Age-dependent differences and similarities in the plasma proteomic signature of postoperative delirium – Scientific Reports

This study compared part of the plasma proteome of postoperative delirium in two age groups and reveals both similarities and differences in the delirium signature with age. Consistent with previous results, the plasma proteome of delirium in older patients was complex and included numerous proteins and pathways, mostly related to inflammation and immune regulation. In contrast, the proteomic signature of delirium in middle-aged subjects was less intricate and included only a relatively small number of the proteins and pathways enriched in the older patients, though both fit the delirium phenotype. In fact, only seven plasma proteins and twelve protein networks were common to delirium in both ages. Notable among these are IL-6 and IL-8, which have been linked to delirium in older surgical patients and been proposed as potential biomarkers for it in several prior studies^12,14,27,28. Our results indicate those associations are accurate irrespective of age and suggest that these and the other age-independent delirium-associated plasma proteins and pathways reflect core pathophysiological features of the syndrome and could be universal, age-independent plasma biomarkers for it. On the other hand, the proteins and pathways unique to delirium of older age may reflect, and potentially contribute to, the greater vulnerability and/or morbidity of the syndrome in older persons. Either way, our results reveal age differences in the plasma proteome of this common postoperative syndrome.

Overall, 89 (33%) of the 273 proteins on the panel were associated with postoperative delirium at some point in the perioperative period in one or both age groups and this signature was dynamic. Postoperative plasma and pre- to postoperative concentration changes were most revealing, as only a minority of differences were evident in the preoperative plasma of delirious subjects, and then only in the younger cohort. This suggests the search for predictive biomarkers of delirium may be more fruitful in middle-aged than older surgical patients. Of the 89 delirium-associated proteins, some have been linked to the syndrome previously. Examples include IL-6 and IL-8^{9,12,14,26,27,29}, albeit only in older subjects, and CHI3L1, which has recently been identified as a potential predictive and disease biomarker for delirium¹³ and was positive in our analysis of delirium cases irrespective of age. However, we also uncovered associations between delirium and plasma proteins not reported previously. Examples include IL-15, a proinflammatory cytokine implicated in the pathophysiology of Alzheimer’s Disease and frontotemporal dementia³⁰; CCL23, a chemokine involved in brain injury-induced neuroinflammation and progressive cognitive impairment³¹; TNF receptors 1 and 2 (TNFR 1 and 2)³²; and CSF-1, a molecule critical for activation of cerebral microglia and strongly implicated in neurodegenerative disease³³. GSEA and PPI network analyses similarly demonstrate delirium-associated changes in a multitude of pathways, mainly involving cytokine, chemokine, and interleukin signaling but also a few unexpected networks. One example of the latter is metabolism of lipids, concentrations of which are altered in several neurologic diseases and in the cerebrospinal fluid of older patients postoperatively^34,35. Expression of many of these proteins was decreased in the context of delirium, as observed previously^12,13, suggesting that dysregulation or imbalance of the systemic inflammatory milieu, not just activation, is a pathophysiologic characteristic of the syndrome. Collectively, our findings are consistent with the concept that systemic inflammation is a fundamental driver of postoperative delirium, even in middle-aged persons, and add several new candidates to the growing list of proteins and pathways that may be involved in delirium susceptibility and pathogenesis. However, our results are associative in nature, so it is still unclear whether these molecules are causes or merely markers of delirium. Either way, age proves to be an important factor.

Delirious middle-aged and older surgical patients had remarkably few delirium-related proteins and networks in common. Among proteins, they had just three postoperatively (IL-8, LTBR, and TNF-R2) and four by pre- to postoperative change (IL-8, IL-6, LIF, and ASGR1) in common. IL-6 and IL-8, as mentioned, have been strongly associated with delirium in older subjects previously^12,14,27,28. Our results demonstrate that association holds in middle-aged patients as well, thus strengthening the case that these proteins could be useful, age-independent biomarkers for delirium. None of the other proteins have been linked to delirium before, although TNFR1 (but not TNFR2) is implicated in the delirium of critical illness³⁶. All are inflammation related. LTBR belongs to the TNF receptor superfamily and is enriched in myeloid cells and monocytes, which are key mediators of clinical outcomes of surgery^17,37. There is little information about the regional or cellular specificity of LIF and ASGR1 in the human brain and both are poorly expressed there. LIF is a cytokine in the IL-6 family and a stem cell growth factor that is not detected in immune cells but reportedly has anti-inflammatory and neuroprotective properties^38,39 and ASGR1, a transmembrane protein, is found mainly in myeloid cells and liver but has no known function in brain. Pathway analyses likewise revealed relatively little overlap between the ages. At the postoperative time point, one gene set was common to both age groups and by pre- to postoperative change there were nine. Confirmation is required but the association of these proteins and networks with delirium in both age groups implies they represent a core feature, and possibly an essential part, of the pathophysiology of the syndrome. As such, proteins common to the condition across age could be universal, age-independent biomarkers for the syndrome.

The differences in the delirium signature with age are also interesting. Overall, the profile is more complex in older subjects. By pre- to postoperative change, for instance, there were 12 delirium-associated proteins and 11 delirium-associated networks in middle-aged patients but 51 proteins and 16 networks in the older group. However, 9 of the networks were common to both age groups. Thus, it appears that the additional proteins and networks associated with delirium in older patients may not be essential for it since the same clinical phenotype occurs in younger surgical patients without them. One possibility is that the proteins and networks unique to the delirium of older age represent parallel processes or consequences of delirium, rather than a direct cause of it, and contribute in some still undefined way to the greater vulnerability and morbidity of delirium in older patients. Along these lines, older patients with delirium had worse preoperative cognition by MiniCog and verbal fluency than age-matched patients without delirium. Poor cognition is a well-established predisposing factor for delirium but, in and of itself, is unlikely to account for the biomarker changes we identified because about half of the delirious older patients with biomarker changes had normal cognition (64% by MiniCog, 43% by verbal fluency). Nonetheless, more work is necessary to understand how predisposing and/or precipitating factors for delirium influence its plasma biomarker profile.

This study has both strengths and limitations. Chief among the former is that we included both middle-aged and older patients undergoing similar surgical procedures and having the same clinical syndrome. Additionally, we utilized a high sensitivity assay and a large protein panel and analyzed results at both the level of individual proteins and functional networks. Limitations include that this was a single-center study of subjects with relatively high educational attainment and low racial diversity and that plasma does not necessarily reflect the status of the brain. As such, it is unclear whether these plasma proteins have an etiologic role in the cognitive dysfunction of delirium or are simply peripheral markers of the responsible CNS processes. Furthermore, our study focused on neurosurgical and orthopedic spine surgery, which improves sample homogeneity but limits generalizability. The higher delirium rate in the included versus excluded subjects raises the possibility of selection bias but this seems unlikely because exclusion was based solely on plasma sample availability, enrollment criteria and experimental procedures were otherwise identical, and there was no difference between included and excluded subjects in other demographic, health, and outcome measures. Another limitation is that our sample size for delirium was relatively small (N = 22), though it is similar to or larger than other recent studies of the delirium proteome^12,13,40. We used a validated and widely-accepted combined method (3D-CAM and chart review) to identify delirium^13,41,42,43 but no method is perfect and we could have missed some cases. Work-flow for the proteomic analysis required a 2nd freeze–thaw cycle after plate preparation; while this could introduce pre-analytic variation, all samples were processed identically so it is unlikely to affect within or between group comparisons. Due to the exploratory nature of the work, we did not correct for multiple comparisons in the analyses of individual proteins, gene sets, or protein–protein network interactions and, therefore, cannot exclude the possibility of a type 1 error. Others investigating the plasma proteome of delirium have taken the same approach and our results confirm some previous findings but validation in larger data sets is necessary. Another limitation is that we utilized a curated panel of proteins. Although extensive, our platform included fewer proteins and used an analytical method different from those used in other recent work. Consequently, we did not assay some proteins implicated previously in the development of postoperative delirium (e.g. CRP) and possibly missed others that might be affected because they underwent post-translational modifications (e.g. phosphorylation) that produced proteoforms our assay could not recognize or were not on the panel. There are also differences between our results and previous reports for a few proteins. For example, delirium has been linked to upregulation of CHI3L1¹³, a protein associated with aging and chronic immune disorders, and to an increase in NFL, a potential plasma protein marker of neuronal injury^28,43. We found an increase in CHI3L1 by differential expression when the age groups were combined (Supplemental Fig. 2D) but it did not emerge when the age groups were analyzed separately, possibly due to the smaller sample sizes in each age group. NFL was not a positive biomarker for delirium in any of our analyses. We are not unique in this regard^44,45 and nor do all report a relationship between plasma NFL and postoperative cognitive outcomes⁴⁶ but the different results could be explained by variations in the type and invasiveness of surgical procedures (cardiac vs. joint replacement vs. spine surgery), assay method (SOMAscan vs. proximity extension assay), or duration of postoperative follow-up. The latter is relevant because we collected our final plasma sample on postoperative day 1 and changes could have been missed because NFL peaks later.

In summary, we demonstrate that the plasma protein signature of postoperative delirium is age dependent. Some proteins and pathways, including plasma IL-6 and IL-8, are common to delirium in both age groups but older delirious patients had changes in more proteins and pathways than middle-aged subjects with the same clinical phenotype. This implies that there are age-related differences in the pathogenesis of delirium and/or its clinical consequences but mechanistic details are not clear. We speculate that the proteins and networks common to delirium in both ages reflect core pathogenic attributes of the condition, and could be age-independent biomarkers for it, whereas those unique to older delirious subjects may represent the greater susceptibility and/or morbidity of the syndrome in that age group. Additional studies of delirium in patients of different age are needed to validate these results and to distinguish the plasma signature of delirium per se from secondary but important and age-related aspects of this complex syndrome.