IFN-α administration was associated with significant alterations in diurnal HPA axis activity including flattening of the ACTH and cortisol slope and increases in evening ACTH and cortisol concentrations. Both flattening of the cortisol slope and increases in the evening cortisol were in turn significantly correlated with IFN-α-induced increases in depression and fatigue. IFN-α-induced increases in plasma concentrations of IFN-α as well as TNF-α and sTNFR2 were also significantly correlated with behavioral changes, however, no relationship was found between these or other immune variables and changes in HPA axis parameters. Given the prospective, longitudinal design of the study, these data provide some of the first evidence that chronic exposure to innate immune cytokines such as IFN-α may be a relevant pathophysiologic pathway by which HPA axis dysregulation (especially flattening of the cortisol slope) occurs in patients with various medical illnesses. Nevertheless, given the lack of a direct relationship between HPA axis and immune measures, the mechanism by which chronic cytokine exposure influences HPA axis function remains to be determined.
A number of studies suggest that dysregulation of HPA axis diurnal activity represents an important link between illness and behavioral disturbance. Flattening of the cortisol rhythm has been observed in patients with—or at risk for—a number of medical disorders, including cancer, type 2 diabetes and cardiovascular disease.23–26
In addition, flattening of the cortisol slope has been associated with depression, fatigue, anxiety and maladaptive coping styles in both medically ill and medically healthy individuals. 30–32,42–44
The data reported here replicate the association between flattening of the cortisol slope and behavioral alterations including depression and fatigue. Moreover, the changes in cortisol slope as a function of IFN-α treatment are very much in line with previous reports in other clinical populations that have exhibited alterations in diurnal cortisol rhythm. These similarities were apparent despite the fact that in the majority of previous studies, cortisol slope was derived from diurnal sampling of cortisol in saliva over several (typically 3) days in the subject’s home environment. For example, patients with meta-static breast cancer in the study by Abercrombie et al
exhibited a slope of −0.092 log ug per 100ml per hour (s.d. 0.033) compared to −0.113 (s.d. 0.030) in healthy control subjects. These results are similar to the cortisol slopes in IFN-α/ribavirin treated patients (−0.07 (s.d. 0.03)) versus controls (−0.11 (s.d. 0.06)) at visit 2 in the current study. Likewise, in the study by Giese-Davis et al
slopes for metastatic breast cancer patients labeled as ‘high anxious’ or ‘repressor’ (slope = −0.07 for both groups (s.d. 0.08 and 0.04, respectively)) were exactly the same as IFN-α-treated patients in the current study, whereas patients labeled as self-αssured or non-extreme exhibited slopes of −0.10 (s.d. 0.05) and 0.11 (s.d. 0.05), respectively; virtually the same as our control subjects. In the Matthews CARDIA study, individuals with cardiac calcifications exhibited a slope of −0.061 (s.d. 0.064) and those without calcifications exhibited a slope of −0.084 (s.d. 0.054), a smaller effect size than the one observed between IFN-α-treated and control groups.23
Finally, in the study by Bower et al
cortisol slope was −0.21 (s.d. 0.13) in controls versus −0.14 (s.d. 0.06) in breast cancer survivors with fatigue.32
Although the slope values are to an extent ‘steeper’ in the study by Bower and co-workers than those observed in the current study (and the studies by Abercrombie et al
Giese-Davis et al
and Matthews et al
), the 67% difference in slopes between fatigued patients versus controls is similar to the 64% difference observed between IFN-α-treated and control patients at visit 2. Based on these data, it is reasonable to conclude that IFN-α had a significant effect on the HPA axis and cortisol slope that is very much in line with the magnitude of effects that have been observed in other clinical populations in previous studies.
It is of great interest that the association between flattening of the cortisol slope and the development of depression and fatigue as a function of cytokine exposure appeared to be primarily related to increased cortisol production late in the day. Less attention has been paid to cortisol activity near the circadian nadir than to either the overall diurnal slope or the morning peak of cortisol release. However, increased pulsatile release of cortisol late in the day has emerged as a primary HPA axis abnormality in major depression.30
In addition, cortisol elevation late in the day correlates with age-related disruption of sleep integrity and loss of slow-wave sleep and has been shown to have more adverse metabolic effects than elevation of cortisol in the morning.45,46
For example, in subjects given metyrapone to suppress endogenous cortisol production, administration of hydrocortisone in the afternoon produced significantly greater impairments in glucose metabolism/insulin sensitivity when compared to an identical dose of hydrocortisone administered in the morning.46
Thus, increased cortisol activity late in the day may be an important mechanism by which activation of innate immune responses contribute to both behavioral alterations and vulnerability to medical illnesses.
Correlations between IFN-α-induced HPA axis changes and behavior were found using scales of both depression and fatigue. Nevertheless, the MADRS (depression) and MFI (fatigue) were highly inter-correlated. However, even when the fatigue item (lassitude) was removed from the MADRS, correlations between HPA axis changes and depression largely persisted. These data indicate that the MADRS and MFI are measuring overlapping constructs, which are both overexpressed as a function of IFN-α exposure and cannot easily be disentangled.
It is important to note that HPA axis changes as a function of chronic IFN-α exposure are in contrast to the changes seen following acute IFN-α administration, where plasma ACTH and cortisol levels as well as IL-6, were dramatically elevated for up to 3 hours.10–13
These data suggest that significant adaptation of the HPA axis occurs following chronic IFN-α treatment and emphasize the distinction between acute and chronic cytokine effects on HPA axis function. Moreover, given the similarity to HPA axis changes following acute versus chronic stress (where acute stress is typically associated with marked HPA axis activation and chronic stress is associated with flattening of the cortisol slope), the data suggest that chronic IFN-α administration acts on the HPA axis much like a chronic stressor. Moreover, the data highlight the relevance of chronic cytokine (IFN-α) exposure as a model for individuals exposed to chronic immune activation as a function of either medical illness and/or chronic stress.
The fact that many of the diseases associated with flattening of the diurnal cortisol rhythm are characterized by activation of the body’s innate immune/inflammatory response,47–53
further supports the notion that chronic exposure to innate immune cytokines can disrupt normal HPA axis function. Specific mechanisms by which cytokines, such as IFN-α, might disrupt normal HPA axis rhythms are currently unknown; however, several possibilities warrant consideration including effects of cytokines on factors that regulate HPA axis function as well as cytokine effects on sleep-wake cycles.
Innate immune cytokines have been shown to activate HPA axis pathways, in part, through their induction of corticotropin releasing hormone.54,55
Moreover, acting through several intracellular signaling pathways, cytokines of the innate immune response have been found to inhibit glucocorticoid receptor function,56,57
which, in turn can disrupt negative feedback regulation of HPA axis function, potentially leading to increased concentrations of ACTH and cortisol late in the day. These possibilities are consistent with recent studies in which IL-6 was found to correlate with a dampened cortisol rhythm in patients with metastatic colon cancer and increased p.m. cortisol in patients with coronary artery disease.28,29
Nevertheless, the current study found no significant correlations between immune and HPA axis variables. Although the sample size for this study was relatively small (in part related to the intensive sampling demands and strict entry criteria), correlation coefficients between immune variables and measures of diurnal cortisol secretion, all of which were non-significant, ranged from r
= 0.0 to r
= 0.22, thereby accounting for less than 5% of the variance. These correlation coefficients are in contrast to the much stronger (and statistically significant) relationships that were observed between measures of diurnal cortisol secretion and behavior as well as the immune parameters and behavior, where correlation coefficients ranged from r
= 0.36 to r
= 0.64, accounting for 13–41% of the variance in measures of depression and fatigue. Taken together, these data suggest that either innate immune cytokines and HPA axis alterations are independent contributors to changes in behavior, or there are additional, yet to be identified, pathways or factors which serve to link HPA axis and immune measures. One possibility in this regard includes additional cytokines that were not measured as part of the study design. For example, in a study by Wichers et al
IL-8 (which was not assessed in the current study) was found to significantly correlate with daily average cortisol over time in a longitudinal study of IFN-α-treated patients with hepatitis C. Interestingly, Wichers et al
also found positive correlations between the awakening cortisol response and IL-6 as well as IL-8, IL-10 and sIL-2R. These data suggest that while correlations between HPA axis and immune parameters may exist, they may be, in part, a function of the HPA axis or immune parameter assessed.
Another potential pathway by which innate immune cytokines such as IFN-α might influence diurnal HPA axis activity is by affecting sleep. Complex bi-directional relationships exist between inflammation and the sleep-wake cycle.58,59
For example, innate immune cytokines including IL-6 and IFN-α have been found to disrupt sleep efficiency and architecture in humans.7,60
Sleep disruption, in turn, has been shown to promote further cytokine production and release.7,58,61,62
Sleep disruption has also been shown to disrupt the diurnal cortisol rhythm,63
and as noted above, is especially likely to increase cortisol levels late in the day63
—a pattern of changes identical to those observed in the current study following chronic IFN-α exposure. Interestingly, in a recent case report of two patients with subacute sclerosing panencephalitis, intracerebroventricular administration of IFN-α was found to markedly inhibit the production of orexin, a neuropeptide involved in the maintenance of sleep-wake cycles.64
Thus, the effects of innate immune cytokines (for example, IFN-α) on neuropeptides and/or neurotransmitters that regulate sleep may in turn influence diurnal cortisol secretion. Finally, IFN-α has been shown to influence daily rhythms of locomotor activity and body temperature in association with effects on the expression of relevant clock-genes in the suprachiasmatic nucleus, a hypothalamic brain region intimately involved in the regulation of circadian rhythms.65
Such effects of cytokines on daily rhythms, sleep and other relevant behaviors that have been associated with altered diurnal cortisol secretion (for example, depression)30
may explain, in part, the lack of a direct correlation between cytokines and HPA axis function.
Regarding the relationship between cytokines and behavior, to our knowledge, this is the first study to report a significant correlation between plasma concentrations of IFN-α and depression and fatigue. These data indicate that there is a dose–response relationship between IFN-α and behavior, and thus IFN-α dosage reduction during IFN-α therapy for hepatitis C (or other conditions) appears to be an appropriate strategy for the initial management of IFN-α-induced behavioral change. A similar relationship was found between behavioral changes and plasma concentrations of both TNF-α and sTNF-R2. Of note, mean plasma concentrations of TNF-α and sTNFR2 in HCV control patients in the current study were similar to those reported previously for healthy control subjects, indicating that HCV infection itself was not associated with elevated plasma concentrations of these immune parameters in this study population.66–68
Furthermore, the magnitude of elevation in sTNFR2 from visit 1 to visit 2 in IFN-α-treated patients (0.8 ng ml−1
) corresponds closely with differences that have been observed between control subjects and either patients with type 2 diabetes or cancer survivors with fatigue (~0.7 ng ml−1
), supporting the clinical relevance of the observed IFN-α-induced increases in sTNFr2.66,67
Soluble TNFRs remain elevated for long periods of time after TNF-α administration and are believed to reflect previous TNF-α effects.69,70
Thus, elevations in TNFR2 in the current study, along with significant correlations between TNF-α and depression and fatigue, provide evidence that activation of TNF-α and its signaling pathways may be an important component of the effects of IFN-α on behavior.
Several strengths and limitations of the current study should be noted. The use of a prospective, longitudinal design allowed for a ‘within-subjects’ design that reduced the potential effects of baseline variables such as age and body mass index that have been cross-sectionally associated with diurnal rhythm of the HPA axis in prior studies.71,72
Environmental factors such as sleep and wake times and the timing of meals can also influence the pattern of diurnal cortisol activity and are typically not controlled in studies of cortisol slope conducted in outpatients.32
The use of an inpatient environment in the current study may have strengthened findings by controlling these variables.
In terms of study limitations, several issues warrant comment. First, the sample size was relatively small, and therefore the inability to detect significant correlations between HPA axis parameters and proinflammatory cytokines (and their receptors) may have been secondary to a lack of power. Nevertheless, as noted above, the correlations that were obtained between HPA axis and immune parameters were relatively small, and even if found statistically significant with a larger sample size would only account for a relatively small amount of the variance in diurnal HPA axis activity compared to what was found with the behavioral measures. Another limitation is that relatively few cytokine parameters were assessed, and as noted above there could be other cytokines that may correlate with HPA axis function. Similarly, immune parameters were only measured at one time point during IFN-α treatment, and it is possible that cytokines, such as IL-6, which were not elevated following 12 weeks of treatment may have been increased at an earlier time point and may have contributed to the initial development of neuroendocrine and/or behavioral changes, with these changes being subsequently maintained by other cytokine, neurotransmitter or neuroendocrine systems. Another limitation of the study is that group assignment was not randomized. The lack of randomization resulted in the occurrence of five patients with a history of past major depression in the IFN-α/ribavirin treatment grou and no patients with a major depression history in the control group. Nevertheless, further evaluation of IFN-α-treated patients with and without a past history of major depression revealed no significant differences in terms of the effects of IFN-α/ribavirin on any of the HPA axis, immune or behavioral variables.
It should be noted that all subjects included in this study were infected with HCV, and analyses of behavioral and neurobiological parameters were not controlled for viral load. Therefore, it remains possible that different levels of HCV infection may have variably augmented the effect of chronic IFN-α exposure on diurnal HPA axis activity and immunologic parameters. Nevertheless, based on the similarity of cortisol slope values in HCV control subjects in the current study versus cortisol slope values in healthy controls from the published literature (see above),26
there is no clear evidence that chronic viral infection itself was associated with alterations in diurnal HPA axis activity in the absence of IFN-α administration. As per standard of care, all subjects who received IFN-α were also treated with the antiviral agent, ribavirin. Although there is no evidence that ribavirin affects the HPA axis, the possibility that ribavirin influenced the results cannot be ruled out. Nevertheless, individuals treated with IFN-α monotherapy for cancer show similar behavioral changes to those reported here.73
A final limitation of the study is that HPA axis hormones were only assessed hourly for 12h during the day, and no information is available on the 24-h circadian rhythm of these hormones. The 12-h protocol was used to limit the potential influence of overnight blood sampling on sleep (which in turn may influence cortisol rhythm) and to coincide with previous studies that have used daytime salivary sampling strategies to examine the relationship between diurnal cortisol rhythm and both medical and psychiatric/psychological outcomes. Nevertheless, more frequent blood sampling over a 24h period as conducted by Licinio et al.74
might have revealed more nuanced changes in hormone secretion including changes in hormonal spike frequencies and amplitude as well as phase shifting in the circadian rhythm.
In summary, results of the current study indicate that administration of the innate immune cytokine IFN-α leads to profound behavioral changes that are associated with flattening of the diurnal cortisol slope, with the primary change being an increase in evening cortisol plasma concentrations. Potential mechanisms for these effects may include cytokine-induced disruption of the sleep-wake cycle as well as inhibitory effects of cytokines on glucocorticoid receptor functioning. Further studies clarifying these potential mechanisms may provide novel insights into the pathogenesis of behavioral disturbances in the context of sickness as well as provide clues to novel treatment targets for addressing the consequences of cytokines on the brain and behavior.