Value of Immunological Markers in Predicting Responsiveness to Influenza Vaccination in Elderly Individuals

Abstract

Elderly individuals are at high risk for morbidity and mortality when infected with influenza virus. Vaccinations with inactivated virus are less effective in the elderly due to the declining competency of the aging immune system. We have explored whether immunological parameters predict poor anti-influenza virus vaccine responses and can be used as biological markers of immunosenescence. One hundred fifty-three residents of community-based retirement facilities aged 65 to 98 years received a trivalent influenza vaccine. Vaccine-induced antibody responses were determined by comparing hemagglutination inhibition titers before and 28 days after immunization. The composition of the T-cell compartment was analyzed by flow cytometry and the sizes of three T-cell subsets, CD4⁺ CD45RO⁺ cells, CD4⁺ CD28^null cells, and CD8⁺ CD28^null cells, were determined. Only 17% of the vaccine recipients were able to generate an increase in titers of antibody to all three vaccine components, and 46% of the immunized individuals failed to respond to any of the three hemagglutinins. The likelihood of successful vaccination declined with age and was independently correlated with the expansion of a particular T-cell subset, CD8⁺ CD28^null T cells. The sizes of the CD4⁺ CD45RO⁺ memory T-cell and CD4⁺ CD28^null T-cell subsets had no effect on the ability to mount anti-influenza virus antibody responses. Frequencies of CD8⁺ CD28^null T cells are useful biological markers of compromised immunocompetence, identifying individuals at risk for insufficient antibody responses.

Influenza is a highly contagious respiratory disease that is self-limiting in an immunocompetent individual but is associated with substantial morbidity and mortality in the elderly (6). Epidemics between 1972 and 1992 caused up to 11,800 deaths in the United States alone. Estimates by the Centers for Disease Control and Prevention suggest an average of 20,000 deaths of all causes per year during influenza seasons (40). Influenza viruses are unique in their ability to undergo point mutations and antigenic variations in their hemagglutinin and neuraminidase genes, thereby escaping protective immunity (4, 48). In addition, new pathogenic strains that have acquired new hemagglutinin genes can arise. Annual influenza epidemics usually result from antigenic drift, whereas antigenic shift or the acquisition of a new hemagglutinin is associated with a more severe pandemic. Therefore, immunity to influenza virus variants is a constant challenge for the immune system and is highly dependent on the ability to generate primary immune responses to new antigens. Due to senescence of the immune system, the ability to mount novel immune responses may be particularly compromised in elderly individuals.

Currently, annual vaccinations with inactivated vaccines are the most effective means of preventing influenza infections. The efficacies of the vaccines range from 60 to 90% in healthy adults (5, 22). Although immunogenicity is generally much lower in the elderly population, vaccination has been shown to be cost-effective and to reduce mortality and hospital admissions (27, 32, 33). Gross et al. (13) did a meta-analysis of 20 observational studies and 3 case-control studies and found a 50 to 68% decrease in respiratory illness, pneumonia, hospitalization, and death among elderly individuals who were vaccinated with influenza virus. Nevertheless, many vaccinated elderly individuals do not develop sufficient antibody titers and are only marginally protected from influenza or may have a protracted illness in spite of the vaccination.

The aging immune system undergoes changes that critically affect the development of adaptive immune responses (29). The formation of high-affinity antibodies in germinal centers is impaired, and delayed-type hypersensitivity reactions are reduced (50). Although the mechanisms underlying immunosenescence are multifactorial, the dramatic involution of the thymus has a prominent role (14, 15). The immune system is in constant need of replenishment; however, during adulthood, thymic output decreases by at least 2 logs and the void is filled by replication of mature lymphocytes and expansion of memory T cells (20, 49). These dynamic changes are reflected in phenotypic changes of peripheral lymphocytes. Memory and effector T cells with antigen experience express the CD45RO molecule, and their numbers increase with age (9, 24, 35). With replicative senescence, T cells lose the expression of the functionally important CD28 molecule and subsets of CD4⁺ CD28^null and CD8⁺ CD28^null T cells emerge (1, 8, 31, 34, 42, 43).

The present study was designed to identify biological markers of immunosenescence that are stable, easy to measure, and predictive of poor vaccine response. Our data suggest that the frequency of CD8⁺ CD28^null T cells is such a biological marker.

MATERIALS AND METHODS

Study population.

The study population consisted of 153 (112 female) healthy, ambulatory adults 65 years of age or older, living in community-based retirement facilities in Rochester, Minn. The protocol was approved by the Mayo Clinic Institutional Review Board, and all participants provided informed consent. They were seen at a single medical center. A medical history was obtained, and individuals with a serious disease, such as cancer (except localized skin cancer or nonmetastatic prostate cancer), a chronic inflammatory disease, advanced atherosclerotic disease or congestive heart failure, poorly controlled diabetes mellitus, acute or progressive renal or hepatic disease, or chronic obstructive pulmonary disease requiring the use of oxygen, were excluded. The mean age of the study population was 79.7 ± 7.6 years.

Immunization.

All individuals were immunized with a single intramuscular injection of a trivalent influenza vaccine, containing 15 μg each of the three influenza virus antigens A/Texas/36/91 (H1N1), A/Nanchang/933/95 (H3N2), and B/Harbin/7/94. Peripheral blood specimens were obtained prior to and 28 days after the vaccination.

Antibody assays.

Serum antibody responses to the three antigens contained in the vaccine were assayed by determining hemagglutination inhibition titers as described by Dowdle et al. (7) except that 0.4% guinea pig red blood cells were used instead of chicken red blood cells.

Flow cytometry.

Peripheral blood mononuclear cells obtained before the vaccination were labeled with the following monoclonal antibody combinations: fluorescein isothiocyanate (FITC)-conjugated anti-CD4 and phycoerythrin (PE)-conjugated anti-CD28, FITC-conjugated anti-CD8 and PE-conjugated CD28, or FITC-conjugated anti-CD4 and PE-conjugated anti-CD45RO (all from Becton Dickinson, San Jose, Calif.). The cells were analyzed on a Becton Dickinson FACScan flow cytometer, and 50,000 events were collected. The data were analyzed using WinMDI software (Joseph Trotter, Scripps Research Institute, La Jolla, Calif.).

Statistical analysis.

Logistic models and proportional odds models were used to identify variables that correlated with a decreased likelihood of a humoral vaccine response. In the logistic models, antibody titer increases of at least fourfold were defined as indicating a vaccine response. We first investigated univariately each potential independent variable (age, sex, baseline titer, CD4⁺ CD45RO⁺ T cells, CD4⁺ CD28^null T cells, and CD8⁺ CD28^null T cells) with each vaccine. Subsequently, we forced age, sex, and baseline titer in each model and allowed the three T-cell subsets to enter in a stepwise process. In the proportional odds models, individuals were ranked into four ordered groups, dependent upon how many of the three antigens they responded to with an increased antibody titer of at least fourfold. The Pearson product moment correlation was used to calculate the correlations between age and the three T-cell subsets. Analysis was done using SAS statistical software.

RESULTS

Influenza vaccine response in elderly adults.

One hundred fifty-three residents of several community-based retirement facilities were immunized with a trivalent inactivated influenza vaccine containing three hemagglutinin antigens. The humoral immune response to the vaccine was assessed by comparing hemagglutination inhibition titers before and 28 days after immunization. Most individuals already had a significant antibody titer before vaccination. Only 26 (for A/Texas/36/91), 29 (for A/Nanchang/933/95), and 41 (for B/Harbin/7/94) of the 153 individuals had a titer of <1 in 40. Results for the vaccine responses, defined as titer increases, are shown in Table 1. Overall, the majority of the vaccinated individuals had no or only marginal (titer increase of twofold) vaccine responses. There were variations in the potency of the three hemagglutinin antigens to induce antibody production. If a vaccine response was defined as a titer increase of at least fourfold, A/Texas/36/91 was immunogenic in only 26% of vaccine recipients, whereas 33% of individuals responded to B/Harbin/7/94 and 42% responded to A/Nanchang/933/95. Seventy of the 153 individuals (45.8%) were nonresponders and did not generate an immune response to any of the antigens. Only 26 of the 153 vaccine recipients (17.0%) had at least a fourfold increase in titer of antibody to all three antigens. Different influenza virus vaccine strains have various degrees of immunogenicity, raising the possibility that the low immunogenicity, in particular of A/Texas/36/91, is characteristic of the vaccine strains rather than of the population vaccinated. However, a recent study of young children vaccinated during the same season showed fourfold titer increases in 64% (A/Texas/36/91) and 82% (B/Harbin/7/94) of vaccinated children (17, 18).

TABLE 1.

Influenza vaccine response in elderly adults

Titer increase (fold)	% Responders
	B/Harbin/7/94	A/Texas/36/91	A/Nanchang/933/95
≤1	30.7	34.6	26.1
2	36.0	39.2	31.4
4	19.6	12.4	17.0
8	7.2	4.6	14.4
16	3.3	2.0	4.6
32	0.7	2.0	4.6
64	0.7	2.6	1.3
128+	2.0	2.6	0.7

Influence of age on vaccine responses.

Individuals enrolled in this study ranged in age from 65 to 98 years, with a mean age of 79.7 years. To address the question of whether immune responsiveness continued to decline with increasing age, a titer increase of less than fourfold was defined as a nonresponse, and the impact of age on responsiveness was determined. The data shown in Table 2 indicate that vaccine responses to all three antigens declined with age. The probability of a titer increase decreased by about 25% for each 5 years of age for each vaccine.

TABLE 2.

Influence of age on the humoral response to influenza vaccines

Vaccine	β (SE of β)	Odds ratio (95% CI)a	P
A/Texas/36/91	−0.061 (0.025)	0.74 (0.58–0.95)b	0.016
A/Nanchang/933/95	−0.067 (0.023)	0.72 (0.58–0.90)	0.004
B/Harbin/7/94	−0.055 (0.024)	0.76 (0.60–0.96)	0.020
All 3 vaccines	−0.067 (0.020)	0.72 (0.59–0.88)c	0.001

^aOdds ratios and confidence intervals (CI) are per 5 years of age. 

^bCalculated with the logistic model. 

^cCalculated with the proportional odds model. 

Because immunocompetence may be better reflected by a combined assessment of the immune responses to all three antigens than separate analyses for each antigen, vaccine responses were ranked using the following definition. Individuals who did not have at least a fourfold increase in titer of antibody to any of the three antigens were considered nonresponders (n = 70 [45.8%]). Poor (fourfold titer increase for one hemagglutinin, n = 36 [23.5%]), intermediate (fourfold titer increase for two antigens, n = 21 [13.7%]), and good (fourfold titer increase for all three antigens, n = 26 [17.0%]) responders were defined accordingly. Proportional odds analysis was performed. The probability of a vaccine response declined by about 25% for each 5 years of age (P = 0.001; odds ratio, 0.72). In contrast, the sex of the donor did not have an influence. Within the study cohort, 112 participants were female and 41 were male. Maleness was associated with a poorer vaccine response (odds ratio, 0.62; 95% confidence interval, 0.31 to 1.22), but this did not reach significance (P = 0.16).

Influence of baseline hemagglutination titers on vaccine responses.

It has been hypothesized that vaccine immunogenicity is reduced in chronically immunized individuals, accounting in part for the poor vaccine response in the elderly (16, 21, 37). Indeed, Smith et al. (41) have shown that the vaccine immunogenicity in individuals given different vaccinations is determined by the antigenic difference of the vaccine strains, i.e., vaccine responses will be reduced if the vaccine strain is closely related to previously used strains. To determine the impact of vaccination history and preexisting immunity on vaccine responses, prevaccination hemagglutination titers were correlated with vaccine responses, again defined as a fourfold titer increase. The mean prevaccination titers were as follows: A/Texas/36/91, 201 ± 447; A/Nanchang/933/95, 109 ± 104; B/Harbin/7/94, 127 ± 210. For all three vaccine components, the preexisting titer had a negative impact on vaccine responses (A/Texas/36/91: P < 0.001; odds ratio 0.54; 95% confidence interval, 0.42 to 0.70; A/Nanchang/933/95: P < 0.001; odds ratio, 0.49; 95% confidence interval, 0.36 to 0.66; B/Harbin/7/94: P < 0.001; odds ratio, 0.55; 95% confidence interval, 0.42 to 0.71).

Phenotypic T-cell markers and aging.

The T-cell compartment undergoes numerous changes during aging. Due to thymic involution and cumulative antigenic exposure, the number of memory (CD45RO⁺) T cells increases with a reciprocal decline in naive (CD45RO^null) T cells (29). In addition, replicative immunosenescence has been associated with a loss of cell surface expression of the costimulatory molecule CD28 on CD4 and on CD8 T cells (12, 42, 43). To assess whether expansion of any of these subsets correlated with compromised humoral immunity, peripheral T cells were analyzed for the expression of these cell surface markers. The frequency of CD45RO⁺ cells within the CD4 T-cell subset varied widely (mean ± standard deviation, 58% ± 17%; range, 14 to 99%). The age of the individuals did not account for this variability (Fig. 1) (R² = 0.001, P = 0.68), suggesting that at least for the age group older than 65 years there is no further increase in CD4⁺ CD45RO⁺-T-cell frequencies. CD4⁺ CD28^null T cells were present at frequencies from 0 to 46%, with a slight but not significant increase with older age of the donor (R² = 0.012, P = 0.18). The distribution was skewed, with a median frequency of 1.9%, suggesting that only a proportion of the cohort was prone to develop increased frequencies of CD4⁺ CD28^null T cells. The frequency of CD8⁺ CD28^null T cells increased with age (R² = 0.166, P <0.001); however, age could not fully account for the variability in the frequencies, which ranged from 5.5 to 95.3% (52.9% ± 23.7%). Regression analysis of the frequencies of the three cell surface types did not yield a significant correlation, suggesting that they represented independent variables.

FIG. 1.

Influence of aging on the frequencies of functional T-cell subsets. Peripheral blood mononuclear cells from vaccinated individuals were analyzed for the frequencies of three functional T-cell subsets previously reported to be increased in the elderly: CD4⁺ CD28^null, CD8⁺ CD28^null, and CD4⁺ CD45RO⁺. Frequencies are the percentages of total CD4 and CD8 T cells; boxes show medians and 25th and 75th percentiles, and whiskers show 10th and 90th percentiles. Frequencies of all three subsets varied widely; however, only for CD8⁺ CD28^null T cells was there a further increase with age after the age of 65 years (R² = 0.166, P < 0.001).

Correlation of T-cell subset frequencies and vaccine responses.

Frequencies of CD4⁺ CD45RO⁺, CD4⁺ CD28^null, and CD8⁺ CD28^null T cells were analyzed to determine whether they were predictive of poor antibody responsiveness to the influenza vaccine. Vaccine recipients were defined as responders when they had at least a fourfold increase in titer of antibody to the hemagglutinin antigen, and logistic odds analysis for each vaccine strain was performed. The odds ratios and 95% confidence intervals are shown in Table 3. Neither the frequency of CD4⁺ CD45RO⁺ T cells nor that of CD4⁺ CD28^null T cells influenced the humoral response. In contrast, higher frequencies of CD8⁺ CD28^null T cells correlated with a higher probability of nonresponsiveness to all three vaccines. The odds ratios for response to vaccination ranged from 0.75 to 0.82 per 10% increase in the frequencies of CD8⁺ CD28^null T cells. Given the wide range of CD8⁺ CD28^null-T-cell frequencies (5.5 to 95.3%) (Fig. 1), this marker appears to have considerable value for predicting compromised immune responsiveness. In a second analysis, only individuals with a preimmunization titer of <40 were included, and univariate logistic analysis was used to identify variables that predicted a titer increase to ≥1 in 40. None of the T-cell frequencies was significant. This analysis was based on rather small subgroups, and the definition of response was not very robust.

TABLE 3.

Univariate logistic models of humoral vaccine responses in the elderly

T-cell subset	Vaccine	β (SE of β)	Odds ratio (95% CI)a	P
CD4+ CD45RO+	A/Texas/36/91	0.001 (0.011)	1.01 (0.82–1.25)	0.894
	A/Nanchang/933/95	0.002 (0.009)	1.02 (0.85–1.23)	0.842
	B/Harbin/7/94	0.009 (0.010)	1.09 (0.90–1.33)	0.371
CD4+ CD28null	A/Texas/36/91	−0.013 (0.025)	0.88 (0.54–1.42)	0.597
	A/Nanchang/933/95	0.001 (0.020)	1.01 (0.68–1.51)	0.949
	B/Harbin/7/94	−0.018 (0.023)	0.83 (0.53–1.31)	0.424
CD8+ CD28null	A/Texas/36/91	−0.027 (0.008)	0.77 (0.65–0.90)	0.002
	A/Nanchang/933/95	−0.020 (0.007)	0.82 (0.71–0.95)	0.006
	B/Harbin/7/94	−0.029 (0.008)	0.75 (0.64–0.88)	<0.001

^aOdds ratios and confidence intervals (CI) are per 10% increase in T-cell subset frequencies. 

To more globally assess immunocompetence, vaccine responders were ranked as to whether they responded to none, one, two, or all three of the antigens with at least a fourfold titer increase, and proportional odds analysis was performed (Table 4). Again, CD8⁺ CD28^null- but not CD4⁺ CD28^null- or CD4⁺ CD45RO⁺-T-cell frequencies correlated with a poorer vaccine response. An increase of 10% in the frequencies of CD4⁺ CD45RO⁺ T cells had an odds ratio of 1.03 to predict a superior immune response, i.e., optimally an increase in titer of antibody to all three antigens. Similarly, an increase of 10% in the frequency of CD4⁺ CD28^null T cells had an odds ratio of 0.91, which was not significant. In contrast, the frequency of CD8⁺ CD28^null T cells correlated significantly with the vaccine response (P < 0.0001). An increase of 10% yielded an odds ratio of 0.76, i.e., individuals with high numbers of CD8⁺ CD28^null T cells had a low probability of responding to the vaccine. Multivariate regression analysis was used to test the hypothesis that the correlation of CD8⁺ CD28^null-T-cell frequencies and vaccine responses was a reflection of the age dependence of both variables. The frequencies of CD8⁺ CD28^null T cells continued to predict a poor vaccine response after correction for age and sex (odds ratio, 0.80; 95% confidence interval, 0.69 to 0.92; P = 0.002). Also, adjusting for baseline titers of all three antigenic components did not abolish the predictive value of CD8⁺ CD28^null-T-cell frequencies. In the proportional odds model, the frequency of CD8⁺ CD28^null T cells remained significant after adjustment for baseline titers, age, and sex (odds ratio, 0.81; 95% confidence interval, 0.69 to 0.94; P = 0.004).

TABLE 4.

Proportional odds model of humoral vaccine responses in individuals older than 65 years

T-cell subset	β (SE of β)	Odds ratio (95% CI)a	P
CD4+ CD45RO+	0.003 (0.009)	1.03 (0.87–1.22)	0.732
CD4+ CD28null	−0.009 (0.019)	0.91 (0.63–1.32)	0.634
CD8+ CD28null	−0.028 (0.007)	0.76 (0.66–0.87)	0.0001

^aOdds ratios and confidence intervals (CI) are per 10% increase in T-cell subset frequencies. 

DISCUSSION

Our study demonstrates that the ability of elderly individuals to respond to a trivalent influenza vaccine is low, with only 17% of vaccine recipients mounting an increase in titer of antibody to all three vaccine components. Among these aged individuals, 46% failed to generate an immune response to any of the hemagglutinins of a trivalent influenza vaccine. Within this age group (65 to 98 years), age negatively correlated with the induction of anti-influenza virus antibodies, providing evidence for a critical role of immunosenescence in compromising the effectiveness of vaccination. Data presented here suggest that measuring the frequency of CD8⁺ CD28^null T cells, a biological marker that can be easily assessed, provides useful information about the immunocompetence of aged individuals, particularly as it relates to their ability to generate vaccine-induced humoral responses. Confirmatory studies are needed to determine whether frequencies of circulating CD8⁺ CD28^null T cells can be developed into a reliable biological marker to identify individuals at risk for poor anti-influenza virus vaccine response. In addition, our observations could be helpful in designing novel mechanistic approaches for improving immunocompetence in the elderly.

Postvaccination hemagglutination inhibition antibody titers are generally accepted as inversely correlating with the susceptibility to infection, with a titer of ≥40 being considered protective (36). In this study, we wanted to assess the ability to respond to vaccination with a humoral immune response and therefore used the titer increase as the outcome variable. With this criterion, a large proportion of our study population did not respond to the vaccination, although the population excluded individuals with major comorbidity. Immunogenicity may vary between influenza virus vaccine strains, and indeed A/Texas/36/91 has been described as poorly immunogenic (16, 21, 37). Therefore, it cannot be excluded that elderly individuals may have a better response to other vaccine strains. However, it was striking that nearly half of the population did not mount a humoral response to any of the three vaccine strains.

Our findings on vaccine responses of the elderly are consistent with previous studies. In a community-based cohort study by Nichol et al. (33), vaccination was shown to prevent 48 to 57% of all hospitalizations for pneumonia and influenza. Similar results have been reported from several case-control observational studies. Thus, there is no question that vaccination with inactivated influenza virus is effective in a large subset of elderly individuals and is, therefore, cost-effective. It is also evident that the vaccine immunogenicity is far below what is desired to prevent the disease in this risk group. The frequency of CD8⁺ CD28^null T cells appears to be a good biological marker to identify individuals at risk, even after adjustments for other risk factors, such as age. It is obvious that this vaccine nonresponder population does not benefit from the vaccination.

In addition to the immunocompetence of the host, immune responses to vaccination are influenced by several other variables, in particular by preexisting immunity. Prevaccination titers varied substantially in our study population; the majority of individuals already had titers of 40 and more. Vaccinees with higher baseline titers have a reduced chance to respond to the vaccination and, therefore, do not develop immunity to the new antigenic epitopes unique for the vaccine strain (16, 21, 37). Smith et al. (41) recently analyzed this phenomenon using historical data from repeat vaccinations, including the relatedness of the vaccine antigens employed in subsequent years and the vaccination efficacy (3). Using computer simulation, they demonstrated that the similarity of the vaccine strain to previous strains inversely correlated with the efficacy of repeat vaccinations. Our findings are consistent with this model. For all three antigens, the preexisting antibody titer had a negative impact on the vaccine response. Interestingly, there was no difference in this respect between A/Texas, which had been used for vaccinations in previous years and had the highest prevaccination titers, and the two other vaccine strains. Most importantly, multivariate analysis showed that CD8⁺ CD28^null-T-cell frequencies remained predictive of a poor vaccine response even after adjustment for baseline titers.

The correlation between CD8⁺ CD28^null T cells and the defective antibody responses to vaccination may also give mechanistic insights. Few studies have tried to develop immune status indices that are predictive of immunosenescence. The tests either are of limited value or are complicated, e.g., skin testing for delayed-type hypersensitivity or in vitro T-cell assays (10, 47). The three tests selected here represent phenotypic markers of T-cell senescence that can be easily assessed by flow cytometry. CD45RO is a cell surface molecule that is expressed on T cells that have recognized antigen in the past (38). Therefore, it has been assumed that repeated exposure to antigens during a lifetime would result in an increase in the CD45RO⁺ population and lead to an exhaustion of the CD45RO^null (naive) population. The CD45RO⁺ (memory) population is at least 100-fold less diverse than the naive T-cell subset (2, 45). Therefore, an expansion of this subset at the expense of the naive population would compromise the ability to generate an immune response to new antigens, which is required to recognize the antigenic drifts of the influenza virus hemagglutinins. Indeed, animal data support the notion that CD45 isoforms may be useful markers of immunosenescence (30). Our studies have clearly shown that this is not the case.

The CD28 molecule transmits signals that are pivotal for T-cell activation and T-cell survival (23, 25). In the absence of CD28 triggering, T cells are rendered anergic, have a limited ability to proliferate, and are prone to apoptosis. With continuous replication, T cells tend to lose two DNA-binding proteins that are necessary for the transcription of the CD28 gene (43). The loss of these two binding proteins can be more easily induced in CD8 than in CD4 T cells. The emergence of CD8⁺ CD28^null T cells is, therefore, normal with aging and chronic infections. The frequency of CD8⁺ CD28^null T cells followed a Gaussian distribution, yet with a very wide range, indicating marked diversity in the elderly population. CD4⁺ CD28^null T cells are less frequent and emerge in only a subset of healthy individuals (26, 42). The finding that the frequency of CD8⁺ CD28^null T cells is predictive of a poor antibody response to the influenza vaccine may indicate that the loss of CD28 not only is a marker of immunosenescence but also is directly responsible. CD28-mediated costimulation has been shown to be crucial for the formation of germinal centers, where B-cell responses mature and antigen-specific high-affinity antibodies are generated. Mice treated with anti-CD28 antibodies and CD28 knockout mice have defective germinal centers (11, 39). In response to immunization, antibodies are not hypermutated and do not undergo affinity maturation, presumably because of defective T-helper-cell function. T-helper-cell activity has generally been attributed to CD4 T cells, which express cell surface molecules such as CD40 ligand and secrete cytokines that enable them to provide B-cell help (44). Therefore, it was surprising to find that a functional molecule on CD8 T cells correlated with humoral nonresponsiveness. CD8 T cells have been associated with antiviral effector activities rather than with auxiliary regulatory function (28). However, recent observations in patients with rheumatoid arthritis support a role of a distinct subset of CD8 T cells different from the usual cytotoxic CD8 T cell in germinal-center formation (19, 46). These studies await confirmation for normal humoral responses. Strategies to restore CD28 expression may, therefore, be useful for improving vaccine responsiveness in the elderly. Preliminary results from our laboratory have indicated that the loss and reexpression of CD28 are regulated by cytokines.

In summary, we have identified a biological marker that can be easily determined and correlates with a defective antibody response to inactivated influenza virus vaccine in the elderly. Individuals with high frequencies of CD8⁺ CD28^null T cells have an increased likelihood of an insufficient antibody response to the vaccine and may be at risk for infection despite being vaccinated. These individuals may be candidates for prophylactic treatment with neuraminidase inhibitors. Restoring CD28 expression through the use of the appropriate cytokines may be an elegant way to overcome the low immunogenicity of the inactivated vaccine in a subset of the elderly.