Lack of Antibody Production Following Immunization in Old Age: Association with CD8⁺CD28⁻ T Cell Clonal Expansions and an Imbalance in the Production of Th1 and Th2 Cytokines¹

The function of the immune system declines with age (1, 2). This leads to increased frequency and severity of infectious diseases and endangers the protective effect of vaccination (3, 4). Decreased Ab production and shortened duration of protective immunity following immunization are characteristic features in the elderly (5, 6). Despite intensive research work, many of the basic mechanisms of age-related immune dysfunction have not yet been clarified. This is presumably due to the fact that many published reports conflict on theoretically important points. For instance, the production of the type 1 cytokine IFN-γ has been reported to be high (7, 8) or low (9, 10, 11) during aging and multiple similar examples could be given (12). Controversial observations may result from variations among species, strains, organs, or culture systems, but may also be due to interindividual differences in the course of the biological aging process. To circumvent the problems of heterogenous cohorts and large deviations, we have made use of the fact that influenza vaccination induces protective Abs in only 40 to maximally 70% of the elderly (13, 14) and selected a very homogenous “model” cohort of elderly persons, none of whom produced specific Abs 1 mo after influenza vaccination. Other vaccination types might have been equally suitable to define elderly persons with a lack of Ab production following immunization, but influenza vaccination is most frequently performed and easy to monitor.

Using this special cohort, we demonstrate that lack of Ab production following immunization is associated with the increased occurrence of expanded autoreactive CD8⁺CD28⁻ T cell clones. These clones, which are frequently CD45RA⁺, produce large amounts of IFN-γ which might be the basis for a change in the polarization of the immune system in elderly persons and the development of age-related immune deficiency.

A total of 27 elderly (>60 years), apparently well, mentally alert persons who live in the community were included in the study. These persons had been selected from a cohort of 78 elderly persons on the basis of their Ab response to influenza vaccination with a commercially available trivalent split vaccine (Vaxigrip; Aventis-Pasteur, Paris, France) that contained influenza virus of the A/Sydney/184/93, the A/Beijing/262/95, and the B/Beijing/184/93 strain (similar to B Harbin). A total of 13 persons (6 females, 7 males, mean age 68 ± 9; range 61–89 years) were chosen who had failed to raise a humoral immune response to all three influenza strains present in the vaccine and had Ab titers below the protective level of 1:40 1 mo after vaccination. These elderly persons are in the whole manuscript referred to as “old nonresponders” (ONR)³. Aged persons (5 females, 9 males, mean age 67 ± 7; range 61–85 years) who did not have influenza-specific Abs before vaccination, but titers of ≥1:40 to all three influenza strains 4 wk after vaccination were chosen as a humoral responder group and are in the manuscript referred to as “old responders” (OR). Six people from the ONR and seven from the OR group had previously been immunized against influenza. Six young healthy adults (Y; 3 females, 3 males, mean age 31 ± 3; range 28–35 years), who did also not have influenza-specific Abs before, but Ab titers of ≥1:40 1 mo after vaccination were chosen as a young control group. Four of the six young people had previously been immunized against influenza. All participants had given their informed written consent and the study was approved of by the local ethical committee.

Peripheral blood was taken by venipuncture at three different time points, before as well as 1 and 12 mo after influenza vaccination. Serum samples (5 ml) were obtained before as well as 1 mo after vaccination. Before bleeding, a health check was performed in each participant. Preparation of PBMC was performed by Ficoll-Paque gradient centrifugation as previously described (15). Purified cells were stored in liquid nitrogen until pre- and postvaccination Ab titers had been assessed and the final selection of participants had been made. PBMC samples obtained before vaccination were used to perform the full set of experiments in all donors. In two to four people from each group, complementarity-determining region (CDR)3 spectratyping and phenotypic analysis were repeated using cells obtained 1 and 12 mo after vaccination.

Ab responses to the vaccine hemagglutinin components were determined by standard hemagglutination inhibition assay as described previously (15).

The following reagents were used for in vitro cultures: FCS (SEBAK, Stuben, Austria), RPMI 1640 (Life Technologies, Grand Island, NY), PHA (20 ng/ml; Sigma-Aldrich, St. Louis, MO), live influenza virus (kindly provided by Berna, Swiss Serum and Vaccine Institute, Berne, Switzerland), live CMV (kindly provided by Dr. H. P. Huemer, University of Innsbruck, Innsbruck, Austria), live EBV (diluted supernatant; kindly provided by D. Kraft, University of Vienna, Vienna, Austria), and OKT-3 (20 ng/ml; American Type Culture Collection, Manassas, VA).

Cells were preincubated with an Fc block (mAb to CD16-CD32, 2.4G2; BD PharMingen, San Diego, CA) and washed in FACS buffer (PBS, 0.5% BSA, 0.01% sodium azide). Next, they were incubated with specific Ab directly conjugated as described below, washed, and analyzed using FACScan, FACSort, and FACSCalibur cytometers; and data were analyzed with CellQuest software (BD Biosciences, San Jose, CA). The mAbs used were purchased from BD PharMingen: CD4 (CyChrome-, PerCP-, or allophycocyanin-labeled), CD8 (CyChrome-, PerCP-, or allophycocyanin-labeled), CD3 (PE-labeled), CD11a (PE-labeled), CD28 (FITC- or CyChrome-labeled), CD45RO (FITC- or PE-conjugated), CD45RA (FITC- or CyChrome-labeled), CD95 (FITC-conjugated), CD16 (PE-labeled), Vβ1 (PE-labeled), Vβ5 (PE-labeled), Vβ9 (FITC-labeled), and Vβ17 (PE-labeled).

CD4⁺ and CD8⁺ T cells were purified as follows. PBMC were passed through a nylon mesh. CD4⁺ and CD8⁺ T cells were then purified by a positive selection procedure using anti-CD4⁺/CD8⁺-labeled microbeads and the MACS system (Miltenyi Biotec, Bergisch Gladbach, Germany) (purity: 97 ± 3% CD4⁺/CD8⁺/CD3⁺ cells, range 92–100%). For the isolation and characterization of expanded T cell clones, CD4⁻ cells were stained with anti-CD8 and Vβ-specific anti-TCR Abs and double-positive T cells were sorted with a FACS sorter. Only clones that dominated one Vβ family in a way that no other clones were visible (diversity score 3, see CDR3 spectratyping) were selected for purification. The selected T cells were then washed, phenotyped, and placed into culture. The clonal nature of the selected cells was confirmed by immunoscope technology and sequencing.

Total RNA was extracted from T cells using Tri Reagent (Sigma-Aldrich). A total of 1 μg of total RNA was used for first-strand cDNA synthesis using a Reverse Transcription system (Promega, Madison, WI). TCR Vβ transcripts were amplified by PCR using primers (Life Technologies, Vienna, Austria) specific for each of the human Vβ families and a specific primer for the constant region of the β-chain (labeled with the fluorescent dye marker 6-FAM).

An aliquot of the PCR product was diluted in 16 μl deionized formamide and 1.2 fmol internal lane standard GeneScan-350 Tamra (PerkinElmer, Norwalk, CT). The samples were denatured at 90°C for 2 min and snap cooled on ice before loading on a CE 310 Genetic Analyzer (PerkinElmer). Each sample was injected for 5 s at 15 kV and electrophoresed for 24 min at 10 kV using a 36-cm capillary and POP4 (PerkinElmer). Analysis of the raw data was performed applying the GeneScan 2.1 analysis software package (PE Applied Biosystems, Foster City, CA) using the Local Southern method for fragment size estimation. In the case of a normal distribution of individual clones within the Vβ family, a Gaussian profile was depicted. Deviations from the Gaussian profile indicated the presence of large expanded clones. The occurrence of dominant clonal expansions within the different Vβ families was quantified by using a diversity score between 1 and 3; 1 was assigned to a Gaussian distribution, 2 corresponded to a pattern with one to three peaks above the Gaussian background, and 3 corresponded to one predominant peak above the Gaussian distribution (Fig. 1). The mean of all scores assigned to the clonal distributions within the different Vβ families was then calculated for each individual blood donor for CD4⁺, as well as CD8⁺ cells. These individual scores were then used for further analysis, when the clonal diversity of CD4⁺ and CD8⁺ cells in the different groups (Y, OR, ONR) was assessed. The classification of CDR3 spectratypes as Gaussian, oligoclonal, or monoclonal was done “blind” by three different persons. The ratings were then compared. In the rare case of discrepancies, the assessment made by two of the three investigators was used for further calculations. Three different opinions never occurred.

Individual Vβ families that were seemingly dominated by one expanded clone were selected and amplified by PCR. The clonal nature of the observed expansions was confirmed by DNA sequencing. Purified PCR products were ligated with pCRScriptAmp SK vector (Cloning kit; Stratagene, La Jolla, CA) for 60 min at room temperature. One Shot competent cells (Invitrogen, San Diego, CA) were used for the transformation. After PCR screening, positive colonies were selected and expanded. Plasmids were purified, digested with restriction enzymes, and the inserts were sequenced using an ABI Prism Genetic Analyzer CE 310 (PE Applied Biosystems) DNA sequencer. The sequence data were analyzed using GeneScan Analysis Version 2.1 software.

FACS sorted and isolated T cell clones were cultured together with equal numbers of autologous irradiated (35 Gy) PBMC as APC at a density of 5 × 10⁴ cells/well in 96-well plates in the presence or absence of different stimuli as previously described (16). Cells and conditioned supernatants were harvested after 6 days.

Purified CD4⁺ T cells were seeded at a density of 10⁶ cells/well in 24-well plates (Falcon; BD Biosciences) together with 10⁶ autologous irradiated (35 Gy) PBMC and stimulated with PHA. Cells and conditioned supernatants were collected after 6 days for cytokine analysis.

Cytokine concentrations in conditioned supernatants were assessed by commercially available ELISA kits (CYTELISA; Cytimmunesciences, Vienna, Austria). IFN-γ was analyzed as a characteristic type 1, and IL-5 as a characteristic type 2 cytokine.

All statistical analyses were based on Student’s t tests for paired and unpaired data. Differences in Ab titers between the groups were assessed by Fisher’s exact test.

Elderly persons were selected who did not produce protective Abs 1 mo after influenza vaccination. This group is in the following referred to as ONR. Two groups were chosen as controls: elderly persons who did produce protective Abs to influenza following vaccination (OR), and a cohort of young persons (Y) who had a satisfactory humoral immune response to influenza vaccination. None of the persons studied had Abs specific to the three influenza strains present in the vaccine before vaccination. Postvaccination Ab titers in the respective groups are listed in Table I.

The clonal composition of the different CD8⁺ and CD4⁺ TCR families was studied in the three donor groups. Immunoscope technology was applied. A representative example of results obtained with this technique is depicted in Fig. 1,a. In the CD4⁺ population, Vβ subfamilies had a Gaussian or close to Gaussian CDR3-size distribution pattern in all three groups. Clear deviations from the Gaussian profile were only noted in the Vβ families in the CD8⁺ population. Pronounced differences among the groups were observed. Although the Gaussian profile was most frequently observed in young persons, deviations from the Gaussian profile were often present in both elderly groups, but were more pronounced in the ONR group (Fig. 1, a and b). All Vβ families were equally affected. Single large expanded clones dominating a Vβ family (diversity score 3) were detected in 2 of 6 (33%) of the young control persons, and in each person in both elderly groups. However, their frequency was higher in the ONR group, in which the shift affected a large majority of Vβ families in all but one person. In this person, 50% of the Vβ families were affected. Fifty percent of the Vβ families were also dominated by single peaks in one member of the OR group, while the frequency of expanded clones was lower in all other members of this group. One of the two young people who had dominant peaks in their CD8⁺ repertoire showed a pattern similar to elderly responders, while dominant peaks were very rare in the other one. In two donors from each group, CDR3 spectratyping was repeated 1 mo and 1 year after vaccination. However, no alterations from the original pattern were observed in any of the samples studied.

To learn more about the phenotypic and functional characteristics of expanded T cell clones, clones which dominated a Vβ family and were also stably detected over a year (Fig. 2,a) were isolated by cell sorting (Fig. 2,b). Seven clones from five donors (one Y, two OR, and two ONR) were obtained. Characterization of the sorted cells by immunoscope technology always demonstrated one large peak identical in size to the one observed before sorting (Fig. 2,c). The identity of the clones at different time points as well as before and after sorting was additionally confirmed by cloning and sequencing. All sorted clones expressed CD11a and CD95 at high density (Fig. 3,a). However, they did not express CD28. Five of the seven clones (71%) were CD45RA⁺, and two (29%; one from a donor from the Y, the other one from a donor from the OR group) contained CD45RA⁺ as well as CD45RO⁺ cells. The sorted clones retained their phenotype in cell culture in the absence of stimuli over an observation period of 2 wk. During this time the cells stayed viable, but failed to increase in number. Cytokine secretion analysis demonstrated that the clones produced IFN-γ, but failed to secrete the type 2 cytokine IL-5 in response to stimulation with irradiated autologous PBMC (Fig. 3 b). Thus, they were autoreactive, although due to the low numbers of sorted cells available, the exact nature of the autoantigen could not be defined. IFN-γ, but no IL-5, was also secreted when the cells were stimulated with OKT-3; but no cytokines were produced when irradiated allogeneic PBMC were used as stimulus. No increase in the production of IFN-γ by the clones took place when the autologous PBMC which were used as APC were pulsed with a panel of different Ags including influenza virus, CMV, or EBV. IFN-γ production was not accompanied by proliferation. There was no difference in the cytokine secretion pattern between clones from the Y, the OR, and the ONR donor groups, or between CD45RA⁺ clones and clones that also contained CD45RO⁺ cells (data not shown). These data indicated that the phenotypic and functional characteristics of the expanded clones were more or less the same, but that their frequency was different in the three donor groups.

To define whether the predominance of expanded clones observed in the ONR group was reflected in the phenotype of PBMC, we analyzed the expression of CD28 and of CD45RA and CD45RO on peripheral CD4⁺ and CD8⁺ lymphocytes. Although the percentage of CD4⁺CD28⁻ cells was low in all three groups, CD8⁺ cells were frequently CD28⁻, and there was a clear cut difference in the number of CD8⁺CD28⁻ cells among the groups (Fig. 4,a). Within the CD8⁺CD45RO⁺ population, CD28⁻ cells were more frequent in both the OR and the ONR group than in the young group. In contrast, the percentage of CD28⁻ cells within the CD8⁺CD45RA⁺ population was markedly higher in the ONR than in the OR group. CD8⁺CD45RA⁺CD28⁻ cells coexpressed CD11a at high intensity (Fig. 4 b). They did not express CD16. Thus, they were most likely effector T cells with a CD45RA phenotype, and corresponded to the predominant clonotypes described above. Phenotypic analysis was repeated 1 mo and 1 year after vaccination in two to four donors from each group. No significant changes from the original prevaccination samples were observed in these donors.

As the predominance of IFN-γ producing autoreactive CD8⁺ T cell clones may lead to a polarization of immune responses in the direction of Th1 responses, CD4⁺ T cells were studied for their cytokine production. CD4⁺ cells were purified from PBMC and stimulated with PHA. Conditioned supernatants were tested for the presence of IFN-γ and IL-5. Following PHA stimulation, the secretion of IL-5 was similar in the Y and in the OR group, but significantly lower in the ONR group (Fig. 5). In contrast, IFN-γ production did not differ among the groups, suggesting an imbalance in the production of Th1 and Th2 cytokine in elderly persons, who fail to produce Abs following influenza vaccination.

Our results show that expanded CD28⁻ T cell clones dominate the CD8⁺ repertoire in persons who fail to produce protective Abs following influenza vaccination. These clones produce large amounts of IFN-γ upon autoantigenic stimulation. However, they do not produce IL-5. Frequent and ubiquitous restimulation by autoantigen could be a reason for the persistence of these clones in vivo (Fig. 3), and represent a possible basis for a relative overproduction of IFN-γ in old age. Many studies which consistently report that neopterin (a macrophage product the production of which is exclusively triggered by IFN-γ) is elevated in the elderly support this possibility (17, 18, 19, 20). Measuring the production of Abs of the IgG2a istoype could be another way to assess the functional consequences of IFN-γ production in the elderly. A high type 1 cytokine production could change the cytokine microenvironment in lymphatic tissues, and thus be responsible for a decreased production of type 2 cytokines by CD4⁺ T cells (Fig. 5). Type 2 cytokines act on B cells to induce activation and differentiation, whereas low type 2 cytokine production is likely to hamper B cell propagation, and thus may be responsible for a decreased production of Abs following immunization in the elderly. A recent study in aged mice, which fail to raise a Th2 response following in vivo challenge, is also in favor of this concept (21). It would also be of interest to test whether CD28⁻ clones might have a direct effect on the differentiation and polarization of CD4⁺ cells. Experiments presently underway in our laboratory, in which the cytokine production pattern of cocultured CD4⁺ and CD8⁺ T cells is tested, will help to elucidate this possibility.

Th1 polarization may not only be a problem in the context of decreased Ab production in old age, but may also be detrimental for other reasons. Proinflammatory cytokines are now believed to exacerbate the functional pathology and disease course of age-related disorders, such as Alzheimer’s disease (22, 23) and atherosclerosis (24). IFN-γ triggers the production of the Alzheimer β amyloid (Aβ) in combination with TNF-α by human neural and extraneural cells (25), and increases the production of oxygen radicals by microglial cells (26). TNF-α can also increase the toxicity of Aβ (27) as well as stimulate smooth muscle cell proliferation (24), a key event in the development of atherosclerosis.

Although the etiology of CD28⁻ T cells is not fully understood in healthy individuals, there is strong evidence that these cells have arisen in response to continued antigenic stimulation (28, 29). CD28 expression is characteristically lost after many rounds of cell division (30), and the strong expression of CD11a on CD28⁻ cells (Figs. 3 and 4 b) is also consistent with their prior activation by Ag. Telomere shortening has also been described in CD28⁻ T cells and is more pronounced than in CD28⁺ cells from the same donor (31, 32). This also indicates that the former have undergone more rounds of cell division than the latter.

It is striking that CD28⁻ memory/effector cells mostly have a CD45RA phenotype in people who fail to produce specific Abs following influenza vaccination. CD28⁻CD45RA⁺ memory/effector T cells have previously been described to occur at increased frequency in old age, but their origin and significance were unclear (7, 33). It seems now likely that the CD28⁻CD45RA⁺ phenotype is consistent with a state of terminal effector cell differentiation (34), although its role as a reservoir of long-lived memory cells has also been discussed. Characteristics such as a loss of growth potential (7) and an increased resistance to apoptosis-inducing stimuli (35, 36), which are reminiscent of the well-defined phenotype of senescent fibroblasts or keratinocytes (37, 38), do in any case suggest that CD28⁻CD45RA⁺ T cells can be considered as truly “old.”

The accumulation of expanded CD28⁻ clones of old effector/memory T cells in elderly persons could lead to a restriction of the space still available for functioning T cells. This could be another cause for the development of age-related immune deficiency (39). In this context, it seems of interest that predictions of longevity have been made based on the ratio of naive vs memory T cells. Thus, mice characterized by relatively low levels of CD4 and CD8 memory cells and high levels of CD4 naive cells lived longer than conventional controls (40). No information is presently available on the presence or absence of clonal expansions in the CD8⁺ T cell pool of these long-lived mice, but age-related clonal T cell expansions have also been demonstrated in rodents (41). These animal studies in combination with our present data support the concept that the number of memory cells, or more precisely the number of CD28⁻CD45RA⁺CD8⁺ T cells could be used as a biomarker of immune senescence in middle-aged and elderly humans. Work presently underway in our laboratory will demonstrate whether aged persons with a high proportion of CD28⁻CD45RA⁺CD8⁺ T cells in their peripheral blood are also likely to have a low humoral immune response to vaccines such as tetanus, diphtheria, or rabies.

In conclusion, our data illustrate the accumulation of CD8⁺CD28⁻ IFN-γ producing T cells in the aging immune system. Due to their special properties, these cells may drive a Th1 polarization. Of course, we realize that our data make only an indirect case for cause and effect at this time and that we still have to prove that poor in vivo Ab responses are due to effects of the CD28⁻ cells on CD4 cell differentiation. The possibility that CD8⁺ clones and low Ab production may both be due to the same underlying alteration that contributes to differences in several signs of immune aging cannot yet be ruled out. It is still tempting to speculate that an imbalance in the production of pro- and anti-inflammatory cytokines could diminish the chances of elderly persons to be protected from infectious diseases, and increase their likelihood to develop age-related disorders, such as Alzheimer’s disease and atherosclerosis.

We thank P. Beverley, C. Franceschi, J. Evans, and M. Weksler for many stimulating discussions; and B. Jenewein, R. Muehlmann, and A. König for their excellent technical assistance.