Keloid disease (KD) is a cutaneous fibroproliferative tumour characterised by heterogeneity, locally aggressive invasion and therapeutic resistance. Clinical, histological and molecular differences between the keloid scar centre and margin as well as recent evidence of the importance of epithelial-mesenchymal interactions (EMI) in KD pathobiology contribute to the complexity and diversity of KD, which coupled with the lack of a validated animal model have hindered research and effective management. Despite significant progress in the field of KD research, reliance on conventional monolayer cell culture and whole tissue analysis methods have failed to fully reflect the natural architecture, pathology and complexity of KD in vivo. In order to address these challenges, a site-specific in situ approach was therefore employed here for the first time in KD research. The first aim of this work was to compare the value of this contemporary approach with traditional methods of tissue dissection. The second aim was to compare the genomic expression between well-defined, distinct keloid sites and normal skin (NS). The third aim was to develop and explore hypotheses arising from this site-specific gene expression profiling approach, so as to enhance understanding of KD pathobiology as a basis for improved diagnostic and therapeutic strategies in future KD management. The fourth aim was to probe these hypotheses with relevant functional in vitro studies. The current site-specific in situ approach was achieved through a combination of laser capture microdissection and whole genome microarray, allowing separation of epidermis from dermis for keloid centre, margin and extralesional sites compared with NS. This in situ approach yielded selective, accurate and sensitive data, exposing genes that were overlooked with alternative methods of dissection.Identification of significant upregulation of the aldo-keto reductase enzyme AKR1B10 in all three sites of the keloid epidermis (KE) in situ, implicated dysregulation of the retinoic acid (RA) pathway in KD pathogenesis. This hypothesis was supported by showing that induced AKR1B10 overexpression in NS keratinocytes reproduced the keloid RA pathway expression pattern. Moreover, co-transfection with a luciferase reporter plasmid revealed reduced RA response element activity. Paracrine signals released by AKR1B10-overexpressing keratinocytes into conditioned medium resulted in TGFβ1 and collagen upregulation in keloid fibroblasts, suggesting the disturbed RA metabolism exerts a pro-fibrotic effect through pathological EMI, thus further supporting the hypothesis of RA deficiency in KE. Gene expression profiling further revealed an upregulation of NRG1 and ErbB2 in keloid margin dermis. Exogenous NRG1 led to enhanced keloid fibroblast migration with increased Src and PTK2 expression, which were attenuated with ErbB2 siRNA studies. Together with the observed failure to recover this expression with NRG1 treatment, suggested the novel KD pathobiology hypothesis that NRG1/ErbB2/Src/PTK2 signaling plays a role in migration at the keloid margin. In addition to these hypotheses, LCM methodology with comprehensive analysis of the data permitted the development of additional novel working hypotheses that will inform future KD research, including inflammatory gene dysregulation and cancer-like stem cells that may contribute to the therapeutic resistance characteristic of KD.