine schedules as infants, falling nearly to zero incidence among those 15 years and older. These data argue that wP vaccines were far more effective than aP vaccines (overall far lower incidence) and that they induced protective efficacy for significantly longer (i.e. beyond 15 years). Such striking differences also raise the following question: do the two classes of vaccine work immunologically in the same ways? We will return to this issue subsequently. Evolutionary shifts are likely in response to aP vaccines (and to a lesser degree wP vaccines) and plausibly contribute to the pertussis resurgence by degrading the efficacy of certain aP vaccines. But the extent of contribution is uncertain, and the abrupt rises in incidence do not fit well with evolution as the driving factor.

Figure 4. A natural experiment exploring the effect of infant exposure to acellular or whole cell pertussis vaccines and age-specific incidence of pertussis

These data from the Kaiser Permanente health maintenance organization in Northern California reflect a natural experiment centered on the transition from whole cell to acellular pertussis vaccines in the US in the 1990s. These data were all collected over a narrow time window, and so the incidence by age should be understood in relation to what vaccine these individuals had received as infants. Those above the age of 15 could have received only whole cell pertussis vaccines, those under the age of 11 could have received only acellular pertussis vaccines, and those aged in between could have received a blended schedule of both vaccines, reflecting the transition period. The blue line depicts pertussis incidence as a function of age; the red line represents the proportion of each age stratum that received the acellular vaccine (100% of those vaccinated below the age of 11; ~0% of those aged 15 years and older). We make several observations: First, there is a decline in incidence from birth through the first year of life. This is likely explained by the induction of immunity through infant vaccination. Pertussis incidence then increases steadily through age 10. In light of what has been learned, this most likely reflects waning of acellular pertussis vaccine-induced immunity with time. Second, rather than continuing to rise with age, the incidence instead plummets above the age of 11 years, falling essentially to zero in those 15 years and older. It is remarkable to observe that receipt of any whole cell pertussis vaccines during infancy continues to exert such durable protection, to the extent that the 15 years and older birth cohort remain almost completely protected even decades out. This strongly argues that the immunologic effects resulting from whole cell and acellular pertussis vaccines are quite distinct. DTaP, diphtheria, tetanus, and acellular pertussis.

Novel insights into pertussis transmission from modeling studies

In summary, persistence and evolutionary shifts likely contribute to the pertussis resurgence but seem insufficient to explain current disease trends. Something is still missing.

It has been argued that wP vaccines prevent disease but do not block infections. This conclusion rests on a single influential study published by Fine and Clarkson in the Lancet in 1982⁷⁰. Their core argument concerned changes in the time period between pertussis epidemics and whether these expanded during the wP vaccine era (evidence that transmission was being impeded) (see examples in Figure 5). They found some lengthening of the inter-epidemic period but deemed this insufficient to suggest a decline in transmission and so concluded that wP vaccines did not prevent infections.

Figure 5. Effect of infection-blocking versus disease-preventing vaccines on the inter-epidemic cycle lengths of a hypothetical respiratory disease.

These cartoons depict the effect on inter-epidemic cycle lengths by vaccines with different immunological effects in terms of whether they prevent infections (regardless of symptoms) as well as clinical disease (by definition symptomatic) or that only prevent clinical disease. At a population level, immunity waxes and wanes over time. With increased infection rates, the population acquires immunity and disease incidence subsequently falls. Over time, population immunity wanes (for example, because of the introduction of non-immune infants born into the population), leading to a resurgence of disease. The inter-epidemic cycle length is the average time between peaks in this cycle. The amplitude of each peak reflects the number of observed symptomatic cases in the population.

  •  Panel A depicts the base case absent any vaccination. For pertussis absent vaccination, the inter-epidemic cycle length has been estimated at between 3 and 5 years.

  •  Panel B depicts the effect of a vaccine that prevents infections and clinical disease. By blocking infections, the pace of spread through the population is slowed, leading to an extension of the inter-epidemic cycle length as well as a decline in amplitude of peaks due to prevention of clinical disease.

  •  Panel C depicts a vaccine that fails to block infection but effectively prevents clinical disease. This vaccine only reduces the amplitude of peaks but has no impact on period length, since transmission is not affected.

This conclusion is oft-cited and well engrained in the literature and official texts^71,72 but is arguably flawed. Co-author Rohani and his team re-examined period lengths across multiple cities in England and Wales before and after wP vaccine introduction, with each city serving as its own control⁷³. When the data were aligned in this way, lengthening of inter-epidemic periods became more apparent, with average increases in each city of 1.5–2.5 years. This observation has since been replicated in 64 countries^10,74,75. In our view, the preponderance of evidence argues that wP vaccines impede disease transmission as well as prevent clinical disease⁷⁶.

This issue was readdressed by the modelers Althouse and Scarpino but now studying the question in reverse by focusing on the transition from wP to aP vaccines¹². Using a signal-processing technique called “wavelet analysis”, they found that the switch to aP vaccines preceded a contraction of the inter-epidemic period length, the opposite of introducing wP vaccines to a vaccine-naïve population.

The Althouse/Scarpino model further combined population-level incidence data and phylogenetic data mapping the genetic evolution of B. pertussis as a means to account for the burden of disease through observed contact linkages over space and time. Their surprising finding was a far higher pace of genetic mutation than could be accounted for via visible transmission pathways between known symptomatic cases. This excess genetic diversity implies longer chains of transmission between observed cases and therefore argues for the existence of asymptomatic transmission routes. This conclusion echoes that of the British epidemiologists Miller and Gay, writing in 2000, 4 years before the introduction of aP vaccines into the infant schedule⁷⁷. Disease among infants <3 months old (i.e. too young to have been vaccinated) increased sharply in the late 1970s, coincident with declines in wP coverage rates. As wP coverage improved in the 1980s, pertussis incidence fell, including among young infants. Surprisingly, though, the infant rates stabilized in the 100–200 cases/100,000 infants/year range, even as disease among older age groups fell almost to zero. Since those infants are unlikely to be their own reservoir for infection, a more plausible explanation was transmission from a sustained pool of asymptomatic older individuals in the population⁷⁷. This also emphasizes that even very high rates of wP vaccination may fail to completely interrupt pertussis transmission.

These models also offer a possible explanation for the surprising failure of “cocooning” to protect infants from pertussis. Cocooning refers to the strategy of administering Tdap to all household contacts (particularly the mother) of a newborn to prevent spread to the infant during the early months of life. Cocooning is logical but presupposes that aP vaccines prevent asymptomatic infections. Unfortunately, several controlled trials of cocooning in the US found no efficacy^15,16,78. These counterintuitive results conflict with expectations if aP vaccines block carriage and transmission but fit well if aP vaccines only prevent disease but have more limited ability to block infections.

Recent experimental animal data show differential effects of aP and wP vaccines

Perhaps the most direct evidence of asymptomatic carriage and its implications emerged from an elegant series of experiments by Warfel and Merkel at the FDA using a newly developed infant baboon model. The key experiments in this series involved groups of infant baboons who were unvaccinated, vaccinated with aP vaccines, or vaccinated with wP vaccines and subsequently infected via exposure to aerosols of B. pertussis⁷⁹. In each case, vaccine-naïve baboons developed clinical illness, whereas aP- or wP-vaccinated animals remained asymptomatic (no coughing or lymphocytosis).

However, the animals also underwent serial NP sampling over the ensuing days and weeks, and here clear differences between the aP- and wP-vaccinated animals emerged. Among the wP-vaccinated baboons, NP carriage was detectable for a mean of 18 days (at low carriage densities) versus 30 and 35 days (at high densities) for naïve and aP-vaccinated animals, respectively. In fact, aP-vaccinated baboons eradicated NP colonization no more quickly than unvaccinated animals.

Follow-on experiments shed further light. Infant baboons immunized with aP vaccines and then infected with B. pertussis (but asymptomatic) were co-housed with vaccine-naïve baboons who had not been exposed to the B. pertussis aerosols: the unvaccinated cage mate animals also became infected, showing that aP-vaccinated animals remained contagious.

Approaching this in the opposite direction, the investigators infected a vaccine-naïve animal with pertussis and then co-housed it with both an aP-vaccinated animal and a second vaccine-naïve animal⁸⁰. Both of these animals acquired NP carriage at the same colonization density.

In a final experiment, female baboons were vaccinated with aP vaccines during pregnancy. Post-partum, the infant baboons were exposed to pertussis aerosols but were fully protected from clinical disease. However, all became colonized at similar densities as the infant baboons of unvaccinated mothers⁸¹.

In summary, these animal experiments demonstrate that aP vaccines prevent disease but not infection and that this clinical protection is mediated by antibody, is transferable over the placenta, and can prevent disease post-partum in infants. But the aP vaccine-derived antibodies do not block colonization of the nasopharynx, allowing aP-vaccinated and passively protected animals to become infected and to propagate transmission. By contrast, wP-vaccinated animals diminished carriage in both bacterial burden and duration, clearing infections in half the time required by vaccine-naïve or aP-vaccinated animals. To further quantify this, they compared the cumulative area under the curve for B. pertussis shedding from the nasopharynx of vaccinated and infected animals: wP vaccines reduced the total burden of bacterial shedding by >1,000-fold compared with DTaP⁸².

Empiric human evidence from a recent longitudinal study supports these interpretations. Looking at secondary attack rates among vaccine failures, the investigators calculated that wP vaccination reduced secondary infections by 86% compared with only 6% from aP vaccines⁸³.

Recent immunologic insights and a possible missing link

As suggested by Klein’s natural experiment data (Figure 4), the effects of wP and aP vaccines appear distinct. But what immunologic mechanism mediates these differences?

For decades, it was understood that naïve T-helper cells can differentiate into distinct lineages with distinct phenotypes and functional roles based on exposure to different combinations of antigenic stimuli, cytokine-mediated extracellular signaling from antigen presenting cells (APCs), and the specific combination of co-receptor molecules present at the interface between T-cells and the APC. The net result of these forces induces T-helper cells to adopt one of two functional phenotypes: TH1 or TH2⁸⁴. TH1 cells are optimized for cell-mediated immunity targeting intracellular pathogens (such as tuberculosis), with actions mediated indirectly by T8 effector cells. By contrast, TH2 cells are optimized to stimulate B-cells for extracellular killing mediated primarily via antibodies and play a central role in allergic responses and controlling parasitic helminth infections. Yet, while helpful, this model has evident gaps, since it does not clearly explain a number of opportunistic conditions that characterize HIV/AIDS but lack a clear link to either TH1 or TH2 phenotypes.

More recently, a third TH cell developmental pathway was identified, whose inclusion seems to fill many of the gaps left by the TH1/TH2 paradigm. This was named TH17 in reference to the dominant cytokine (interleukin 17) released by APCs. TH17 lymphocytes direct neutrophil-mediated extracellular immune responses on mucosal surfaces, and deficits in TH17 manifest clinically as an inability to eradicate mucosal carriage and increased susceptibility to pathogens that colonize mucosal surfaces⁸⁵. Since HIV degrades TH17 and TH1 responses, whereas TH2 responses may actually increase in the setting of HIV/AIDS, this provides a far more satisfying explanation for such HIV-associated conditions as Candida albicans esophagitis, Pneumocystis jirovecii pneumonia, Cryptococcus neoformans meningitis, and invasive pneumococcal disease, conditions strongly associated with pathogen proliferation upon, and invasion across, mucosal surfaces. TH17 also proves directly relevant to pertussis^86,87.

Warfel and Merkel assayed T-cell phenotypes in their infant baboons before and after infection with B. pertussis, or wP or aP vaccination. Pertussis infection triggered a purely TH17 response, whereas wP vaccination led to a predominantly TH17 response, with a lesser TH1 response. By contrast, aP vaccines induced only TH2 responses, which resulted in (mostly IgE) antibody generation and would not be expected to interfere with pathogens on mucosal surfaces⁸⁸.

These data suggest a definable immunologic mechanism distinguishing wP and aP vaccines that explains their differential effects on mucosal immunity and herd immunity and go a long way towards harmonizing the experimental and modeling data.