The four human IgG subclasses are remarkably similar in sequence and differ significantly only in the hinge exons ( Reviewed by Burton et al [34] and by Jefferis [35]). However, they exhibit major differences in their effector functions (as described above). By comparing effector functions and sequences it is possible to speculate as to which of the sequence differences are responsible for the observed functional differences [34],[35]. Using recombinant DNA techniques, including site-directed mutagenesis and exon shuffling, these speculations may be tested experimentally. Several groups have carried out such experiments using different combinations of antibodies with either the anti-dansyl specificity (Reviewed by Shin et al [19] and also by Morrison et al Chapter 6) or the anti-NP specificity described above and also elsewhere in this volume (Lund & Jefferis Chapter 7). These studies have identified some of the structural features which determine the different antibody effector functions.
Although the hinges of the four human IgG subclasses are quite different in sequence, early speculation that this might account for the large functional differences was not upheld experimentally. Thus in experiments with antibodies specific for dansyl, the hinge regions of IgG3 and IgG4 could be exchanged without significantly affecting complement activation, i.e. the activity of IgG3 with an IgG4 hinge was only slightly reduced, and IgG4 with an IgG3 hinge still did not activate complement [36]. In another series of experiments with antibodies specific for NP, the long internally repeated hinge of IgG3 was reduced in length by sequential deletion of the repeated exons. In this case the activation of complement increased at shorter hinge lengths [37]. Further studies have indicated that the CH2 domain, including the lower hinge region, contains the sequences responsible for differences in binding to Fc receptors [38],[39],[40],[41],[42], and to complement [43],[44]. These studies are reviewed in Chapters 6 and 7 this volume.
The CAMPATH-1 (CD52) specificity represents an ideal natural target antigen for investigating the structural requirements within antibodies for recruitment and activation of autologous effector functions (in this case for the destruction of human lymphocytes and lymphoid tumours). The activation of such autologous effector functions can be conveniently measured in-vitro [18] and at least for the rat IgG isotypes there has been a correlation between the ability to activate in-vitro effector functions such as ADCC and the in-vivo effectiveness of these antibodies [24],[25],[32]. Thus results from studies in this system are of direct relevance to an understanding of how antibodies work in-vivo.
Wild-type IgG1 with CAMPATH-1 specificity works well in both complement mediated lysis and ADCC [18],[31] whereas IgG4 does not work in complement mediated lysis and only works in ADCC with some donors [31]. By exploiting a combination of naturally occurring and specially engineered restriction sites a series of domain swap mutants between these antibodies was constructed and tested [31]. The results are summarised in Figure 3 . Initial swapping of the exons indicated that despite earlier speculation on segmental flexibility, the hinge region was not a major contributory factor to the observed functional differences in IgG1 and IgG4. The major differences seemed to correlate with the C H 2 domain of the antibody. A convenient conserved restriction site within the C H 2 domains of gamma 1 and gamma 4 allowed the N-terminal and C-terminal halves of the domain to be swapped independently. The functional differences in both complement and ADCC correlated with the C-terminal half of the domain. There are only four residues which differ in this region between IgG1 (Tyr 296; Ala 327; Ala 330; Pro 331) and IgG4 (Phe 296; Gly 327; Ser 330; Ser 331). Of significance perhaps is the Pro/Ser 331 change which, in the X-ray crystallographic structures of IgG1 Fc, occurs at the elbow of a loop just adjacent to the hinge region [45]. Perhaps there is an overall distortion of the structure as a result of this change which affects other epitopes involved in interactions with complement and with Fc gamma receptors. Unfortunately there are only a limited number of solved structures for human IgG and they are all of the IgG1 subclass. Models of IgG2, 3 or 4 would, of course, be of tremendous value in predicting how the amino acid differences might affect interactions with other molecules.