Diabetes mellitus is a metabolic disease that affects every part of the human body including the
skin. One of the most common complications of diabetes is skin disorders, including skin infections and
diabetic foot ulcers (DFUs) that account for the majority of diabetes-associated limb amputation and
hospitalizations, and result in a reduction in quality of life and economic burden.
The etiology of DFU involves the patients’ genetics, microbiome, and environmental factors that
together affect the severity, response to treatment, and outcome of ulcers. Long-term high blood sugar in
diabetics can lead to changes in skin texture, appearance, and ability to heal, yet the underlying molecular
mechanisms are understudied. To date, studies on the genetics of DFUs consider the expression of genes
individually and ignore the effect of their coordinated expression in biological systems. Here, we use
network analysis and topological properties to systematically investigate the dysregulated gene coexpression patterns in type II diabetic skin with transcriptome profiles of skin samples from the GenotypeTissue Expression database. Our work reveals a novel mechanism (miR-21-PPARA-NCOA6) associated
with the dysregulation of keratinocyte proliferation, differentiation, and migration in diabetic skin, where
NCOA6 is the hub gene and KHSRP and SIN3B are key coordinators. Additionally, we build a TF-miRNAmRNA regulatory network to describe its interactive connections.
With respect to the skin microbiome, the impaired immune response of diabetics may fail to prevent
bacterial colonization in affected tissue resulting in chronic infection and biofilm production. First, we
review the latest literature on DFU microbiology unveiled by next-generation sequencing technologies and
discuss the limitations and the promises of these approaches in measuring and monitoring wound
progression. Then, we conduct a meta-analysis across publicly available DFU microbiome datasets to assess
the effect of demographic and technical factors on the resulting microbiome and leverage this harmonized
dataset to train a predictive model for the wound outcome. We show that the wound microbiome predictive
model can classify DFUs as healing or seriously infected, and that the presence of two bacterial genera,
Actinomyces and Brevibacterium (Actinomycetia), are strong predictors of wound status. Our work also
demonstrates the significant impact of the study cohort, geographic location, sampling method, and
sequencing region on the observed DFU microbiome, and reveals how such factors can obscure pathogen
detection. Lastly, we develop a web-based database called WoundDB, which is a repository for both crosssectional and longitudinal datasets that will enable the reuse of wound-relevant microbiome data in future
studies.
This dissertation explores different facets of DFUs including host genetics and skin microbiome,
and offers (1) new insights into molecular mechanisms associated with susceptibility to cutaneous diseases
in diabetic populations; (2) a deeper understanding of DFU microbiome and molecular sequencing
techniques used for profiling; (3) a predictive model that associates DFU healing status with corresponding
dynamic microbiome for wound outcome classification; and (4) an online database called WoundDB that
provides a user-friendly interface for querying and downloading wound-relevant microbiome data. This
work expands our knowledge of the etiology of DFUs towards the development of wound management and
diagnostics.
Click to expand...