Background P. gingivalis, a non-motile, rod-shaped, anaerobic, Gram-negative bacterium is one of the principal sources of periodontal disease. It possesses a number of potential virulence factors thought to be important in the disease process including 5 hemagglutinins (Hag). One of these is HagB. It is a well characterized nonfimbrial adhesin expressed on the surface of P. gingivalis. HagB is very pro-inflammatory and induces robust chemokine and cytokine responses in vitro and in vivo. Since the chemokine and cytokine responses seen from single cells grown in tissue culture often are not representative of the chemokine and cytokine profiles seen in clinical samples or biopsy specimens, we devised a co-culture model of keratinocytes, dendritic cells, and T-cells to test the hypothesis that chemokine and cytokine responses from co-cultured cells would be more representative of responses seen in clinical samples from individuals with periodontal disease than single cell models. Methods and materials HagB was prepared by cloning hagb of P. gingivalis (1.4 kb) into the vector pQE31 (QIAGEN Inc., Valencia, CA); expressed in E. coli M15(pREP4)pQE31-TX1; and isolated from E. coli lysates by affinity chromatography using a Ni-charged resin (Profinity IMAC Resin, BioRad, Hercules, CA) and examined by SDS-PAGE. Co-culture models were treated with 10 µg/ml HagB (Test) or 10 µg/ml HagB diluent (Control). At 64 hours the supernatants were collected. Chemokine and cytokine biomarkers GM-CSF, CCL3 (MIP-1α), CCL4 (MIP-1β), CCL5 (RANTES), IL-1α, IL-6, IL-8, TNFα, IL-12(p40), and VEGF responses were determined using Milliplex immunoassays. HagB responses were corrected by subtracting the constitutive responses detected in supernatants incubated with HagB diluent. Statistical differences among groups were determined on Log10 transformed biomarker concentration using JMP 10 (version 10.0, SAS, CAR; NC). Results Buffers (e.g. HagB diluent) did not induce a chemokine or cytokine response, however there was a gradual increase in chemokine and cytokine responses from cells at 64 hours. These were subtracted from HagB induced responses. Responses generally fell in 2 groups; in one group containing VEGF, IL-12(p40), IL-6, RANTES and GM-CSF, there were no significant differences among groups (p>0.05). In another group containing IL-1α, IL-8, MIP-1α, MIP-1β and TNF-α, there were significant differences among groups (p< 0.05). Interestingly these resulting responses fell in 2 categories- GM-CSF, IL-12, IL-1α, IL-6 were less than 25pg/ml and IL-8, MIP-1α, MIP-1β, RANTES, TNFα and VEGF were more than 25pg/ml. Some responses were driven by a particular cell type e.g. GM-CSF produced by dendritic cells, RANTES produced by T- cells and VEGF produced by T cells. There were similar responses in HagB-induced IL-8, MIP-1β, MIP-1α and TNFα responses by dendritic cells + keratinocytes and dendritic cells + keratinocytes + T cells. Conclusions Co-culture models can more realistically produce chemokine and cytokine responses to agonists than individual cultures of cells, which is important for predicting and assessing novel therapeutic treatments of periodontal disease.