Whereas multiple mechanisms may contribute to lymphomagenesis in HIV-infected persons, there are two major mechanisms that appear to be involved in the development of these cancers: loss of immunoregulatory control of Epstein–Barr virus (EBV) and/or HHV8, and chronic B-cell activation due to the immune dysfunction resulting from HIV infection [14, 27]. Many forms of AIDS-NHL are characterized by the presence of recurrent genetic alterations, which may be due, in part, to errors in normal processes that occur in activated B cells, which involve modification of somatic DNA, such as immunoglobulin gene (Ig) class-switch recombination (CSR) and somatic hypermutation (SHM). Activation-induced cytidine deaminase (AID) is a DNA-editing enzyme normally active during B-cell activation, which plays a central role in both Ig SHM and CSR [28]. AID also promotes c-MYC/IgH translocations, and is required for the genesis of lymphomas of GC origin [29]. Moreover, AID is able to produce DNA double-strand breaks, resulting in widespread genome instability, which may contribute to some of the non-Ig-related genetic modifications that are present in AIDS-NHL [29]. Notably, AID expression is elevated in peripheral blood mononuclear cells (PBMCs) of HIV-positive individuals prior to the diagnosis of NHL [30]. Therefore, changes in HIV infection that result in chronic B-cell hyperactivation have the potential to contribute to the genesis of AIDS-NHL, particularly those forms that are EBV-negative, and are therefore not driven by the expression of EBV-encoded oncogenes.

It is not known whether effective HAART, which results in profound plasma HIV RNA suppression and consequent partial restoration of immune competence, allows normalization of serum levels of these markers of immune activation and inflammation, or of B-cell activity. It has been suggested by Biggar [31] that the continued increased risk for NHL among HAART-treated HIV-infected persons may be a reflection of incomplete immune reconstitution, and possibly a chronic state of B-cell hyperactivation. The continuation of immune activation and inflammation following HAART initiation may contribute to ongoing B-cell hyperactivation, accounting, in part, for a persistent incidence in the post-HAART era of those AIDS-lymphoma subtypes associated with chronic B-cell hyperactivation (mainly EBV-negative, systemic NHL) [14, 31, 32].

Table 1 describes the characteristics of the study population at the first MACS visit in which HAART use was reported, defined as the date of HAART initiation. The median date of HAART initiation was July 1997, the mean age was 43.8 years, and the median CD4 and HIV RNA levels were 296 cells/µl and 21 906 copies/ml, respectively. Sixteen percent (n = 47) of the study group had been diagnosed with AIDS, and the most common first AIDS-defining illness was Pneumocystis carinii (jiroveci) pneumonia (PCP). The NHL subgroup tended to be older, with higher baseline HIV RNA levels, and lower CD4. In this group, the mean age was 48.5 years, and the median CD4 and HIV RNA levels were 167 cells/µl and 88 080 copies/ml, respectively.

Table 2 describes selected population characteristics at each measurement time point. The majority of measurements occurred between 1995 (T1) and 1999 (T4). Median pre-HAART CD4 and HIV RNA levels at the measurement closest to HAART initiation were 322 cells/µl and 27 276 copies/ml, respectively, with the median CD4 cell count increasing to 474 cells/µl and nondetectable HIV RNA levels apparent 1.5–3 years after HAART initiation. For persons who were antiretroviral therapy (ART) experienced prior to HAART initiation, nondetectable HIV RNA levels were achieved in 49% (T3) and 53% (T4) of persons after starting HAART, compared with approximately 80% of those study participants who were ART-naive prior to HAART. About half of the participants were not receiving any antiretroviral therapy at the measured pre-HAART visits.

As shown in Fig. 1a–g, we observed a heterogeneous pattern of change in crude biomarker levels over time. Compared with the HIV-seronegative controls, after HAART initiation there was a reduction, but not normalization, of all biomarkers assayed with the exception of IL-6 and CRP. IL-6 remained elevated, and CRP marginally elevated, across all four time points, and the serum levels of these molecules appeared to be unaffected by HAART use.

Similar results were seen in the subgroup of men who developed NHL. Pre-HAART biomarker levels tended to be higher than in the total study group. However, their response to HAART was similar to that seen in the larger group: comparing results from the latest post-HAART (T4) to the earliest pre-HAART (T1) time-point, median [interquartile range (IQR)] sCD27 decreased from 581 (382–697) to 320 (276–525) U/ml; sCD30 from 92 (80–123) to 48 (28–88) U/ml; CXCL13 from 272 (133–489) to 121 (72–228) pg/ml; IgG from 34 257 (20 629–32 323) to 22 845 (20 780–24 564) ng/ml; and IgA from 984 (736–1651) to 870 (408–1114) ng/ml. IL-6 remained relatively stable and CRP showed an increase over time: IL-6 from 3.23 (2.03–3.82) to 2.82 (2.00–5.13) pg/ml, and CRP from 1.80 (1.22–2.31) to 3.13 (2.02–5.84) µg/ml.

Subgroup analyses of 100 men with effective viral responses and post-HAART CD4 cell count at least 350 cells/µl showed a similar heterogeneous pattern: at 2–3 years after HAART initiation, 29% of sCD27, 48% of CXCL13, and 51% of IgG analyte levels continued to exceed reference limits, whereas sCD30 approached normalization.

To determine if the biomarkers tested displayed distinct patterns of co-expression, we assessed the correlation of serum levels of these molecules. Levels of sCD27 and sCD30 were strongly correlated with each other (r = 0.44, P < 0.001). Both of these molecules are members of the TNF receptor superfamily, are associated with immune activation, and demonstrate functional similarities [33, 34], so the strong correlation noted is consistent with their biological properties. Additionally, both sCD27 and sCD30 were moderately correlated with IgG (r = 0.32, P < 0.001 and r = 0.29, P < 0.001, respectively) and with IgM (r = 0.20, P < 0.001 and r = 0.30, P < 0.001, respectively), consistent with these soluble receptors being associated with enhanced B-cell activity [33, 34]. Serum levels of CXCL13 were seen to correlate with both sCD27 (r = 0.18, P < 0.001) and sCD30 (r = 0.42, P < 0.001), as well as with IgG (r = 0.22, P < 0.001) and IgM (r = 0.19, P < 0.001), consistent with the B-cell-stimulatory properties of this chemokine [15, 42]. There was a moderate correlation between IgG and IgA (r = 0.26, P < 0.001), and a strong correlation between IgG and IgM (r = 0.47, P < 0.001). As expected, there was a moderate correlation between IL-6 and CRP (r = 0.21, P < 0.001). This is consistent with the known role of IL-6 in the induction of the acute-phase response [26], which includes the elevated production of CRP.

Results of multivariable analyses, controlling for baseline CD4 and HIV RNA levels, CD4 slope, and AIDS status pre-HAART are shown in Table 3 and illustrated in Fig. 2. In addition, as age was shown to be independently associated with IL-6 and CRP, we controlled for age in the analyses of these two biomarkers. Results are depicted for the total population (Fig. 2a), stratified by HIV viral response (Fig. 2b), and stratified by ART status at HAART initiation (Fig. 2c). Separate models were run for each immune activation biomarker. There was a statistically significant effect of HAART on all biomarkers except IL-6, CRP and IgA. These ranged from a post-HAART decrease of 17.5 (10.7–24.3)% for IgG to 41.6 (31.7–51.5)% for CXCL13 in the total population, 19.6 (7.2–31.9)% for IgG to 55.8 (38.7–72.9)% for CXCL13 in viral responders, and 29.3 (14.2–44.3)% for IgG to 62.5 (45.0–79.9)% for CXCL13 in those naïve to ART prior to HAART initiation.

Highly active antiretroviral therapy was seen to be associated with a significant decrease in serum levels of sCD27 and sCD30, even in the absence of an apparent viral response. This observation may be explained either by a direct effect of antiretroviral drugs on the immune cells that produce these molecules, or more likely, by a HAART-mediated restoration of immunity resulting from virological suppression, even when this suppression was suboptimal. In support of this, a substantial rebound in CD4 numbers was seen, in the absence of adequate virologic response (i.e. even among persons who achieve only partial viral suppression on HAART), as shown in Fig. 3. Among participants who were considered viral nonresponders post-HAART CD4 cell count, compared with pre-HAART levels, increased by 61 (21–101) cells/ml to a median (IQR) of 364 (230–537) cells/µl 2–3 years post HAART (P = 0.003).

Baseline CD4 cell count appears to modify the effect of HAART use on CXCL13 level, where higher CD4 was associated with a greater biomarker response to HAART. When stratifying by baseline CD4, a decrease of 34.8 (14.8–54.8) % was seen in CXCL13 in those with baseline CD4 cell count below 200 cells/µl; in contrast, a decrease of 48.2 (34.1–62.3)% was seen in those with baseline CD4 cell count at least 350 cells/µl.

Only one-third of study participants achieved immediate and sustained virologic responses. In the subgroup that did not achieve optimal HIV suppression, increases in CD4 cell number were still apparent, with significant reductions in serum sCD27 and sCD30 levels. Results of subgroup analyses suggest that the effects of HAART on serum biomarkers was most pronounced in those persons who were naïve to ART prior to HAART initiation, consistent with previous studies [43]. Therefore, persons who have experienced therapeutic failure with prior nonfully suppressive ART regimens, with resultant accumulation of drug resistance, may have a decreased probability of immune normalization with subsequent regimens.

There is a paucity of data examining the effect of HAART on the biomarkers measured in this study, and no published study has measured this group of biomarkers simultaneously. Atlas et al. [44] and De Milito et al. [45] examined the effect of HAART on plasma sCD27 levels. Atlas et al. reported results from a study of 64 persons (primarily of African ethnicity living in Sweden) with mixed HIV-1 subtypes, who were ART-naive at HAART initiation and experienced effective viral responses to HAART. When comparing results from a subgroup infected with HIV-1 subtype B, the predominant viral subtype in North America, to our results, Atlas et al. showed a more modest 9% decrease in sCD27 levels. De Milito et al. found that HAART-induced significant and progressive reductions, but not normalization, of plasma sCD27 levels in 26 HIV-infected persons after 27 months.

The results of our study are consistent with a previous pilot study of 34 MACS participants that examined the effect of HAART on CXCL13 serum levels [15], which reported a 21% decrease in CXCL13 serum levels after 1 year of HAART. Our results, in a larger study population, demonstrated greater than 40% decreases in CXCL13 after 2–3 years of HAART. Interestingly, baseline CD4 cell count appeared to modify the effect of HAART use on CXCL13 level, with higher CD4 being associated a greater decrease in CXCL13 levels following HAART initiation. CXCL13 is produced by follicular helper T cells (T_FH), which is the T-cell subset that interacts with activated germinal-center B cells in secondary lymphoid organs [35, 42]. Little is known about the effect of HIV infection on T_FH cells, or about how these cells are normally regulated. A recent study identified a novel regulatory CD8⁺ T-cell subset in mice, which can suppress the activity of T_FH cells [46]. When these CD8⁺ regulatory T cells were disrupted, this resulted in the enhanced production of auto-antibodies and in autoimmunity, and in a 10-fold increase in the size of germinal-center area [46]. It is interesting to speculate that more advanced HIV disease, for which low CD4 number can act as a marker, results in a more pronounced loss of such regulatory T-cell subsets, making it more difficult for HAART to restore the activity of such cells and dampen T_FH activity, thereby resulting in a less marked decrease in CXCL13 levels post HAART. Clearly, additional research is needed to better define the effect of HIV infection, and of HAART, on T_FH cells and on potential regulatory T-cell subsets that may act on those cells.

Several studies have examined the association between HAART and hypergammaglobulinemia [47–50]. Overall, these studies report a reduction in IgG with HAART use, and in some cases, normalization. Our results are consistent with these prior studies, but show much more modest decreases in serum IgG levels. In a case–control study, Chong et al. [47] reported 65% lower IgG levels among virologic responders who were compared with an ART-untreated group. Notermans et al. [48] reported complete normalization of IgG in 13 persons with virologic responses to HAART after 1.5 years of therapy, using a population-based reference range determined for the specific methodology in the testing hospital. Jacobson et al. [50] and Redgrave et al. [49] reported findings that were fairly consistent with ours, noting significant decreases, but not normalization, of IgG levels in persons who achieved effective HAART responses, with 45% (Jacobson et al.) and 37% (Redgrave et al.) of IgG values continuing to exceed reference limits.

Our results indicate that serum CRP levels were unaffected by HAART, and in fact appear to increase over time following HAART initiation. A similar pattern was seen in a study by Lau et al. [37], who reported a general increase in CRP over time in untreated HIV-infected men in the MACS, with a 4.5 and 8% increase per year for men who remained AIDS-free and for men who progressed to AIDS, respectively.

Overall, these findings suggest that HAART resulted in near normalization of some biomarkers, whereas others continue to be markedly elevated post-HAART. It is possible that such sustained immune system alterations occurring among HAART recipients may lead to ongoing or new pathological processes. It is also possible that immune dysfunction is resolving slowly, and will eventually normalize over time, with resultant normalization in markers of immune activation and inflammation. However, our findings do not support this, in that we saw fairly consistent results between the two post-HAART measurements. Perhaps more extended longitudinal study of immune activation biomarkers among successfully treated HAART recipients is warranted, to determine whether and when full immune function is restored. It would be expected that the longer a state of chronic B-cell hyperactivation exists, the greater the risk for accumulation of genetic errors that can lead to the development of AIDS-related NHL. Therefore, even if immune normalization is ultimately achieved, it would be expected that the excess risk of NHL among HIV-infected persons would be reduced, but not eliminated, with HAART use. This is consistent with the results of epidemiologic studies, which have noted whereas HAART is associated with a decreased incidence of AIDS-NHL, a significant elevation in NHL risk remains in HIV-positive persons receiving HAART [31, 32, 51].

Previous studies have shown that although HAART results in viral suppression to undetectable levels in peripheral blood, viral eradication is not achieved, even after years of continuous therapy [52], and there is quick rebound of plasma viremia with therapy interruption. Viral persistence is thought to be the result of long-term survival of infected memory CD4 cells, permitting a steady-state low level of viral replication in lymphoid tissue that is not detectable in the peripheral blood [53]. This low-level HIV replication may contribute to persisting B-cell hyperactivation in those on HAART. The progressive increases in serum CRP levels seen in this study suggest ongoing immune activation may continue post-HAART.

We found no apparent effect of HAART upon IL-6 levels. The SMART study, a study of HAART administration, with one arm based on minimizing drug use using target CD4 number (drug conservation arm) and another aiming at maximal suppression of HIV plasma levels (viral suppression arm), found that IL-6 levels increased by 30% over the first month of the trial in the drug conservation arm but were unchanged in the viral suppression arm, and also that the risk of death was associated with higher IL-6 levels at study entry [36]. In another study, plasma IL-6 levels did not seem to change in most HAART recipients following initiation of treatment, except in those who developed immune restoration and inflammatory syndrome (IRIS), who showed an increase in plasma IL-6 post-HAART [54]. IL-6 is a B-cell-stimulatory cytokine that is involved in promoting IgH class switching and also has antiapoptotic effects on B cells, effects that have the potential to promote the development of B-cell cancers. In addition, a persistent excess of IL-6 may promote nascent lymphoma clones through its anti-apoptotic effects. Therefore, even if levels of other cytokines are normalized in HAART recipients, there still may be increased risks for those forms of AIDS-related NHL that are associated with errors in IgH class switch DNA recombination, which may be promoted by IL-6. If this is true, one would expect to observe excess levels of those forms of AIDS-NHL associated with this genetic error among HAART-treated persons. In light of this, it is informative to note that whereas the incidence of EBV-associated AIDS-related NHLs has decreased in the first decade of the HAART era, the incidence of Burkitt’s lymphoma has not decreased significantly [32].

Because this study did not include concurrent HAART-naive HIV-infected controls, we cannot adequately account for temporal changes that may have confounded this analysis (i.e. changes in patient management strategies over time, and better management of comorbidities including other opportunistic illnesses). However, our results were generally consistent within each time period studied, and significant changes in biomarker levels were seen within the fairly brief time periods closest to HAART initiation, often as short as a 1-year interval. Results from historical controls selected from the MACS cohort suggest that these biomarker levels are relatively stable over a period of three to five years (data not shown).

Most of the HIV-infected men in our study sample are long-term survivors. In this group of 467 participants, 81% were HIV-infected by early 1988, and may have been infected as early as 1978. Hence, this is a population that survived between 8 (seroincident participants) and 14 years (seroprevalent participants), respectively, prior to the introduction of HAART. These survivors are healthier and slower progressors than other MACS participants, who progressed to AIDS and died prior to HAART availability, and therefore could not have participated in this study. Obviously, the results obtained in this study cannot provide information on HAART’s impact upon serum B-cell activation associated molecules in these faster progressors.

Another potential source of error can occur with spurious events that cause transient increases in B-cell activation, such as responses to acute infection. Since effective HAART will potentially decrease the likelihood of acute infections, it is probable that acute infections will more likely occur prior to exposure to HAART.

Ultimately we would like to have been able to evaluate most closely those persons who developed AIDS-related NHL during the course of their infection and HAART treatment. However, with only 12 men this group developing AIDS-NHL, it was not possible to provide a formal analysis of this subgroup. Descriptive analyses suggest the NHL subgroup had a similar HAART response to that seen in the larger group of HIV-infected persons, with substantial decreases, but not normalization, of biomarker levels following HAART. Further studies in a larger NHL group are needed to support these findings.