Swine erysipelas is one of the major swine diseases. Both attenuated and inactivated vaccines are available, but the exposure of animals to the whole microorganism Erysipelothrix rhusiopathiae can aggravate arthritic problems. In this context, many efforts have been done to develop a subunit vaccine that would be safer and able to prevent this disease. Furthermore, by this approach, risky cultivations of pathogenic bacteria could be replaced by safer GMO cultures. Thus, the main objective of the present study was to investigate the expression of SpaA surface protein in fed-batch cultivations of E. coli. The 1026 pb fragment codifying for the antigenic region of the SpaA protein was amplified by PCR from the E. rhusiopathiae chromosomal DNA and cloned into pGEM-T vector. After sequencing, the spaA fragment was cloned into the expression vector pET28a and used to transform E. coli BL21(DE3). Fed-batch cultivations were conducted at 37 β’ C, pH 7.0, using a modified LB medium. IPTG 0.25 mM was the inductor. The experiments were carried out in an automated 2-L bioreactor (Applikon, using LabView) connected to a gas analyser (Sick/Maihak S.710) and equipped with a capacitance probe (Fogale, Nanotech) for on-line viable cell mass monitoring. Key process variables were measured: glucose, organic acids, biomass (optical density, viable cell counting and dry cell weight). The rSpaA protein over-produced as inclusion bodies was purified in denaturating conditions using a Ni 2+ sepharose column, after washing with Triton-X100 and solubilization in urea. Unstructured modelling and parameter estimation by a simulated annealing algorithm were used to simulate growth, metabolite formation and glucose consumption under different feed profiles. From the simulation results, a two-stage feeding strategy was implemented, starting with an exponential feed rate (0.05e -0.13t L/h), which was switched to a constant feed rate (0.1 L/h) 1.5 h after inductor addition. The proposed feed profile ensured adequate glucose supply and prevented viability loss during the 4 h of induction, when the growth rate was sharply reduced. Acetic acid, lactic acid and formic acid accumulated at the beginning of the cultivations, but their concentrations decreased to very low values during the fed-batch phase. High levels of protein expression were detected after inductor addition. Furthermore, DO was successfully kept at 30% of saturation throughout the experiment, almost without oscillation, by a control algorithm implemented in Labview. The capacitance probe followed closely cell growth and it turned out to be a more reliable tool for process monitoring than the respiratory quotient.