Personalized Healthcare in Autoimmune Diseases - Genomic Biomarkers for Pharmaceutical Development - page 62

the drug. However, not all non-responders develop NAbs, and if induced it takes at least six
months for NAbs to develop, which can also disappear again over time [59,60] .
Efforts to understand differential responsiveness have focused primarily on the mech-
anistic (that is, the primary) phase of unresponsiveness. Van Baarsen and colleagues were
the first to report that the baseline IFN type I signature determined the pharmacological
response to IFNβ [61,62] . The results revealed that an increased IFN response gene activity
at baseline was associated with the absence of a pharmacological effect of IFNβ-treatment
in RRMS. Comparing baseline expression levels of an optimal performing set of 15 type I
IFN response genes with pharmacological response outcome revealed a significant nega-
tive correlation (R = −0.7208; p = 0.0016). The negative correlation between pharmacological
response and baseline levels of IFN-induced genes is consistently observed over time, at one,
three and six months after the initiation of the therapy and was confirmed in an independent
cohort of 30 RRMS patients (p < 0.0085 and R = −0.4719) ( Fig. 3.7 ) . These results were con-
firmed by other investigators, who observed that an increase in the expression of type I IFN
genes was much lower or absent in the IFN high RRMS patients [63-65] .
Given the ameliorative role of IFN-β on disease activity in RRMS, these results appear
counterintuitive. One possible explanation for these observations is that the type I IFN activ-
ity in the IFN high RRMS patients was already saturated prior to the initiation of the therapy,
resulting in the absence of a pharmacological effect of administrated IFN-β in these patients.
Indeed, it was demonstrated that peripheral blood mononuclear cells (PBMC) of IFN high
RRMS patients had lost the capacity to respond to in vitro stimulation with IFN-β, consistent
with the in vivo findings [62] . Moreover, Zula and colleagues reported that IFN-α / β therapy in
patients with RRMS induced the expression of a STAT dependent gene signature in myeloid
cells of responders only [66] . Others observed that the IFN high status correlated with endog-
enous IFN-β and elevated IFN receptor 1 (IFNAR1) expression by monocytes [63] . Also the
capacity of IFN-β to induce its own expression was deficient in cells from IFN high patients [64] .
Overall the data convincingly support the conclusion that the in vivo pharmacological
response to IFN-β treatment is dependent on the intrinsic type I IFN response gene activity
status prior to the start of the therapy [67] . Obviously this could also have consequences for
the clinical outcome of IFN-β therapy. Indeed, baseline type I IFN response gene expression
levels were predictive of the clinical outcome of IFN-β treatment in RRMS [63,68] . Responders
were defined as having no increase in the Expanded Disability Status Scale (EDSS) score and
no relapses during the 24 month follow-up period. Non-responders had at least one relapse
and one point increase in EDSS score during that period. As anticipated from the pharmaco-
logical outcome study, Comabella and colleagues observed that non-responders were charac-
terized by an increased expression of type I IFN response genes at baseline. The predictive
accuracy of a gene set, consisting of predominantly type I IFN genes, reached 78%. These find-
ings were replicated in an independent group of RRMS patients (predictive accuracy 63%).
The genes that were selectively induced by type I IFN were found to be the best predictors of
efficacy. These results warrant further studies to validate the clinical utility of the IFN signa-
ture as a biomarker to predict the response to IFN-β treatment in RRMS [63] .
Accordingly, baseline phospho-STAT1 levels were found to be significantly higher in mono-
cytes derived from non-responders when compared to responders. Upon stimulation of the
monocytes with IFN-β, no differences between the phospho-STAT1 levels were observed
between responders and non-responders [63] . These findings are consistent with previous