Osteoporosis has been associated with pulmonary silicosis in California horses exposed to soils rich in cytotoxic silica dioxide crystals, a syndrome termed silicate associated osteoporosis (SAO). The causal mechanism for the development of osteoporosis is unknown. Osteoporotic lesions are primarily located in bone marrow-rich sites such as ribs, scapula and pelvis. Gene transcription patterns within bone marrow and pulmonary lymph nodes of affected horses may offer clues to disease pathobiology. Bone marrow core and tracheobronchial lymph node tissue samples harvested postmortem from affected and unaffected horses were examined histologically and subjected to RNA sequencing (RNA-seq). Sequenced data were analyzed for differential gene expression and gene ontology. Metatranscriptomic and metagenomic assays evaluated samples for infectious agents. Thirteen of 17 differentially expressed transcripts in bone marrow were linked to bone and cartilage formation such as integrin binding bone sialoprotein (log2FC = 3.39, PFDR = 0.013) and chondroadherin (log2FC = 4.48, PFDR = 0.031). Equus caballus solute carrier family 9, subfamily A2 (log2FC = 3.77, PFDR = 0.0034) was one of the four differentially expressed transcripts linked to osteoclast activity. Osteoblasts were hyperplastic and hypertrophic in bone marrow from affected horses. Biological pathways associated with skeletal morphogenesis were significantly enriched in affected horses. The 30 differentially expressed genes in affected lymph nodes were associated with inflammatory responses. Evidence of infectious agents was not found. The SAO affected bone marrow molecular signature demonstrated increased transcription and heightened activation of osteoblasts. Increased osteoblastic activity could be part of the pathological mechanism for osteoporosis or a compensatory response to the accelerated osteolysis. Transcriptome data offer gene targets for inquiries into the role of osteocytes and osteoblasts in SAO pathogenesis. Viral or bacterial infectious etiology in SAO is less likely based on metatranscriptomic and metagenomic data but cannot be completely ruled out.