Dysregulated Humoral Immunity to Nontyphoidal Salmonella in HIV-Infected African Adults

See allHide authors and affiliations

Science  23 Apr 2010:
Vol. 328, Issue 5977, pp. 508-512
DOI: 10.1126/science.1180346

HIV and Salmonella

HIV-positive individuals who are infected with nontyphoidal strains of Salmonella enterica often succumb to high morbidity and mortality. Why this is the case is unknown. MacLennan et al. (p. 508; see the Perspective by Moir and Fauci) have uncovered a dysregulated antibody response to Salmonella that is the likely culprit. Sera from HIV-infected individuals do a poor job of killing S. Typhimurium, despite surprisingly elevated antibody titers. Experiments showed that HIV-infected serum inhibited the power of normal serum to kill Salmonella. Inhibition was specific to antibodies against lipopolysaccharide (LPS), a component of the cell wall of Salmonella. Hence, HIV-infected sera was able to kill Salmonella strains lacking LPS, and removing LPS immunoglobulin G from infected sera permitted Salmonella killing. Thus, not only does HIV cause defects in cell-mediated immunity but it also seems to impair humoral immunity, with severe consequences for multiple infections.


Nontyphoidal Salmonellae are a major cause of life-threatening bacteremia among HIV-infected individuals. Although cell-mediated immunity controls intracellular infection, antibodies protect against Salmonella bacteremia. We report that high-titer antibodies specific for Salmonella lipopolysaccharide (LPS) are associated with a lack of Salmonella-killing in HIV-infected African adults. Killing was restored by genetically shortening LPS from the target Salmonella or removing LPS-specific antibodies from serum. Complement-mediated killing of Salmonella by healthy serum is shown to be induced specifically by antibodies against outer membrane proteins. This killing is lost when excess antibody against Salmonella LPS is added. Thus, our study indicates that impaired immunity against nontyphoidal Salmonella bacteremia in HIV infection results from excess inhibitory antibodies against Salmonella LPS, whereas serum killing of Salmonella is induced by antibodies against outer membrane proteins.

The association between HIV infection and fatal disease with nontyphoidal strains of Salmonella (NTS) was first described at the outset of the AIDS pandemic 26 years ago (1, 2). This is a global problem affecting affluent countries (3, 4), but particularly Africa (58). The underlying mechanisms are not known. NTS, especially Salmonella enterica serovars Typhimurium and Enteritidis, are a major cause of invasive bacterial disease in Africa affecting young children (5, 9) as well as HIV-infected adults. Case fatality and recrudescence rates are high (10), antibiotic resistance is an increasing problem (5), and currently no vaccine is available. Although Salmonellae are facultative intracellular pathogens (11) and cell-mediated immunity is important for controlling infection (1214), we recently demonstrated an important protective role for antibody-induced complement-mediated killing of NTS in African children (15). We investigated antibodies to Salmonella in the context of HIV infection because HIV causes extensive defects in the humoral immune system (1618). Our studies reveal aberrant humoral immunity to NTS in HIV-infected African adults, characterized by a lack of bactericidal activity resulting from dysregulated antibody production, with excess immunoglobulin G (IgG) directed against S. Typhimurium lipopolysaccharide (LPS). We also show that antibodies against S. Typhimurium outer membrane proteins induce killing of NTS in HIV-uninfected African adults.

To determine whether HIV infection affects humoral immunity to NTS, we assessed in vitro killing of two invasive Malawian S. Typhimurium isolates by sera from Malawian adults (19). Isolate A23753 was killed by all sera from HIV-uninfected adults, with a log10 kill at 180 min of ≥0.9 (designated “normal kill”) (Fig. 1A), and all effected a 3.0 log10 kill of A19520 by 45 min (Fig. 1B). In contrast, there was considerable variation in the ability of sera from HIV-infected adults to kill both isolates. Twenty-eight percent of sera failed to effect a 0.9 log10 kill of A23753 by 180 min (Fig. 1C), and 59% failed to produce a 3.0 log10 kill of A19520 by 45 min (Fig. 1D). All sera had normal total and alternative pathway hemolytic complement activity (table S1), excluding complement degradation or impaired synthesis as reasons for impaired killing. HIV targets CD4+ T lymphocytes, and lowered blood CD4+ lymphocyte numbers (CD4 counts) are associated with increased susceptibility to NTS bacteremia (20). CD4 counts of HIV-infected people with impaired serum killing of A23753 were lower than those with normal killing (P = 0.05) (fig. S1).

Fig. 1

Dysregulated humoral immunity to NTS in HIV infection. Killing of (A and C) S. Typhimurium A23753 and (B and D) serum-sensitive S. Typhimurium A19520 by sera at 45, 90, and 180 min. Negative values correspond with a decrease in viable Salmonellae as compared with the initial concentration. (E and F) Serum titers of Salmonella A23753 IgG compared with killing of S. Typhimurium A23753 at 180 min and (G and H) C5b-9 MAC deposition on A23753. CI, confidence interval. (A), (B), (E), and (G) Sera from HIV-uninfected Africans (n = 58). (C), (D), (F), and (H) Sera from HIV-infected Africans (n = 58). Each line or point represents data for serum from one individual. All lines are superimposed in (B). The horizontal dashed line in (E) and (F) indicates the threshold for impaired killing of S. Typhimurium A23753 (–0.9 log10 change in Salmonellae colony-forming units per milliliter).

Next, IgG binding to S. Typhimurium A23753 was measured in all sera to determine whether a lack of antibody was the reason for impaired Salmonella killing. S. Typhimurium–specific IgG was present in all sera, and, paradoxically, IgG titer positively correlated with impaired Salmonella-killing by HIV-infected sera (P = 0.002) (Fig. 1, E and F). S. Enteritidis D24954-specific IgG was also present in all sera and positively correlated with S. Typhimurium IgG titer for HIV-uninfected and -infected sera (fig. S2). Some impairment of killing of S. Enteritidis D24954 was observed with a subset of HIV-infected sera that could not kill S. Typhimurium D23580. S. Enteritidis IgG titer correlated with impaired killing of S. Enteritidis (fig. S3).

In case Salmonella-specific antibody in HIV-infected sera could not activate complement, we measured deposition on A23753 of C5b-9 membrane attack complex (MAC), the final effector of complement-mediated bactericidal activity. MAC deposition was detected for all sera and strongly correlated with Salmonella-specific IgG titer for HIV-infected and -uninfected sera (Fig. 1,G and H). We also detected IgG binding and C3 complement deposition for HIV-infected and -uninfected sera by confocal microscopy (fig. S4). This indicates that a failure to deposit complement is not responsible for a lack of Salmonella-killing by HIV-infected serum.

Killing of Salmonella A23753 was impaired when different proportions of HIV-infected sera that could not kill Salmonella were mixed with HIV-uninfected serum (Fig. 2A). For some HIV-infected sera, this impairment was observed with one part HIV-infected serum to nine parts control serum. Thus, an inhibitor in HIV-infected serum blocks killing. The inhibitory factor was found to be between 100 and 300 kD (Fig. 2, B and C). We tested whether this was an antibody, because IgG is approximately 160 kD. Total IgG at 10 g/liter extracted from inhibitory HIV-infected sera blocked killing of S. Typhimurium D23580 and D19774 by control sera (Fig. 2, D and E). Conversely, IgG from HIV-uninfected sera had no effect on killing.

Fig. 2

Inhibition of HIV-uninfected control serum killing of NTS by HIV-infected sera with impaired Salmonella-killing ability. (A) Killing of S. Typhimurium A23753 at 180 min by mixed sera consisting of different percentages of HIV-infected serum (n = 12, serum from one HIV-infected person per line) and control HIV-uninfected serum. The horizontal dashed line indicates the threshold for impaired killing of S. Typhimurium A23753. (B to E) Inhibition of control serum killing of S. Typhimurium D23580 [(B) and (D)] and serum-sensitive S. Typhimurium D19774 [(C) and (E)] by size-fractionated [(B) and (C)] and IgG fraction [(D) and (E)] of HIV-infected serum. PBS, phosphate-buffered saline. Data are means ± SD of three experiments. Inhibition of killing of both strains of Salmonella by HIV-uninfected sera with a <300-kD fraction of HIV-infected serum as compared with a <100-kD fraction, and with an IgG fraction of HIV-infected serum as compared with an IgG fraction of HIV-uninfected serum was significant by Student’s t test (P < 0.0001).

We then tested whether inhibition results from excess total serum Ig, because hypergammaglobulinemia is a well-recognized feature of HIV infection (16, 17). Although higher total IgG titers were present in HIV-infected as compared with -uninfected sera (P < 0.0001), there was only a small, yet significant correlation between total serum IgG and IgA, but not IgM, and impaired killing of S. Typhimurium (fig. S5). This suggests that inhibitory IgG binds specific targets on S. Typhimurium. We hypothesized that antibody-targeting structures away from the bacterial membrane might prevent killing. NTS are surrounded by LPS with long polysaccharide side chains (O-antigen) extending from the outer membrane along with flagella (consisting of flagellin, which is the H-antigen) (21). LPS and flagellin are highly immunogenic (22). We previously showed that the O-antigen of invasive African S. Typhimurium protects against complement-mediated killing in the absence of antibody (15). Earlier studies found that MAC deposited on LPS of S. Minnesota does not insert into the bacterial membrane (23), and rabbit LPS IgG can inhibit the bactericidal effect of bovine serum on S. Typhimurium (24). These considerations led us to test whether LPS and flagellin are targets of inhibitory IgG.

S. Typhimurium LPS IgG titers were selectively elevated in HIV-infected compared with -uninfected sera (P < 0.002) (Fig. 3, A and B), whereas flagellin-specific IgG titers were comparable (fig. S6). Impaired Salmonella-killing in HIV-infected sera correlated with LPS IgG titer (P = 0.0002) but not with flagellin IgG titer. We confirmed the correlation between LPS IgG and impairment of Salmonella-killing by measuring LPS IgG in a subset of HIV-infected and -uninfected sera by fluorescent-bead–based immunoassay (fig. S7). These results are consistent with LPS IgG being the key inhibitor. Median IgM titers to LPS [as previously reported (25)] and flagellin were respectively higher or not significantly different in HIV-uninfected as compared with HIV-infected sera (fig. S8), arguing against a role for IgM in the inhibition of Salmonella-killing.

Fig. 3

LPS antibodies prevent killing of NTS by HIV-infected serum. (A and B) Killing of S. Typhimurium A23753 at 180 min by HIV-uninfected sera (A) and HIV-infected sera (B) compared with S. Typhimurium LPS IgG titer determined by enzyme-linked immunosorbent assay. The horizontal dashed line indicates the threshold for impaired killing of S. Typhimurium A23753. r is the Spearman correlation coefficient. The median IgG titer was higher in HIV-infected sera than in HIV-uninfected sera (Mann-Whitney test, P < 0.002; difference in medians, 20 units; 95% CI, 6.0 to 39). (C) Killing of indicated S. Typhimurium D23580 by HIV-infected serum. (D) Effect of preabsorbing HIV-infected serum with 100 μg/ml of S. Typhimurium flagellin or outer membrane proteins on killing of S. Typhimurium D23580. (E and F) Effect of preabsorbing HIV-infected serum with 1, 10, or 100 μg/ml LPS on serum killing of (E) D23580 and (F) D19774. (G and H) Effect of adding LPS antibodies at 1-10th the concentration in inhibitory HIV-infected source serum to HIV-uninfected serum on killing of (G) D23580 and (H) D19774. Data represent means ± SD of three experiments. Killing of both strains of Salmonella by HIV-infected sera preabsorbed with 100 μg/ml of LPS as compared with unabsorbed serum, and inhibition of killing of both strains of Salmonella by HIV-uninfected serum with exogenous LPS antibody added as compared with PBS added was significant by Student’s t test (P < 0.0001).

To test further whether LPS IgG inhibits Salmonella-killing, the ability of HIV-infected serum to kill without the LPS target antigen was examined using a galE mutant of S. Typhimurium D23580 lacking O-antigen polysaccharide (15). The mutant was fully susceptible to killing by inhibitory HIV-infected serum (Fig. 3C). Wild-type, flgBCD, and ompR mutants of D23580, deficient in the expression of flagellin and certain outer membrane proteins, respectively, served as controls and could not be killed. These results indicate that inhibitory HIV-infected sera have an inherent capacity to kill Salmonella and suggest that inhibitory antibodies target O-antigen, further implicating LPS IgG as the inhibitor. We investigated the effect of absorbing LPS antibodies from HIV-infected serum. Preabsorption with S. Typhimurium flagellin and outer membrane proteins at 100 μg/ml did not affect bactericidal activity (Fig. 3D), but preabsorption with LPS fully restored killing of S. Typhimurium D23580 and D19774 (Fig. 3, E and F, and fig. S9). For HIV-infected sera with partially impaired Salmonella-killing ability, 1 μg/ml of LPS restored normal killing (fig. S9).

Finally, LPS IgG extracted from inhibitory HIV-infected serum was added to HIV-uninfected serum. Inhibition of killing of S. Typhimurium D23580 and D19774 was induced at 1-10th the LPS IgG concentration present in source serum (Fig. 3, G and H), confirming LPS IgG as the inhibitor of Salmonella-killing. Killing of both strains by HIV-uninfected serum was also inhibited by LPS IgG from autologous HIV-uninfected serum at 10 times the original concentration in source serum (fig. S10) (the relative concentration of LPS IgG in HIV-uninfected serum was 1-60th that in the HIV-infected serum used). This indicates that LPS IgG titer rather than the source of this antibody is critical for inhibition of Salmonella-killing.

Elevated IgG titers in HIV infection are characterized by antibodies to HIV viral proteins (26, 27) and self-antigens (27, 28). This occurs in parallel with the loss of antigen-specific antibodies; for example, to tetanus toxoid and measles (29). The global reduction in T-dependent (30, 31) and T-independent (18, 31, 32) antibody responses after immunization in HIV-infected individuals contrasts with increased antibody to Salmonella-specific LPS. This indicates that immune dysregulation, not immune deficiency, accounts for impaired humoral immunity to nontyphoidal Salmonella. The high proportion of HIV-infected people with elevated LPS IgG suggests that high titers are not the consequence of random expansion of antigen-specific B cell clones. The explanation may relate to elevated plasma LPS titers in HIV infection secondary to microbial translocation from the gastrointestinal tract (25). We found no correlation between serum LPS and S. Typhimurium LPS antibody titers (fig. S11) and no difference between serum LPS levels in HIV-infected and -uninfected sera (P = 0.33). However, LPS is likely to be cleared from the blood by the antibody it induces in immune complexes and become localized in secondary lymphoid tissue.

We hypothesized that LPS antibodies prevent the killing of Salmonella by two possible mechanisms that are not mutually exclusive. One would act by diverting complement deposition away from the bacterial membrane, thereby preventing insertion of MAC into the membrane (fig. S12). The other would impede the access of antibody and/or complement to the outer membrane by cross-linking O-antigen, the distal portion of the LPS molecule. To test these hypotheses, we investigated whether the inhibitory antibodies bind O-antigen, rather than proximal lipid A and core oligosaccharide moieties (21). Preabsorption of inhibitory HIV-infected sera with smooth S. Typhimurium LPS at 100 μg/ml enabled killing of S. Typhimurium D23580 (Fig. 3, E and F). However, preabsorption with 100 μg/ml of lipid A or LPS from Rb, Rc, Rd, and Re rough forms of Salmonella, where LPS is truncated in the core oligosaccharide (33), did not induce killing of Salmonella (fig. S13, A and B). These findings indicate that inhibitory antibodies target O-antigen. We also found that inhibitory antibodies could not be removed by preabsorbing with LPS from S. Enteritidis (group D Salmonella) and S. Minnesota (group L Salmonella) (fig. S13, C and D). This provides further evidence that O-antigen is targeted by inhibitory antibodies, because LPSs from these three Salmonella serovars are distinguished by their non–cross-reactive O-antigens.

The concept that inhibitory antibodies act by binding O-antigen, a target distal to the Salmonella membrane, implies that protective bactericidal antibodies target molecules proximal to the membrane (fig. S12), an idea we have previously suggested (15). This conclusion is consistent with recent reports that antibodies against S. Typhimurium outer membrane proteins, in particular porins OMP F, C, and D, protect against Salmonella in the mouse (34). Consequently, we investigated whether such antibodies are responsible for S. Typhimurium–killing by serum from Africans. First, we preabsorbed serum from HIV-uninfected Malawian adults with S. Typhimurium outer membrane proteins, LPS, or flagellin. Although preabsorption with LPS and flagellin had no effect, killing was abrogated by preabsorption with outer membrane proteins (Fig. 4A). This indicates that antibodies against these proteins are bactericidal. Next, we immunized mice with OMP F, C, and D porins, boosted at day 14, and used heat-inactivated sera from mice at day 21 as a source of OMP F, C, and D–specific antibodies. Immunized sera, but not sera from unimmunized littermates, enabled antibody-deficient human serum to kill S. Typhimurium D23580 (Fig. 4B). This provides further evidence that antibodies against outer membrane proteins, in particular porins, cause Salmonella-killing.

Fig. 4

Antibodies targeted against outer membrane proteins mediate African serum killing of NTS. (A) Effect of preabsorbing HIV-uninfected serum with 200 μg/ml of S. Typhimurium outer membrane proteins, flagellin, or LPS on killing of S. Typhimurium D23580. (B) Effect of adding OMP F, C, and D antibodies from four mice immunized with S. Typhimurium OMP F, C, and D to antibody-deficient HIV-uninfected human serum on killing of S. Typhimurium D23580 as compared with adding antibody from four unimmunized mice. Each line represents a log10 change of Salmonella induced by antibody from one mouse. (C) Effect of adding outer membrane protein antibodies from four HIV-uninfected sera (one per line) at 1-10th the concentration present in source serum, to antibody-deficient serum on killing of S. Typhimurium D23580. In (B) and (C), lines are superimposed. (D and E) Effect of adding outer membrane protein antibodies and LPS antibodies extracted and purified from inhibitory HIV-infected serum to antibody-deficient serum (D) or immune HIV-uninfected serum (E) on killing of S. Typhimurium D23580. In (A), (D), and (E), data represent means ± SD of three experiments.

Finally, we purified antibodies to outer membrane proteins from HIV-uninfected and -infected Malawian sera. These antibodies, when added to antibody-deficient serum at 1-10th the concentration in source serum, enabled killing of D23580 (Fig. 4C), even when extracted from HIV-infected inhibitory serum (Fig. 4D). The outer membrane protein antibodies had no effect when added to immune HIV-uninfected serum (Fig. 4E). This contrasts with the lack of killing of Salmonella observed after adding LPS antibody to antibody-deficient and immune serum (Fig. 4, D and E). The findings also indicate that individual sera contain antibodies that can kill Salmonella and block killing of Salmonella (fig. S14).

These results suggest that killing of Salmonella by inhibitory HIV-infected sera could be restored by adding IgG from HIV-uninfected serum. We added human normal IgG immunoglobulin pooled from HIV-uninfected donors to inhibitory HIV-infected sera. This induced killing in a dose-dependent manner in three inhibitory sera but not in a fourth serum, which had an LPS antibody titer over 10 times higher than the other sera (fig. S15). Finally, killing of Salmonella in antibody-deficient serum could be induced or prevented by adding combinations of IgG from HIV-uninfected and inhibitory HIV-infected sera depending on the proportion of IgG from each serum (fig. S16). This supports the concept of competition between blocking antibodies and killing antibodies to Salmonella.

Dysregulated humoral immunity in HIV-infected Africans could contribute to their susceptibility to invasive Salmonella by undermining protective antibody-mediated immunity that develops within the first 2 years of life (15). Together with impaired cellular immunity in HIV infection, it is unsurprising that HIV-infected adults suffer from repeated episodes of Salmonella infection with associated high mortality (6, 10). A vaccine for nontyphoidal Salmonella is urgently required for Africa. The current study indicates that although an O-antigen polysaccharide-based vaccine might be ineffective and increase susceptibility to life-threatening extracellular Salmonella growth, an outer membrane protein–based vaccine could induce protective antibodies.

Supporting Online Material

Materials and Methods

Figs. S1 to S16

Table S1


References and Notes

  1. Information on materials and methods is available on Science Online.
  2. We thank the individuals who participated in this study and the staff at the Malawi–Liverpool–Wellcome Trust Clinical Research Programme and Queen Elizabeth Central Hospital, Blantyre, Malawi, for their support. We are grateful to T. Plant, K. Stephens, A. Seeley, and the Clinical Immunology Service, Birmingham, UK, for technical assistance and to F. Gotch and P. Kelleher for helpful discussions. This work was supported by a Clinical Research Fellowship from GlaxoSmithKline (C.A.M.); Tropical Research Fellowships from the Wellcome Trust (C.A.M. and M.A.G.); a Programme Grant from the Wellcome Trust (M.E.M.); a Scientific Projects Committee Research Grant from the University of Birmingham (C.A.M.); the Medical Research Council and UK Biotechnology and Biological Sciences Research Council, UK (A.F.C., D.L.L., and I.R.H.); National Institute for Health Research Cambridge Biomedical Research Center (R.D.); and the National Council for Science and Technology (CONACyT) Mexico, grant number SALUD-2007-C01-69779 (C.L.-M.).
View Abstract

Navigate This Article