This high-throughput proteomics approach identified a three-protein signature, namely CDCP1, SIRT2 and LEPR, significantly associated with baseline CES-D score and persistently elevated or recurrent depressive symptoms in individuals with recent-onset diabetes. The positive association of CDCP1 and SIRT2 and the inverse association of LEPR with baseline and persistent/recurrent depressive symptoms were independent of confounding variables. In addition, our study revealed seven additional biomarkers mostly negatively associated with baseline CES-D score and 10 additional biomarkers all positively associated with persistent/recurrent depressive symptoms. However, statistical significance was lost after adjustment for multiple testing.

SIRT2 is an enzyme that is highly expressed in cerebral oligodendrocytes and also found in the circulation.38 In the brain, SIRT2 plays a role in apoptosis, autophagy and neuroinflammation.38 It is hypothesized that SIRT2 has a pro-inflammatory effect, opposite to the anti-inflammatory effect of SIRT1.39 SIRT2 inhibition has been found to induce an antidepressant-like action40 and has been suggested for the treatment of neurodegenerative diseases.41 However, studies have also reported beneficial effects of SIRT2 on insulin resistance, fatty acid oxidation and inflammation in the context of type 2 diabetes.42 43 These and other data suggest that SIRT2 action may be context-specific, and the potential of SIRT2 as pharmacological target appears difficult to assess in the absence of further knowlegde of the mechanistic links between this protein and both diabetes and depression.44

CDCP1 is part of the Inflammation panel, and higher levels were found to be associated with higher baseline and persistent/recurrent depressive symptoms. CDCP1 is a transmembrane protein involved in various processes, such as immune cell migration and chemotaxis.45 As ligand for CD6 on T cells, it specifically contributes to T-cell-mediated inflammatory processes.46 Currently, it is unknown if CDCP1 has a role in diabetes (type 1 and type 2) or depression. However, there are multiple studies that identified associations between CDCP1 and various other neurological conditions. One study found higher levels of CDCP1 to be associated with a higher risk of incident all-cause and Alzheimer’s dementia.47 Further studies reported associations between higher circulating CDCP1 levels and symptoms of psychotic disorder,48 lower cognitive function49 and smaller total cerebral brain volume.50 Of note, CDCP1 is also upregulated in cancer cells from multiple different organs.51 Taken together, these studies suggest a role of CDCP1 in neurological processes and pathways and in oncogenesis, which might be attributed to shared pro-inflammatory mechanisms.

Leptin is an adipokine produced and secreted from white adipose tissue.52 It affects mood and cognition positively by causing structural and functional alterations in the brain.52 Leptin binds to the leptin receptor in brain regions including the hypothalamus and hippocampus.53 The inverse association between LEPR and baseline and persistent/recurrent depressive symptoms is in line with evidence from mouse models where a selective deletion of LEPR in the hippocampus was found to induce depression-like behavior. This provides evidence that the signaling of leptin in the hippocampus may be essential for maintaining positive mood states. Furthermore, this indicates that LEPR is involved in the molecular mechanism of leptin’s antidepressant effect.54 Signaling through LEPR is also essential in the crosstalk between adipose tissue and the central nervous system to regulate appetite and satiety, but in people with obesity and/or type 2 diabetes the sensitivity to leptin appears to be decreased.55 Although leptin administration has been reported to lower excess weight and improve insulin resistance, these effects are only observed in people with leptin deficiency.56 Leptin signaling plays multiple key physiological roles, such as in the regulation of neurological function and metabolism, but remains a challenging target for therapeutic intervention.

Our study revealed several other biomarkers to be associated with either baseline or persistent/recurrent depressive symptoms. For instance, MMP1 and SCF were previously found to be involved in various mechanisms related to the development and modulation of inflammatory processes and neurogenesis.57 58 The chemokines CCL11, CCL19 and CCL20 were positively associated with persistent/recurrent depressive symptoms. Chemokines and their receptors are widely expressed in the CNS and are involved in many roles, including immune surveillance, immune cell chemotaxis and blood-brain barrier permeability.59 Despite differences in samples and methodologies, several human and animal studies have associated higher levels of chemokines with depressive behavioral symptoms.59 Indeed, elevated levels of chemokines were found in individuals with major depressive disorder and other mental disorders, including bipolar disorder and schizophrenia.59 Our study did not reveal a significant association between NfL and depression in contrast to other studies. We speculate that this is because neurodegenerative processes might be less pronounced in individuals with recent-onset diabetes with good glycemic control, or these processes are not a major hallmark of depression pathogenesis in patients with diabetes.

Most of the prior studies investigating associations of protein biomarkers with depression in individuals with diabetes have primarily focused on the inflammatory biomarkers IL-6, IL-1RA and CRP, among which only IL-6 was included in our panels. The lack of an association between IL-6 and depressive symptoms in the present study aligns with our and previous reports among individuals with diabetes.15 16 19 34 These findings were consistent despite differing biomarker measurement techniques (ELISA and PEA technology) and diabetes duration.

Previous research has shown the potential for proteins to be used to help develop new treatments for depression.60 Implementing proteomic-based techniques is crucial to examine the intricate intracellular networks concerning immune-inflammatory pathways.60 By using omics-based technology to examine biomarkers, we can better assess these networks and pathways and therefore, research can be feasibly translated to patients with depression with or without comorbid disorders to help with drug discovery.60 Furthermore, specifically for SIRT2, various SIRT2 functions have been identified in cells that show promise for it to be a potential target for drug development with therapeutic effects on several human diseases, including depression.61 SIRT2 inhibitors have been found to induce an antidepressant-like action, and agents such as 33i, sirtinol and AGK2 are novel SIRT2 inhibitors that pose potential therapeutic options to treat depression.61 However, the context-specific action of SIRT2 and potential adverse effects on other outcomes need to be considered.44

The strength of our study is the repetitive measurement of depressive symptoms at up to six time points, allowing a reliable definition of persistent/recurrent depressive symptoms. Another study has previously assessed persistent/recurrent depressive symptoms in people with diabetes, but not in the context of biomarkers,36 and defined persistent/recurrent depressive symptoms based on only three time points, a limited window to reliably define an individual as having persistent or recurrent depressive symptoms. Other strengths include investigating a diverse panel of protein biomarkers covering various inflammatory and neurological pathogenic processes measured with high-throughput proteomic technology. Most comparable studies concentrated on a single or few biomarkers, potentially overlooking the intricate complexity of biological mechanisms. The extensive phenotyping of our study cohort allowed for a comprehensive adjustment for covariates.

Our study has some limitations. The CES-D questionnaire evaluates recent self-reported depressive symptoms (current or last week’s episode), which is not clinically validated. Nevertheless, the CES-D score represents a reliable tool used in the clinical setting among individuals with diabetes.33 Different cut-offs for the CES-D score have been proposed to define depressive disorder,31 however a lower cut-off was chosen in the present study since the outcome was focused on examining the presence of depressive symptoms. The analysis of persistent and recurrent depressive symptoms has a risk of misclassification for individuals with data on less than six time points. Our study includes individuals recently diagnosed with diabetes with relatively good glycemic control, which may limit the generalizability of the findings to individuals with long-standing diabetes or those with poor glycemic control. We did not perform a power calculation because we did not test a specific hypothesis as in a clinical trial but rather used all available cohort data for our study aims.62 Therefore, the findings should be interpreted within this context. The potential limitations in sample size are reflected by the wide 95% CIs, which have been consistently reported throughout the manuscript, and by the fact that our results are not statistically significant after Bonferroni correction. Therefore, additional cohort studies and meta-analyses using these and our data will be desirable for more certain results. Previous studies have indicated associations between other biomarkers of inflammation and oxidative stress such as IL-1RA or isoprostane with depression, but these data were not available in the GDS. Lastly, given the known role of genetic susceptibility in the development of depression, our results based on individuals of primarily European descent may not be generalizable to non-European populations.