The toll of measles on the immune system
Many of the deaths attributable to measles virus are caused by secondary infections because the virus infects and functionally impairs immune cells. Whether measles infection causes long-term damage to immune memory has been unclear. This question has become increasingly important given the resurgence in measles epidemics worldwide. Using a blood test called VirScan, Mina et al. comprehensively analyzed the antibody repertoire in children before and after natural infection with measles virus as well as in children before and after measles vaccination. They found that measles infection can greatly diminish previously acquired immune memory, potentially leaving individuals at risk for infection by other pathogens. These adverse effects on the immune system were not seen in vaccinated children.
Science, this issue p. 599
Measles virus is directly responsible for more than 100,000 deaths yearly. Epidemiological studies have associated measles with increased morbidity and mortality for years after infection, but the reasons why are poorly understood. Measles virus infects immune cells, causing acute immune suppression. To identify and quantify long-term effects of measles on the immune system, we used VirScan, an assay that tracks antibodies to thousands of pathogen epitopes in blood. We studied 77 unvaccinated children before and 2 months after natural measles virus infection. Measles caused elimination of 11 to 73% of the antibody repertoire across individuals. Recovery of antibodies was detected after natural reexposure to pathogens. Notably, these immune system effects were not observed in infants vaccinated against MMR (measles, mumps, and rubella), but were confirmed in measles-infected macaques. The reduction in humoral immune memory after measles infection generates potential vulnerability to future infections, underscoring the need for widespread vaccination.
In the time before vaccination, nearly every child experienced measles, which resulted in millions of deaths. Global measles vaccination efforts have led to logarithmic reductions in the incidence of measles virus (MV) infections and measles-related mortality. However, measles remains endemic in much of the world, affecting >7 million people annually and causing >100,000 deaths (1–3). After decades of decline, the number of worldwide cases of measles has increased by nearly 300% since 2018 as a result of reduced vaccination (2). This increase is likely to be accompanied by substantial mortality risks (3). The resurgence of measles underscores the importance of understanding the full consequences of MV infection and accurately estimating the value of measles vaccination (4).
Immunosuppression was first documented when children with measles showed negative cutaneous tuberculin reactions after previously testing positive (5). Subsequent studies have shown decreased interferon signaling, skewed cytokine responses, lymphopenia, and suppression of lymphocyte proliferation shortly after infection (6). The MV receptor CD150/SLAMF1 (signaling lymphocytic activation molecule family member 1) is highly expressed on memory T, B, and plasma cells, resulting in their infection and depletion without an effect on total immunoglobulin G (IgG) levels (7–12). Recovery of the functional immune response, including resolution of lymphopenia, occurs 2 to 4 weeks after viral clearance (6, 10, 13, 14). However, MV replication in immune cells has been hypothesized to impair immune memory, potentially causing “immunological amnesia” (10, 15, 16).
Most bona fide immune memory cells reside in the lymphoid tissues and bone marrow (17–20). Peripheral blood mononuclear cells are often used for evaluating immunological memory repertoires. However, these cells are in relative flux owing to recent infections, which limits their utility for measuring long-term immune memory. Antibodies are thought to better represent long-lived humoral memory (18, 20). Most antibodies in the peripheral blood are produced by bone marrow long-lived plasma cells (LLPCs) and are impervious to disruptions in peripheral memory cells (17–22). Changes in pathogen-specific antibodies measured in the peripheral blood reflect changes in the long-lived permanent memory repertoire.
Epidemiological evidence has associated MV infections with increases in morbidity and mortality for as long as 5 years (15, 23) and suggests that in the pre-vaccine era, MV may have been associated with up to 50% of all childhood deaths from infectious diseases, mostly from non-MV infections (15). This phenomenon might be explained by immune amnesia. However, to date, no study has successfully resolved whether measles-induced immune amnesia—a reduction in the diversity of the immune memory repertoire after measles infections—indeed exists. To address this issue, we have studied paired blood samples collected before and after MV infection using a seroprofiling tool that allows the detection of thousands of pathogen-specific antibodies.
Measuring the consequences of measles on immune memory
During a recent measles outbreak in the Netherlands, families in communities with low vaccination rates consented to provide blood samples. Plasma was collected before and after laboratory-confirmed MV infection from 77 unimmunized children with a mean age of 9 (SD ± 2) years, plus five unimmunized children who remained uninfected during the study (24). Of the 77 children, 34 were reported to have mild measles and 43 to have severe measles [detailed in (24)]. The mean time between sample collections was 10 weeks, and mean time of collection after MV infection was 7 weeks (table S1).
To measure the diversity and magnitude of the epitope-specific antibody repertoires in these children and controls, we used VirScan (25), a phage-display immunoprecipitation and sequencing (PhIP-Seq) technology (26) developed for virome-wide detection of antibodies against viral epitopes. VirScan primarily detects antibodies to short contiguous epitopes as opposed to conformational epitopes. The cells producing antibodies to all epitopes are phenotypically similar, aside from their antibody product. Thus, changes in the antibody repertoire detected by VirScan represent changes across the spectrum of antibodies, and these include neutralizing and non-neutralizing antibodies. For this study, we generated an expanded VirScan library that encodes the full proteomes of most known human pathogenic viruses (~400 species and strains) plus many bacterial proteins. For each sample, we obtained a comprehensive measure of the individual’s antipathogen antibody repertoire diversity (i.e., the total epitope hits across all pathogen peptides). We also derived an antibody epitope binding signal (EBS), which is a relative measure of antibody titer for each epitope.
We used VirScan to profile the immune memory antibody repertoires before (time 1) and after (time 2) MV infection. Paired samples were also obtained from four control cohorts (n = 119 paired specimens; table S1). These samples were derived from: (i) approximately age-matched controls sampled at similar intervals (~3 months) as the measles cohorts (control A; n = 28 paired specimens); (ii) age-matched controls with samples collected ~1 year apart (control B; n = 31); (iii) adult controls with collection intervals similar to the measles-infected individuals (control C; n = 22); (iv) young children before and after their first measles-mumps-rubella (MMR) vaccination (MMR vaccinated; n = 33); and (v) unvaccinated children from the same community as the MV cases but who remained seronegative for MV (MV negative; n = 5). Control cohorts A, B, and C were individuals with no known exposure to MV.
Measles modulates the diversity of the antibody repertoire and causes loss of preexisting antibodies
We assessed changes in antibody repertoire diversity (measured as the total number of unique pathogen epitopes recognized, or epitope hits) before and after measles relative to those observed in controls, standardizing the total number of epitope hits per individual by cohort for comparison (Fig. 1A). We detected substantial reductions in the number of pathogen epitopes recognized after measles but limited changes in the absence of measles. MV infections were associated with a mean reduction of ~20% in the overall diversity or size of the antibody repertoire measured by VirScan (Fig. 1B), and this was consistent across individual pathogens (Fig. 1, D and E, and fig. S1). However, effect sizes varied. Notably, 12 of the 77 children (16%) lost >40% of their overall antibody repertoire diversity. We detected increases in MV-specific epitopes in children after measles infection or MMR vaccination (Fig. 1C). No changes in the total IgG, IgA, or IgM levels were detected, as determined by quantitative ELISA (enzyme-linked immunosorbent assay) (fig. S2). These results suggest that, rather than a simple loss of total IgG, there is a restructuring of the antibody repertoire after measles.
To measure the full effect of measles on the pre-measles repertoire and to circumvent interference from new exposures during follow-up, we next restricted analysis to epitopes detected at the first time point and quantified retention or loss of epitope recognition. After severe or mild measles, children lost a median of 40% (range: 11 to 62%) or 33% (range: 12 to 73%), respectively, of their total preexisting pathogen-specific antibody repertoires (Fig. 2A and fig. S3). In contrast, controls retained ~90% of their repertoires over similar or longer durations.
Controlling for interval duration in a binomial random effects model (see materials and methods), we estimated the per-pathogen probability of antibody retention for each child (Fig. 2, B and C). Loss of antibodies after MV infection varied widely for specific pathogens and between children. A small fraction of MV-infected individuals retained antibodies similar to the bottom quartile of the controls. However, the most-affected 20% of children lost >50% of the pathogen-specific antibodies for most pathogens. In some of these children, up to 70% loss was detected for specific pathogens.
Antibody repertoire retention (~90%) was similar in control cohorts A and B despite the longer sampling interval in B compared with A (1 year versus 3 months). The retained antibody repertoire could represent a core stable LLPC repertoire (~90%), and the 10% that was lost could represent a transient repertoire derived from IgG-secreting B cells and plasmablasts. This result is consistent with previous studies that showed no effect of immunosuppressive therapy, such as B cell–depleting anti-CD19 or anti-CD20 treatment, on retention of the majority of the antibody repertoire (17, 19, 21, 22). Therefore, measles is associated with greater loss of antibody-mediated epitope recognition than can be explained by ablation of B cells, suggesting a direct effect on the LLPC compartment.
Measles decreases the strength of epitope recognition
Simply counting epitope numbers recognized before and after measles underestimates immune memory impairment because epitope recognition can be detected even when a large fraction of cellular clones producing the relevant antibody are eliminated. To quantify the relative abundance of particular antibodies, we generated an EBS metric, which is a z-score that measures the relative enrichment of an epitope in a VirScan immunoprecipitation (see materials and methods). Thus, EBS represents a VirScan analog of relative antibody titers and can be measured over time to detect changes in epitope recognition. Significant increases in EBS for a given epitope usually indicate a new exposure during the follow-up interval. In controls, significant reductions are normally detected when a recent exposure has occurred before the first time point, resulting in a peak around time 1 followed by an expected rapid decay of the antibody signal that is detected at time 2.
To evaluate changes in EBS for each pathogen species in each child, EBS signals were clustered by pathogen species and child (EBSPC) and a paired t test (paired per epitope) was used to test for significant differences at time 2 versus time 1. False discovery rate (fdr)–adjusted P values, sample size, and effect size (calculated as the fold change of the geometric mean at time 2 versus time 1) are shown for each EBSPC cluster (Fig. 3A and fig. S4).
d EBS, and found good correlation (correlation coefficient r = 0.73; P < 0.001) (fig. S5, A and B).
To further clarify cohort-level changes, we combined the epitopes from all individuals and evaluated changes on a per-pathogen-species basis across each cohort (fig. S6). Among MV-infected individuals, the median fold change in EBS measured across all species was a strongly negative effect (median fold change: 0.69 or −31%; P < 0.0001). In contrast, among controls, the antibody EBS was significantly changed only for relatively few pathogen species, and these were roughly balanced between positive and negative effects.
Measles disrupts recognition of pathogen epitopes across the cohort
At the cellular level, MV would not be expected to preferentially infect particular antibody-secreting cells on the basis of their pathogen targets or the type of antibodies produced (i.e., neutralizing or not). We therefore tested for cohort-wide differences in EBSs on a per-epitope basis. Across approximately 1100 epitopes each from the control A and B cohorts, none significantly changed (Fig. 3, B and C, and fig. S7), indicating that antibody epitope recognition across a group of individuals is remarkably stable, even when changes are detected at the pathogen species level, as in fig. S6. In contrast, EBS was substantially reduced in 12% of the 855 epitopes evaluated after mild MV infections and in 39% of the 1079 epitopes evaluated after severe MV infections (P < 0.0001) (Fig. 3, B and C). Only one epitope in mild measles and three in severe measles showed a significant increase, <0.2% overall.
VirScan detects neutralizing antibodies when those antibodies target short contiguous epitopes encoded in the phage display. A well-characterized neutralizing short contiguous epitope is the 24-amino-acid target of the anti-RSV monoclonal antibody therapies palivizumab and motavizumab. Among 22 MV-infected children with antibodies against this neutralizing epitope and without evidence of new RSV exposure, we detected a fold change in EBS of 0.59 (SD ± 0.18; P < 0.001) (fig. S8), in line with the observations above and indicating that the effects of MV are the same for the neutralizing and non-neutralizing antibody repertoires.
The MMR vaccine does not impair the immune repertoire
d increase in the overall antibody repertoire diversity was noted in MMR-vaccinated controls (Fig. 1), indicating that a similar loss of antibodies does not appear to accompany receipt of MMR vaccines compared with MV infections. However, in infants and young children, such as the MMR-vaccinated cohort in this study, the antibody repertoire continues to add antibody diversity over time (fig. S9). This is particularly true during the second year of life, following depletion of maternal antibodies, when measles vaccines are first given. These overall increases in antibody diversity over time could be obscuring potential minor impairments from measles vaccine, especially given the relatively long sampling interval for the vaccinated controls. To determine whether viral responses before or after MMR vaccination might be impaired, we analyzed EBS for antibodies detected surrounding MMR vaccine. We found no overall change in EBS after MMR vaccination (fig. S10), which is consistent with observations that the MV vaccine strains are not associated with widespread infection of CD150+ lymphocytes (27, 28). Thus, this higher-resolution analysis failed to detect immune impairment at the epitope level. Combined with increased epitope diversity after MMR vaccination (Fig. 1A), this analysis supports decades of observations that MMR vaccines do not increase susceptibility to subsequent infections (29, 30). Nevertheless, because of the naturally rapidly increasing antibody diversity in this young cohort at the time of MMR receipt, we cannot definitively rule out the potential for minor reductions of antibody-producing cells with measles vaccination.
uction of the antibody repertoire with high-transmission pathogens is spatially clustered
Although antibody diversity and abundance were negatively affected for almost all pathogens, a subset of children had increased EBS and/or epitope hits for particular pathogens after measles (Figs. 1, D and E, and 3A; and figs. S4, S8, and S11), suggestive of potential restoration of immune memory after the initial immune depletion.
Across all of the pathogen-child combinations with significantly increased EBS after measles, 80% were attributable to only six pathogens: adenovirus C, influenza A virus, respiratory syncytial virus, human herpesvirus 4 (Epstein-Barr virus; HHV-4), Streptococcus pneumoniae and Staphylococcus aureus. In these children, the probability of retaining preexisting antibodies for these pathogens was relatively high (fig. S12).
If reconstruction of the antibody repertoire requires new exposures, children experiencing repertoire reconstruction for transmissible pathogens should cluster spatially. Using postal codes and household identifiers, we observed clustering of pathogen-specific repertoire reconstruction at both the postal code or school level and the household level (Fig. 4). Among 19 children with increased adenovirus C antibody EBS, eight (42%) came from a single postal code (number 7) that included only 12 (16%) of the 77 measles cases (Fig. 4A). Thus, children in this postal code had significantly increased odds [odds ratio (OR) 9.4; Fisher’s exact test, P < 0.001] of recovering their adenovirus C repertoire versus other postal codes, suggesting recovery is associated with pathogen transmission. Moreover, six of these children (75%) shared a household with another child with increased adenovirus C EBS. We found similar clustering effects for other respiratory pathogens: influenza A virus [of eight children with increased influenza A EBS, five (63%) were from a single postal code representing only 26% of the measles cases (OR 5.8; P < 0.05) and two were from the same household; Fig. 4B]; RSV [of nine with increased EBS, five (56%) were from the same postal code (OR 4.5; P < 0.05) and four of them (44%) shared a house; Fig. 4C]; rhinovirus [of 12 with increased EBS, eight (67%) shared a household with at least one other child with increased rhinovirus EBS]; and S. pneumoniae [of 13, seven (54%) came from a single postal code (OR 4.5; P < 0.05) and two shared a household]. Combined, these indicate local pathogen transmission and suggest that reconstruction of the antibody repertoire occurs on a per-pathogen basis and is associated with new exposures.