The family II cellulose-binding modules (CBM) from Thermobifida fusca Cel5A and Cel48A were cloned in the Escherichia coli/Streptomyces shuttle vector pD730, and the plasmids were transformed into Streptomyces lividans TKM31. CBMCel5A, and CBMCel48A, CBMCel6B were expressed and purified from S. lividans. The molecular masses were determined by mass spectrometry, and the values were 10595 ± 2, 10915 ± 2, and 11291 ± 2 Da for CBMCel5A, CBMCel6B, and CBMCel48A, respectively. Three different binding models (Langmuir, Interstice Penetration, and Interstice Saturation) were tested to describe the binding isotherms of these CBMs on bacterial microcrystalline cellulose (BMCC). The experimental binding isotherms of T. fusca family II CBMs on BMCC are best modeled by the Interstice Saturation model, which includes binding to the constrained interstice surface of BMCC as well as traditional Langmuir binding on the freely accessible surface. The Interstice Saturation model consists of three different steps (Langmuir binding, interstice binding, and interstice saturation). Full reversibility only occurred in the Langmuir region. The irreversibility in the interstice binding and saturation regions probably was caused by interstice entrapment. Temperature shift experiments in different binding regions support the interstice entrapment assumption. There was no systematic difference in binding between the two types of exocellulase CBMs—one that hydrolyzes cellulose from the nonreducing (CBMCel6B) end and one that hydrolyzes cellulose from the reducing end (CBMCel48A). © 2002 Wiley Periodicals, Inc. Biotechnol Bioeng 80: 380–392, 2002.