Identification and characterization of sebaceous gland atrophy-sparing DGAT1 inhibitors

Abstract

Inhibition of Diacylglycerol O-acyltransferase 1 (DGAT1) has been a mechanism of interest for metabolic disorders. DGAT1 inhibition has been shown to be a key regulator in an array of metabolic pathways; however, based on the DGAT1 KO mouse phenotype the anticipation is that pharmacological inhibition of DGAT1 could potentially lead to skin related adverse effects. One of the aims in developing small molecule DGAT1 inhibitors that target key metabolic tissues is to avoid activity on skin-localized DGAT1 enzyme. In this report we describe a modeling-based approach to identify molecules with physical properties leading to differential exposure distribution. In addition, we demonstrate histological and RNA based biomarker approaches that can detect sebaceous gland atrophy pre-clinically that could be used as potential biomarkers in a clinical setting.

Figure 1. Representative compound structures from three chemical compound series as shown in Table 1.

Figure 2. DGAT1 inhibitors with high lipophilicity induce sebaceous gland atrophy.

Shown are hematoxylin and eosin stains of dorsal skin biopsies from DIO mice treated with either vehicle (A), Cpd1 (B), Cpd2 (C), or Cpd3 (D) for 14 days at doses indicated. Scoring refers to the histological adverse effect score as described in Table 1. Bar  = 50 µm. The corresponding sebaceous gland sizes (area) are plotted in (E) (and shown in Table S1).

Figure 3. Histopathology versus lipophilicity.

Calculated logD (clogD, A) or measured logD (mlogD, B) versus observed skin adverse effects. Data is jittered on the y-axis to improve visualization.

Figure 4. Distribution of logD values by method.

The frequency of logD values of DGAT1 compounds for mlogD (blue) and clogD (red).

Figure 5. RNA biomarkers for sebaceous gland atrophy in skin.

Shown are the 42 probesets, identified in the Training Set (Studies 1 and 2), that were regulated by skin-positive compound treatments (those that produced sebaceous gland atrophy) but not by the skin-negative compound treatments (the one that did not produce sebaceous gland atrophy). After excluding the absent probes (low intensity), these 42 probesets met the following cutoffs: 1.2 fold change and ANOVA p<0.01 between all 3 skin-positive compound treatments (red arrows) and their respective vehicle treatments, and ANOVA p>0.1 between the skin negative compound treatment (black arrow) and its respective vehicle treatment. The probesets for RIKEN genes were excluded. Plotted are the LogRatio values (+/− 4 fold fold scale) with magenta representing up-regulated probesets and cyan representing down-regulated probesets. Treatments from the independent Test Set (Study 4) are included for comparison but were not used to identify the 42 probesets.

Figure 6. Validation of the 42 probeset-composite score in an independent Test Set.

The 42 probesets from the Training Set, shown in Figure 5, were used to generate a composite score across the treatments from the independent Test Set (Study 4). The p value between Cpd6 treatment (skin-positive) and Cpd18 treatment (skin-negative), in Study 4, is less than 0.0001. Expression data from this set of RNA biomarkers is predictive for sebaceous gland atrophy in mice following DGAT inhibitor treatment.

Figure 7. Immune-regulated genes are up-regulated, while lipid metabolism genes are down-regulated, with sebaceous gland atrophy.

Box plots of probesets regulated by the skin-positive DGAT1 inhibitors (those that produce sebaceous gland atrophy) but not by the skin-negative compounds (those that do not produce sebaceous gland atrophy). Plotted are the LogIntensity values across the replicates in each group, and across the three studies. Ccl1 (A; chemokine (C-C motif) ligand 1) is involved in the recruitment of T cells in skin inflammation; and Scd3 (B; stearoyl-coenzyme A desaturase 3) is a sebaceous gland specific gene.