albumin. It was proposed that the MHC-related Fc receptor for IgG (FcRn) protects albumin from intracellular catabolic degradation, as it does for IgG, accounting for the uniquely long half-lives of both molecules and explaining their similar concentration-catabolism relationships (73). Both albumin and IgG do indeed bind to FcRn; however, it should be noted that they possess different binding sites so that binding to one does not influence that to the other.

Impact of Charge (pI), Size, and Valency on Binding/Pharmacokinetic Correlations

Valency, size, pI, and affinity for target antigen are all important parameters of a given mAb, affecting the extent and penetration/permeation kinetics of target uptake. Originally, the concept that changes in the pI value of an Ab, realized through charge modification, could improve target distribution was investigated in 1991 by Khawli et al. who used both in vitro and in vivo assays to confirm a preservation of antigen binding following chemical modification of chimeric tumor necrosis treatment or therapy (chTNT)-l mAb (74-76). This chimeric antibody (chAb; mouse variable, human IgG1/kappa constant regions, Cotara®) targets solid tumors by binding to common histone antigens found in the central necrotic core. Using several experimental approaches including in vitro immunoreactivities, binding affinities, and serum stabilities, the investigators related the effects of chemical and charge modification of chTNT, through conjugation to a small organic molecule (i.e., biotin), on the PK and targeting characteristics with the intent of determining their clinical potentials (76-78). Avidity binding studies demonstrated that biotinylation did not interfere with in vitro binding to fixed Raji cells, and immunoreactivities of both biotin-modified and unmodified chTNT Abs were 65% to 69%, respectively (76). PK studies also revealed that the charge-modified Abs exhibited a faster blood clearance, while in vivo biodistribution studies indicated a preservation of tumor uptake with lower normal organ uptake (75,76). These phenomena were attributed to a reduction in the net positive charge of the Ab, which lessens the influence of electrostatic interactions with negatively charged mammalian cell membranes (79).

Several investigators have similarly demonstrated that modification of the net charge of an Ab also correlates with an altered PK without affecting the in vitro antigen binding. For instance, Lee and Pardridge have related the immu-noreactivities of charge-modified anti-epidermal growth factor receptor (EGFR) Abs to their PK profiles (80). The 528 murine mAb recognizing the human EGFR was sequentially cationized with hexamethylenediamine and conjugated with diethylenetriaminepentaacetic acid (DTPA) as a potential radioimmunoconjugate for imaging EGFR-expressing cancer (80). An immunoradiometric assay showed comparable affinities for the modified and native anti-hEGFR 528 mAb; however, significant differences were observed in the PK behaviors of the two formulations (80). In this example, the discrepancy between in vitro and in vivo data is rationalized by the fact that the charge modification influenced the in vivo PK profiles without affecting the in vitro binding affinities. More recently, a similar in vitro-in vivo approach was applied to a series of genetically engineered mAb chTNT-3 derivatives of varying size and valency. In this regard, several variants were developed and evaluated including the scFv, diabody, triabody, Fab, and F(ab')2 (81,82). In general, results from these studies demonstrated a preservation of immunoreactivity and better tumor penetration for

28 Boswell et al.

these variants; however, lower overall tumor uptakes were observed in vivo because of faster elimination kinetics (81).

In Vitro Binding Correlated to In Vivo Binding/Targeting

Preclinical evaluation of mAbs as candidates for diagnosis and therapy often includes the correlation of in vitro antigen binding to in vivo tissue distribution. In this context, a commonly utilized strategy involves an in vitro cell- or antigen-binding assay followed by assessment of in vivo targeting by molecular imaging. Radioactive (12) and/or fluorescent (83) probes are often employed because of high sensitivity for both in vitro and in vivo measurements. Although high in vitro binding affinity is often a strong early indicator of in vivo targeting ability, data interpretation in such scenarios must be approached with caution because of the influence of stroma or other connective tissues, the presence/absence of endogenous factors, and the influence of antigen shedding. In addition, receptor expression levels tend to fluctuate by varying degrees between in vitro and in vivo models and can affect serum clearance and half-life of mAbs in clinical situations (18,84). A further complication limiting the utility of in vitro models is the absence of several physiological (in vivo) parameters including blood flow, vascular permeability, and interstitial pressure (85-87). In addition, the quantitative correlation of in vitro and in vivo targeting using fluorescent probes is quite challenging because of fluorescence attenuation by tissues, quantum yield modulation by local conditions (e.g., low pH in lysosomes), and other complicating factors. Radioactive probes lack such complications, typically require less overall structural modification than fluorescent probes, and are therefore the preferred method of detection for many applications requiring high sensitivity with maximally preserved immunoreactivity.

Binding/Targeting Correlations

Several studies have demonstrated that in vitro antigen-binding affinities can have an important bearing on in vivo mAb targeting. For example, d-related human leukocyte antigen (HLA-Dr), an antigen expressed on all human B cells, monocytes, and activated T cells, has been explored as a target of the mAb Lym-1. Khawli et al. related the in vitro binding (immunoreactivities via a live cell-binding assay) of this murine monoclonal IgG2a Lym-1 labeled through different methods of radioiodination to their in vivo targeting abilities (via biodistribution and imaging studies in tumor-bearing mice) (74). Furthermore, the same group also related the in vitro immunoreactivities of a chimeric Lym-1 (IgG1, k) and other Abs to their in vivo clearance rates (75). A similar approach was applied by Lewis et al. to 1A3, an IgG1 of the k isotype that binds to an unspecified colorectal adenocarcinoma antigen. They related the immunoreactivities and serum stabilities to the in vivo distributions of 1A3 conjugated to 64Cu-TETA (1,4,8,11-tetraazacyclotetradecane-1,4,8,11-tetraacetic acid) using two different linkers (88). Overall, the in vitro serum stabilities and immunoreactivities of the two different preparations were in good agreement, and the in vivo biodistribution profiles of both radioimmunoconjugates were also nearly identical (88). In general, the results from both Lym-1 and 1A3 demonstrated favorable binding both in vitro and in vivo and supported the concept that the in vitro binding is a valuable predictor of specific Ab targeting.

Binding/Targeting Noncorrelations

Some examples of targets with poor IVIVC have also been revealed, including the failure of in vitro antigen-binding assays to predict in vivo localization of the anti-phosphocholine murine IgM, HPCM2 (hybridoma phosphocholine monoclonal 2) (89). In these studies, Rodwell et al. compared in vitro binding and in vivo targeting of HPCM2 labeled by different methods of attachment (through oligosaccharides, tyrosines, or lysines) using 125I or mIn-DTPA (89). For instance, results from this work showed greater localization efficiency for both indirect 125I-labeled and mIn-DTPA-labeled radioimmunoconjugates (having 125I or mIn attached through residualizing probes) relative to the corresponding conjugate labeled with 125I nonselectively on tyrosines (89). These findings did not agree well with in vitro cell-binding data that did not distinguish between the two methods of radioiodination, demonstrating that in vivo localization to small subcutaneous xenografts is a more stringent test of Ab binding than in vitro cell-binding studies (89). In this case, the chemical method of radionuclide attachment is ultimately responsible for this discrepancy; direct radioiodination of tyrosine residues results in aryl-halogen bonds that are susceptible to in vivo enzymatic cleavage by dehalogenases with subsequent clearance from target sites, whereas radioiodinated oligosaccharides and radiometal-chelate complexes like mIn-DTPA are residualizing, meaning that the residualizing label persists indefinitely within target sites even upon Ab metabolism, typically as an adduct with the amino acid to which it was conjugated (e.g., lysine or cysteine).

In recent work, a lack of correlation between in vitro and in vivo binding was attributed to the high receptor binding affinity of 14C5, a murine IgG1 mAb directed against a yet undefined molecule involved in cell substrate adhesion originally discovered on the surface of cells within malignant breast cancer tissue (90). Burvenich and colleagues studied the in vitro (internalization) and in vivo (biodistribution) targeting properties of radioiodinated 14C5 Ab in non-small-cell lung cancer and colon carcinoma models (90). A significant difference in the level of binding and internalization of 14C5 into cultured A545 human lung carcinoma (high antigen-expressing) and LoVo human colon carcinoma (low antigen-expressing) cells was demonstrated by an in vitro radioimmuno-assay and by confocal microscopy (90). Nevertheless, both the high- and low-14C5-expressing tumors showed good in vivo tumor uptake, both having approximately 10% injected dose per gram of tumor tissue at 24 hours postinjection, and this result was attributed to the high binding affinity of 14C5 for its antigen (90). It has been suggested that high-affinity Abs are capable to effectively bind both high- and low-density antigen, whereas a low-affinity Ab only binds appreciably to high-density antigen because of a requirement for the avidity conferred by divalent binding for effective attachment (91).

Another lack of IVIVC was observed upon evaluation of three IgG mAbs prepared by standard hydridoma methods using osteosarcoma cells resected from an untreated patient (92). Sakahara et al. explored the relationship between the levels of in vitro binding and in vivo tumor accumulation of these radiolabeled anti-osteosarcoma mAbs, concluding a lack of IVIVC (92). Discrepancies between cell binding and in vivo tumor uptake were observed and were related to very different blood clearance rates (92). These results suggest that, while binding studies may be used to exclude Abs having poor in vitro binding, in vivo serum clearance may be a better test for choosing Abs with similar binding.

30 Boswell et al.

In summary, a number of factors can lead to poor correlation between in vitro binding and in vivo targeting. The selected examples above described noncorrelations resulting from (i) the chemical nature and metabolic stability of the probe, (ii) the interplay between binding affinity and antigen density, and (in) limited exposure due to rapid blood clearance rate. In addition, in all of these cases, the noncorrelation can be attributed to inherent differences between in vitro and in vivo models. For instance, (i) dehalogenation may be more pronounced in vivo because of higher enzyme levels and greater opportunity for subsequent clearance, (ii) different thresholds for avidity may exist between in vitro and in vivo models because of variations in antigen density, and (iii) the availability for antigen binding can be PK limited because of unfavorably rapid clearance.

In Vitro Binding Correlated to In Vivo Pharmacodynamics

In general, efficacy can be defined as the relationship between antigen occupancy and the ability to initiate a pharmacological (i.e., PD) response at the molecular, cellular, tissue, or system level. On the other hand, potency is a measure of the concentration of a drug at which it is effective. Bridging these two concepts is PD, the study of the biochemical and physiological drug effects, the mechanisms of drug action, and the relationship between drug concentration and effect.

Correlation of in vivo PD exclusively to in vitro binding is a somewhat rare scenario, as most studies tend to also examine the in vitro potency. One of the few examples is TRX1, a nondepleting humanized antihuman CD4 monoclonal IgG1 Ab being developed to induce tolerance by blocking CD4-mediated functions (93). Ng et al. related TRX1-induced internalization of CD4 and sub-cellular localization of T cell internalized TRX1 Ab in human T cells to the in vivo downmodulation of CD4 and clearance of TRX1 (93). TRX1 displayed nonlinear PK behavior, and TRX1 treatment induced saturation and down-modulation of CD4 receptors on T cells (93). Results from in vitro studies using purified human T cells suggested that CD4-mediated internalization may constitute one pathway by which TRX1 is cleared in vivo. The observed in vivo PD effect, downmodulation of CD4, was also explained by the in vitro studies that indicated binding and subsequent internalization of TRX1 into purified human T cells accompanied by TRX1-induced internalization of CD4 (93).

The anti-general control nondepressible (GCN)4 scFv fragment directed against the Gcn4p dimerization domain represents another example in which in vivo PD is correlated to an in vitro phenomenon other than potency (94). Worn et al. investigated the interplay between in vitro stabilities, binding affinities, and potencies of a series of these cytoplasmically expressed scFv intracellular Abs (i.e., intrabodies) (94). The in vivo performance (i.e., cytoplasmic inhibition) of the Ab fragments, measured as decreasing b-galactosidase reporter gene activity, was related to their in vitro stability, measured by denaturant-induced equilibrium unfolding (94).

0 0

Post a comment