Potential Transcriptional Biomarkers to Guide Glucocorticoid Replacement in Autoimmune Addison's Disease

Åse Bjorvatn Sævik; Anette B. Wolff; Sigridur Björnsdottir; Katerina Simunkova; Martha Schei Hynne; David William Peter Dolan; Eirik Bratland; Per M Knappskog; Paal Methlie; Siri Carlsen; Magnus Isaksson; Sophie Bensing; Olle Kämpe; Eystein S Husebye; Kristian Løvås; Marianne Øksnes


J Endo Soc. 2021;5(3) 

In This Article


Biomarkers reflecting GC replacement in AAD are lacking. In this exploratory study, we mapped the transcriptional landscape of GC exposure and found that DSIPI, DDIT4, and FKBP5—independently and as a triad—are candidate transcriptional biomarkers reflecting levels of GC because (i) all 3 genes were clearly and consistently upregulated after both high-dose HC infusion and in normal cortisol compared to very low cortisol, (ii) gene expression levels significantly correlated with normal cortisol levels, and (iii) they strongly correlated with each other, implying a high degree of co-regulation.

DSIPI, a key mediator of GC anti-inflammatory effects,[25] appears as the most precise biomarker since its expression significantly correlated with serum cortisol both after high-dose HC infusion (step 1) and with normal cortisol levels (step 2). DDIT4, an important inhibitor of the mTOR pathway in response to stress,[26] and FKBP5, an important short-loop feedback inhibitor of GC action,[16] may be more sensitive markers of GC exposure as they revealed the largest fold changes between very low and normal cortisol levels (step 2). Finally, DDIT4 and FKBP5 were the only genes where expression levels significantly correlated with plasma ACTH at very low cortisol.

In addition to DSIPI and FKBP5, we found that that MMP9 had a rapid and sustained increase in gene expression following infusion of 100 mg HC (step 1). This is in contrast to a study by Aljada and coworkers where plasma MMP9 protein levels significantly decreased after 100 mg HC infusion,[27] but in line with another study on healthy individuals where MMP9 levels increased in response to 300 mg HC infusion.[28] Obviously, gene expression and plasma protein levels are not directly comparable, but we would expect the direction of the response to HC infusion to correspond. Taken together, this indicates that more research is needed to determine the true effect of high-dose HC exposure on MMP9 gene expression and MMP9 protein levels.

In step 2, 5 of the 93 GC-regulated genes were significantly upregulated (DDIT4, DSIPI, FKBP5, CEBPB, VDR) whereas 3 were downregulated (ADARB1, POU2F1, ARID5B) in normal compared with very low cortisol. Increased CEBPB expression has been suggested as an early marker of efficacious response to GC treatment in inflammatory bowel disease (IBD) and could be used to identify the 20% of IBD patients who are refractory to GC treatment.[29] With this in mind, we suggest that future studies explore whether difference in CEBPB expression could help identify patients with AAD in need of higher GC replacement dosages due to partial GC resistance.

Expression of VDR has previously been shown to increase following GC exposure in a time- and dose-dependent manner, but only in the presence of GC receptor, indicating that GCs directly regulate VDR expression through GC receptor activation.[30]

The largest effect size on gene expression change was noted for ADARB1, a gene considered key for circadian rhythmicity.[31] Previous studies disagree on whether GCs increase or decrease ADARB1 expression. In an in vitro study on subcutaneous preadipocytes, expression of ADARB1 was twice as high in cells treated with cortisol compared with untreated cell controls.[32] In contrast, our data adds evidence to the notion that GCs suppress ADARB1 expression, as previously demonstrated in human keratinocytes treated with dexamethasone.[33]

For POU2F1, a transcription factor important in stem cell regulation, our findings support that GCs suppress its expression in line with previous studies.[34] The clinical value of POU2F1 has primarily been demonstrated in the setting of cancer, as measurement of POU2F1 protein levels is better than clinical cancer staging in predicting prognosis.[34]

Finally, ARID5B has been associated with susceptibility to and treatment outcomes of acute lymphatic leukemia (ALL), and lower ARID5B expression at diagnosis has been linked to increased risk of ALL relapse.[35] To our knowledge, this is the first study to demonstrate the inhibitory effect of GCs on ARID5B expression.

One patient in step 1 was not cortisol deplete before HC infusion. We do not, however, doubt that the diagnosis of AAD was correct as all study patients were recruited from our national quality registry (the National Registry for Addison's disease [ROAS]) and were carefully characterized both clinically and biochemically. Furthermore, the patient had clearly elevated ACTH (198 pmol/L) before HC infusion that was successfully suppressed after HC infusion (15.1 pmol/L). We propose that this patient had residual GC production, which we recently demonstrated is present in one-third of all patients with AAD.[36] To ensure full privacy protection, all samples were anonymized before analysis, and we were unable to identify the study patient in order to verify residual GC production.

In step 2, the cross-over design allowed us to evaluate changes in gene expression in the same individuals at normal cortisol (the CSHI arm) compared to very low cortisol levels (the OHC arm). We are, however, aware that our results instead could represent changes in gene expression related to the 2 different treatment modalities, CSHI and OHC. This would, however, require that the half-life of gene expression (RNA) levels exceeded 15 hours, corresponding to the minimum length of medication fasting in the OHC arm. Although we do not know the exact half-life of all the included genes, the median half-life of gene expression in general is estimated to be 7.1 hours.[37]

Finally, since this was an exploratory study, we acknowledge that interpretation of our data must be done with caution. In particular, without a control group of healthy individuals, we cannot know whether the observed differences in gene expression are changes toward a more normal physiological state or quite the reverse. In a future clinical study, we therefore advocate the need for a control group of healthy individuals to establish reference ranges for gene expression levels. Nevertheless, by presenting the data as-is, we provide a basis for future studies to explore the true potential and validity of gene expression as biomarkers for GC replacement in patients with AAD. Specifically, we suggest a repeated-measures design where gene expression levels are measured in the same individuals with AAD on repeated occasions following different GC administration types and doses, compared with healthy controls.

In conclusion, we nominate the gene expression of DSIPI, DDIT4, and FKBP5 as candidate transcriptional biomarkers of GC replacement, followed by CEBPB, VDR, POU2F1, ARID5B, ADARB1, and MMP9.