In Vivo Model on Primates

Home , Assam macaque

Gene Therapy (2002) 9, 282–290  2002 Nature Publishing Group All rights reserved 0969-7128/02 $25.00 www.nature.com/gt RESEARCH ARTICLE Antibody-mediated lung endothelium targeting: in vivo model on primates

IV Balyasnikova1, DC Yeomans2, TB McDonald1 and SM Danilov1 1Department of Anesthesiology, University of Illinois at Chicago, IL, USA; and 2Department of Anesthesia, Stanford University, Palo Alto, CA, USA

We have recently provided evidence that angiotensin-con- MAb i2H5, which binds to macaque ACE with substantially verting enzyme (ACE) is a rational target and anti-ACE higher afﬁnity than mAb 9B9, also more effectively accumu- monoclonal antibodies (mAbs) are suitable molecules for lates in their lungs than mAb 9B9. Immunospeciﬁcity of lung directing gene/drug delivery into the pulmonary endothelium accumulation (mAb/control IgG ratio) was 37 for i2H5 and of rodents. As a step towards gene therapy clinical trials 0.5 for 9B9. Lung selectivity of i2H5 uptake (lung/blood ratio) using this approach, the present study evaluated the poten- was around 10. Therefore mAb i2H5 may be useful for in tial of anti-ACE mAbs for in vivo lung endothelium targeting vivo lung targeting in non-human primates, whereas 9B9 in 10 species of primates. Cross-reactivity of 10 distinct may be most useful in primates that are closer to humans mAbs directed to human ACE with ACE from baboon, (chimpanzee). A combination of these two mAbs may be macaques, cercopithecus and chimpanzee revealed that the particularly useful for human clinical trials of gene/drug ther- highest binding with ACE from baboon and macaques was apy for lung disorders such as pulmonary hypertension and with mAb i2H5, from chimpanzee – mAb 9B9, and from lung metastases. human – 9B9 and i2H5. Thereafter, in vivo biodistribution of Gene Therapy (2002) 9, 282–290. DOI: 10.1038/sj/gt/3301657 mAbs i2H5 and 9B9 was estimated in Macaca arctoides.

Keywords: pulmonary endothelium; angiotensin-converting enzyme; monkey; monoclonal antibodies; gene delivery

Introduction injury induced by hydrogen peroxide by the systemic mAb 9B9 conjugated with catalase.14 The angiotensin-converting enzyme (ACE, kininase II, This anti-ACE mAb was successfully used to increase CD143), an important regulator of vascular tone and selectivity and efficacy of transgene expression that have 1,2 remodeling, is expressed on the luminal surface of been incorporated into viral vectors. Thus, using a bispe- 3–6 endothelial cells of different types of blood vessels. The cific conjugate approach for direction of adenoviruses to unique tissue distribution of endothelial ACE, along with ACE, we achieved dramatically (20-fold) enhanced pul- 5,6 preferential expression in lung capillaries, also makes monary gene delivery and expression in vivo along with it an almost ideal target for therapy directed toward significantly reduced (five- to eight-fold) transgene pulmonary endothelium. expression in non-targeted organs in rats.15 Moreover, the Previous studies from our group demonstrated that combination of transductional retargeting adenoviruses monoclonal antibodies directed toward ACE may be (via ACE) and transcriptional retargeting (with the use of highly efficient and specific carriers for the delivery of an endothelial specific promoter for vascular endothelial therapeutic substances (isotopes, proteins, drugs) to the growth factor receptor 1, flt-1) resulted in a remarkable 7–10 lung vasculature. For example, we have shown that (300 000-fold), highly synergistic improvement in selec- after systemic injection, an anti-ACE monoclonal anti- tivity of transgene expression for lung compared with the body (designated 9B9) selectively accumulates in the usual site of vector sequestration, the liver.16 8,11 10 lungs of several mammals, including humans. Fur- The antibody-directed lung-selective in vivo gene deliv- thermore, systemic administration of mAb 9B9 that had ery system via ACE shows great potential in the rat.15,16 been conjugated to the plasminogen activators, catalase However, before this method can be applied to human or superoxide dismutase, results in the specific targeting clinical trials, this work must better approach the human and prolonged association of these drugs with the pul- condition. As recently reviewed by Donahue and Dun- 12,13 monary vasculature of rats. These conjugates retain bar,17 despite success in usual laboratory models on rod- biological activity and demonstrate therapeutic effects. ents and in vitro, ‘the unique predictive value of nonhu- For example, rat lung endothelium was protected from man primate models … has become more apparent, and major advances in gene transfer efficiency have been made utilizing these powerful, but expensive and com- Correspondence: SM Danilov, Anesthesiology Research Center, Univer- sity of Illinois at Chicago, 1819 W Polk Street (M/C 519), Chicago, IL plex systems’. The distinctions between primates and 60612, USA; [email protected] rodents can dramatically impact on the predictive value Received 10 July 2001; accepted 17 December 2001 of gene transfer trials in mice and rats for clinical applica- bility.17,18 The work presented here clearly bears this out. Lung endothelium targeting in primates IV Balyasnikova et al 283 Thus, this study focused on the ability of several mAbs was intermediate between human and rat. ACE activity directed against different epitopes of human ACE, to tar- was much higher in the kidney of studied animals rela- get lung endothelium in primates, and thus for delivery tive to lung tissue or in serum. The observed high ACE of transgenes to the pulmonary endothelium. This in vivo level in the kidney appears to conflict with the absence model performed in both rats and non-human primates of mAb 9B9 in kidney after in vivo antibody injection in should prove helpful in the pursuit of human clinical all animals studied previously.6–10 However, in the kid- studies. ney, ACE is most highly expressed in the epithelium of proximal tubules and is found as well in brush border of Results and discussion intestine.19 Thus, both of these areas are essentially inac- cessible to monoclonal antibodies in circulation.6–13 20–22 Tissue and plasma ACE activity in primates It is likely that the large endothelial surface area, This study was designed to develop an in vivo model for and the consequent high ACE levels in the lung contrib- utes to the selective accumulation of anti-ACE mAbs in ACE-dependent drug/gene delivery to the pulmonary 6–13 endothelium of primates. As we have previously demon- the lung endothelium after systemic injection. Another strated, anti-ACE mAb 9B9 specifically accumulates in contributing factor to this accumulation is the hetero- the rat, hamster and cat lungs after systemic injection due geneous distribution of ACE in human and rat endo- to a high concentration of ACE in the lungs of these spec- thelial cells along the vascular tree. On average, only 10– ies.11 However, data have not yet been described for ACE 20% of human and rat capillary endothelial cells in sys- activity in tissues and blood of primates. temic circulation demonstrate ACE expression; whereas, Data on ACE activity in plasma, lung and kidney of virtually all endothelial cells in pulmonary capillaries 5,6 baboon (Papio anubis), macaques (Macaca arctoides, Macaca strongly express ACE. This finding explains why mulatta, Macaca fascicularis, Macaca nemestrina, Macaca among several anti-endothelial mAbs studied (ICAM-1 assamensis, Macaca fuscata), cercopithecus (Cercopithecus (CD-54), PECAM (CD-31), Thy.1.1 (CD-90), ACE (CD- aethiops), as well as chimpanzee (Pan troglodytes) are 143)), anti-ACE mAbs demonstrated the most selective presented in Table 1. ACE activity in the lung of these lung accumulation after systemic injection.6 The selec- species was between three and eight times higher than tivity of ACE expression in the lung appears to be fairly in human, but between three and 10 times lower than in ubiquitous across mammals, as we have observed a simi- rat lung. It is likely that not only the absolute content of lar pattern of ACE expression in pulmonary versus sys- ACE in the lung, but also the gradient of ACE concen- temic circulation across all 26 species tested, including tration from lung to blood (the organ/blood ACE ratio) human and non-human primates (Franke et al, in determines the efficacy and selectivity of antibody lung preparation). accumulation.11 These data also show that the Therefore, the common cross-species character of endo- tissue/serum ACE activity ratio in non-human primates thelial ACE distribution pattern (Refs 5 and 6, and Franke

Table 1 ACE activity in tissues from different animal species

Animal Tissue ACE activity Tissue/plasma (units/g tissue) ratio

Human Homo sapiens Plasma 0.028 ± 0.01 1 Lung 0.52 ± 0.03 19 Kidney 2.34 ± 0.06 84 Chimpanzee Pan troglodytes Plasma 0.020 ± 0.004 1 Baboon Papio anubis Plasma 0.102 ± 0.001 1 Lung 4.2 ± 0.59 41 Kidney 13.6 ± 3.4 133 Macaque Stump tail/Bear macaque Plasma 0.046 ± 0.014 1 Macaca arctoides Lung 1.34 ± 0.51 29 Kidney 12.1 ± 1.2 263 Rhesus macaque Plasma 0.090 ± 0.01 1 Macaca mulatta Lung 4.3 ± 0.5 48 Kidney 5.6 ± 0.6 62 Cynomolgus macaque Plasma 0.054 ± 0.006 1 Macaca fascicularis Lung 5.56 ± 0.42 103 Kidney 4.83 ± 0.58 86 Japanese Snow macaque Plasma 0.087 ± 0.005 1 Macaca fuscata Lung 5.0 ± 0.204 57 Kidney 9.0 ± 0.8 104 Assam macaque Macaca assamensis Plasma 0.048 ± 0.01 1 Pigtail macaque Macaca nemestrina Plasma 0.075 ± 0.02 1 Cercopithecus African Green Monkey Cercopithecus aethiops Plasma 0.048 ± 0.006 1 Rat Sprague–Dawley rat Plasma 0.262 ± 0.014 1 Rattus norvegicus Lung 13.5 ± 0.62 52 Kidney 0.25 ± 0.05 1

The plasma, kidney and lung tissues were prepared as described in Materials and methods. The plasma samples and supernatants containing solubilized ACE were used for ACE activity measurement using Z-Phe-His-Leu as ACE substrate. The data are presented as mean ± s.d., n = 3.

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 284 et al, manuscript in preparation) and high ACE content in the lung of studied primates (Table 1) suggested that these species will also demonstrate lung endothelium targeting with anti-ACE mAbs.

Cross-reactivity of anti-human ACE mAbs with primates ACE In order to choose the most suitable species for the devel- opment of the in vivo model of ACE-directed lung endothelial targeting on primates, we first estimated cross- reactivity of a panel of 10 anti-human ACE mAbs with ACE from tissue and plasma of the primates, most widely used in biomedical experiments (Table 1). We have previously demonstrated that binding affinity of some anti-ACE mAbs greatly depends on the conformation of ACE. For example, the binding strength of mAbs i1A8 and 5F1 for ACE with hydrophobic transmembrane anchors was three- to five-fold higher than with soluble ACE, whereas mAb 3A5 binds more strongly with soluble ACE.23,24 Thus, we used both membrane-bound ACE (solubilized from the tissues by detergent and equilibrated to 20 mU/ml of ACE activity) and soluble ACE (from plasma, without hydrophobic transmembrane anchor, equilibrated to 2 mU/ml of enzymatic activity) to examine binding of the test antibodies. The results of these experiments demonstrated that two mAbs, i2H5 and 5C5, directed toward almost identical epitopes, demonstrated high levels of binding with plasma and lung ACE from human, baboon, and all studied macaque species, but did not bind chimpanzee ACE (Figure 1). Interestingly, both of these antibodies demonstrated essentially equal binding for both forms of ACE. In contrast, mAb 9B9 (previously studied mostly in rats) Figure 1 Precipitation of ACE activity from tissue samples of different species by a panel of mAbs to ACE (cross-reactivity of anti-human ACE showed strong binding with ACE from human and chim- mAbs). (a) Precipitation of ACE activity from plasma of human, baboon panzee (the most human-like primate) and very weak (Papio anubis), bear macaques (Macaca arctoides), chimpanzee (Pan binding to ACE from other non-human primates. More- troglodytes) with the panel of mAbs to ACE, was detected using ACE over, mAb 9B9 did not precipitate macaque plasma ACE precipitation assay.23 (b) The precipitation of ACE activity by anti-ACE at 2 mU/ml at all, but did precipitate ACE from lung mAbs was repeated with lung homogenates from most of the studied pri- homogenates from all macaque species when tested at a mate species. Microtiter plates, coated by different anti-ACE mAbs, via goat-anti-mouse IgG, were incubated with plasma (2 mU/ml) or kidney higher concentration (20 mU/ml). These results indicate homogenate (20 mU/ml) of studied species. Then, the precipitated ACE that the affinity of mAb 9B9 to macaque ACE is approxi- activity was estimated directly in the wells with Hip-His-Leu as sub- mately two orders of magnitude (40–100-fold, depending strate.23 Negligible background hydrolysis of the substrate in the wells, on the species) lower compared with the affinity of mAb coated by non-immune mouse IgG (negative control), was subtracted from 9B9 for human ACE. each value with specific anti-ACE mAbs. Results (% of precipitated ACE Although we have shown that mAb 9B9 accumulates activity expressed as a % of binding) are shown as a mean of triplicates (variability did not exceed 10% between replicates). Precipitation of ACE selectively in the lungs of human volunteers and patients activity from plasma and lung homogenate of other macaque species, men- 10 with sarcoidosis, most of the other pre-clinical studies tioned in Table 1 (Macaca fascicularis, mulatta, fuscata, nemestrina and with this particular mAb have been performed using rats, assamensis), cercopithecus (Cercopithecus aethiops) was similar, if not where this antibody demonstrates strong selective identical, for all species with one exception. Precipitation of ACE from accumulation in the lung. However, to bring this work Japanese Snow macaque (Macaca fuscata) was significantly lower for to humans, it is clearly critical to examine non-human mAbs that recognize overlapping epitopes and belongs to the IV antigenic region on the N-domain of ACE23– mAbs i2H5, 5C5, IG12, 6A12, 7A2 primates. This gap between laboratory small animals and (not shown). humans could be filled with mAb I2H5 and 5C5, which demonstrated robust binding with ACEs from all non- human primates studied. transmembrane anchor), but also on the flexibility of ACE, in other words, on the availability of the epitope Lung targeting of the systemically injected anti-ACE for a particular mAb in the circulation. For example, mAbs binding of mAb i1A8 with ACE (somatic two-domain It is clear from our previous work and other publications ACE with transmembrane anchor) in solution was the that binding of mAb to ACE in solution does not guaran- lowest among the 10 studied mAbs to ACE, whereas tee the recognition of ACE by this particular mAb on the binding of this antibody with ACE on the surface of cell surface of capillaries in vivo, and subsequent selective CHO-ACE cells was the highest.24 accumulation in the lung. Our recent experiments indi- The fact that the efficacy of the endothelial targeting cate that binding of mAbs to ACE depends not only on is epitope-dependent, was confirmed in the studies with the size of ACE (short, soluble ACE versus long ACE with another anti-endothelial mAbs. Thus, despite similar

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 285 affinities, mAbs to different epitopes of PECAM-1 demonstrated pronounced difference in internalization after binding with endothelial cell surface and subsequently, in pulmonary uptake.6 From four studied monoclonal antibodies, recognizing different rat lung endothelial antigens on tissue slices, Western blotting or in cultured cells, only two of them demonstrated selective pulmonary uptake.25 Therefore, only in vivo experiments can confirm the feasibility of different anti-ACE mAbs for ACE-dependent, antibody-directed lung endothelium targeting. Either mAb 9B9 or i2H5, or non-immune mouse IgG (control) was injected intravenously into stump-tail/Bear macaques (Macaca arctoides). To quantify the tissue distribution (and lung accumulation, in particular) of non-radiolabeled mAbs to ACE, we used a plate precipitation assay (see Materials and methods). The amount of mAbs accumulated in various organs was estimated indirectly by assaying the level of ACE activity, precipitated from tissue homogenates or plasma of studied animals to the plate, covered by anti-mouse IgG. MAb i2H5, in agreement with in vitro results of the cross-reactivity with macaque ACE, demonstrated a substantial and selective accumulation in the lung. More- over, and consistent with our previous results,11 there was no detectable selective accumulation of i2H5 in kidney. Control mouse IgG did not accumulated in the lung, whereas mAb 9B9 demonstrated weak, albeit significant, accumulation in the lung (Figure 2). The immunospecificity index (mAb/mIgG ratio) was 37 for mAb i2H5, but only 0.5 for mAb 9B9. Lung selectivity of i2H5 biodistribution (lung/blood ratio) was around 10. The selectivity of antibody accumulation in the organ Figure 2 The distribution of anti-ACE mAbs i2H5 and 9B9 in tissues of of interest (in our case in lung) greatly depends on the Macaca arctoides after systemic injection. mAbs to ACE (i2H5 and 9B9) injected dose.8 With the relatively modest dose of mAbs and control mouse IgG (1 mg each) were injected intravenously. Within used in this experiment (0.1 mg/kg) lung selectivity is 2 h, blood, kidney and lung tissues were taken from dead animals and were processed as described in Materials and methods. The amount of the quite high. It is interesting to note that the i2H5/9B9 ratio mAbs accumulated in the organs studied was estimated by the amount of lung uptake in Macaca arctoides (Figure 2) was similar of ACE activity, complexed by these mAbs in plasma, lung or kidney (above 30) to the i2H5/9B9 ratio in ACE binding from homogenates, by precipitation on to the plate, covered by rabbit-anti- lung of these species-cross-reactivity in vitro experiments mouse IgG (plate precipitation assay23). ACE activity precipitated from (Figure 1b). Therefore, the relative in vivo lung accumu- homogenates of the organs of monkeys, injected with non-immune mouse lation of mAbs 9B9 and i2H5 in Macaca arctoides very pre- IgG (negative control), was subtracted from each value with specific, anti- ACE mAbs i2H5 and 9B9. (a) Precipitated ACE activity. The data rep- cisely reflects their differences in binding affinity to ACE resent the results of three independent experiments and expressed as mean from this species. Despite the fact that affinity of mAb ± s.d. (b) The ‘immunospecificity’ of the mAbs accumulation (estimated 9B9 to macaques ACE was much lower in comparison as the precipitation of ACE complexed with these mAbs) was also mAb i2H5, the selectivity of the mAb 9B9 accumulation expressed as signal/background ratio. in the lungs of macaques was still very high (see result of the gamma-imaging of lung uptake in macaque rhesus presented in Figure 3). This might be explained by the thelium precludes intracellular delivery of a drug or gen- fact that binding affinity of mAb 9B9 with human ACE etic material.27 However, in the case of PECAM-1, conju- is very high – in the low picomolar range,26 and even a gation of anti-PECAM mAbs to streptavidin markedly significantly lower affinity of binding of mAb 9B9 with facilitates pulmonary targeting and internalization of macaques ACE should still provide high selectivity of anti-PECAM mAbs.6,28 The situation with thrombomodu- pulmonary endothelium targeting. lin is different: although endothelium internalizes anti- The internalization of an antigen or antigen–antibody thrombomodulin mAbs, it undergoes rapid cellular complex, is one of the key determinants (along with intra- degradation,29 thus also precluding intracellular delivery, cellular trafficking and degradation) for successful gene and possible its potential as gene transfer vector. transfer into the cell of interest, and into pulmonary The mAbs to ACE are more suitable for intracellular endothelial cells in particular. Lung vascular immunotar- delivery and possible gene transfer. Endothelial cells in geting has been demonstrated using antibodies to several culture, in perfused rat lungs and rat lungs in vivo different endothelial antigens, such as ACE, thrombomo- internalized mAb 9B9 without significant intracellular dulin, PECAM-1, ICAM-1, caveoli-associated antigens.6,27 degradation.30 Both mAbs studied in the present paper However, many of these molecules have characteristics (9B9 and i2H5) are also very effectively internalized by that suggest that they may not be ideal targets. Thus, CHO cells, transfected with ACE cDNA.24 poor internalization of ICAM-1 and PECAM-1 by endo- However, these studies did not address the mechanism

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 286

Figure 3 Lung accumulation of radiolabeled mAb 9B9 in macaque rhesus (Macaca mulatta). Anterior thoraco-abdominal gamma-scintigrams shows localization of 111-In-anti-ACE mAb 9B9 (right) and non-immune mouse IgG (left) in macaque rhesus 1 h after i.v. injection of radiolabeled antibodies – 200 ␮g, 0.5 mCi. Background radioactivity was subtracted from each image. (Reprinted by permission of the Society of Nuclear Medicine from Ref. 7.)

of anti-ACE mAbs internalization by ACE-expressing cells. Very little is known about the internalization of ACE or ACE ligands, ACE recycling in the endothelial cells, or the effect of ACE ligands on ACE internalization. ACE is a transmembrane glycoprotein that is not con- sidered (given our present knowledge) to possess receptor or signaling function.30 The currently accepted path- way of ACE metabolism in the plasma membrane is its Figure 4 The effect of anti-ACE mAbs 9B9 and i2H5 binding on the shed- shedding to the extracellular medium.31 A systematic ding of ACE from the cell surface of CHO-ACE cells. The CHO-ACE cell line (clone 2C2) was grown on 96-well plate until confluence. mAbs i2H5, study of ACE–antibody interaction on the surface of rab- 9B9 or control mouse IgG at the concentration of 10 ␮g/ml, were added bit endothelial cells was performed in the laboratory of to the cells in serum-free medium (containing 2% BSA) and incubated G Andres32 using polyclonal antibodies to ACE. In a ser- (1 h, 4°C) for mAb binding assay (a) and 4 h at 37°C for ACE shedding ies of publications this group demonstrated that divalent assay (b). (a) mAb binding assay Bound anti-ACE mAbs were revealed polyclonal antibodies increased endothelial permeability with goat-anti-mouse IgG, conjugated with alkaline phosphatase.24 (b) via complement fixation. ACE shedding assay. The culture fluids were collected, centrifuged (for the precipitation of the detached cells) and the enzymatic activity of ACE, Mabs to ACE, in contrast to polyclonal antibodies, does shed from the plasma membrane of cells, were measured as described in not induce complement-mediated injury to cultured Materials and methods. Antibody-induced shedding (mAb i2H5 and 9B9) endothelial cells.11 No pathological changes were is expressed as % from the basal shedding of ACE (in the presence of detected in organs of any animals studied, nor in human control mouse IgG only). The data represent the results of several inde- ± volunteers.7,8,10,11 The mechanism for internalization of pendent experiments and expressed as mean s.d. mAbs to ACE still unclear. One possibility is that the ACE molecule undergoes constitutive turnover and Antibody-induced shedding of ACE from the cell surface recycling in the endothelial plasma membrane, like some As discussed previously, the potential of mAb 9B9 to be other endothelial membrane proteins.33,34 If such consti- used as a carrier for drug/gene delivery to the lung is tutive ACE turnover occurs, ACE could serve as a carrier somewhat limited in that this antibody has been shown for intracellular transfer of anti-ACE mAbs bound to to induce shedding of ACE from the plasma membrane endothelial cells. The second possibility is that ACE itself of endothelial cells, both in vitro and in vivo conditions.11 is not normally internalized, but the interaction of mAbs In order to determine whether our other antibodies to ACE induced internalization of the complex.30 A simi- would have similar limitations, we compared the ACE lar mechanism might be responsible for the internaliz- shedding capacity of mAb i2H5 and 5C5 with that of 9B9. ation of anti-PECAM mAbs: internalization of anti- To do this, we incubated the antibodies with a CHO cell PECAM I mAbs was very low, but dramatically facili- line that expresses ACE (clone 2C2), as well as with tated by conjugation of mAbs to streptavidin.28 HUVEC. The results of this experiment (Figure 4) indi- However, regardless of the mechanism both mAbs to cate that, while mAb i2H5 binds to ACE on the cell sur- ACE-i2H5 and 9B9, internalized by endothelium, selec- face of CHO-ACE cells, this binding does not induce sig- tively accumulates in the lung of macaques after systemic nificant shedding of ACE from the plasma membrane. injection and thus are candidates to be used for intra- Binding to ACE (Figure 1) and the effect on ACE shed- cellular drug/gene delivery into the lung endothelium of ding for mAb 5C5 (not shown) was similar to that for studied non-human primates. mAb i2H5.

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 287 In contrast, mAb 9B9 induced ACE release at a rate compared with the liver and spleen. Reporter gene that was more than twice (216%) the basal, constitutive expression in the lung becomes 200–300-fold more than rate of ACE release (Figure 4). Shedding of ACE, in the liver and spleen even in absolute terms, thus increased by mAb 9B9 binding, was accompanied by achieving the improvement in relative selectivity 300 000- weak, albeit significant loss of surface ACE (not shown). fold for the lung:liver ratio and 6000-fold for lung:spleen The same effect of 9B9 on ACE shedding (and the absence ratio when compared with untargeted vector.16 Thus, in of similar effects for mAb i2H5) was demonstrated for general, a high selectivity of gene delivery to the lung HUVEC (not shown). The normal rate of the ACE shed- endothelium has been achieved, and the efficacy of trans- ding ranges between 5 and 15%/24 h depending on the gene expression via ACE targeting becomes the next cell-type, ie 5–15% of the amount of ACE on the cell priority. membrane appeared in the culture medium after 24 h.24,35 To examine means of improving transgene expression This calculation explains, why a substantial increase in efficacy, we also studied whether the efficacy of ACE shedding, induced by mAb 9B9 (two-fold in com- drug/gene delivery to the pulmonary endothelium of parison with basal ACE shedding) leads to only a small human could be enhanced if we were to use a mixture loss of cell surface ACE. Moreover, there is indirect evi- of mAbs 9B9 and i2H5, both of which recognize human dence of reappearance of ACE on the cell surface, sug- ACE (Figures 1 and 3, this paper, and Refs 23 and 24), gesting the recycling of ACE. Our attempts to deplete but are directed to different, non-overlapping epitopes.23 ACE from the surface of cultivated human umbilical vein Saturation of ACE binding for both mAbs i2H5 and 9B9 endothelial cells by repeated changing of the culture occurs at a concentration of approximately 1 ␮g/ml, both medium (every 4 h) resulted in shedding of the ACE to with purified human kidney ACE adsorbed on plastic the culture medium in amounts greatly exceeding (Figure 5a), as well as on the surface of ACE-expressing (several-fold) the amount of ACE remaining on the sur- CHO cells (not shown). The mixture of these two mAbs face of these cultivated cells (Danilov et al, unpublished at 1 ␮g/ml revealed an additive character of binding to observation). This experiment demonstrates that ACE ACE (Figure 5b and c). Additivity of mAb 9B9 and i2H5 shedded from the surface was very efficiently replaced binding was not complete however (Figure 5c), probably by newly synthesized or recycled ACE. due to the slightly different conformation of human ACE Taken together, the high level of mAb i2H5 binding to adsorbed on the plastic when compared with ACE anch- ACE-expressing cells in vitro, accumulation in the lung in ored to the surface of CHO cells. Nonetheless, these vivo, and the absence of ACE shedding from the cell sur- results (Figure 5) and recognition of monkey’s ACE by faces, indicate that this antibody has significant advan- both mAbs (see above) suggest that a mixture of mAbs tages over mAb 9B9 as a potential carrier for drug/gene i2H5 and 9B9 might substantially increase the efficacy of delivery into the lung of primates, including humans. drug/gene delivery to the pulmonary endothelium in non-human primates models, as well as in future human Combined binding of mAbs to different, non-overlapping clinical trials. epitopes of ACE Therefore, selective accumulation of mAb i2H5 and The efficacy of targeted drug/gene delivery depends not 9B9 in primate lung after systemic administration, with- only on the selectivity of targeting, but also on the absol- out any significant shedding of ACE from the cell surface ute amount of drug or therapeutic gene that can be deliv- (for mAb i2H5), suggests that non-human primates can ered to the site or organ of interest. In other words, the serve as an in vivo model for ACE-dependent, anti-ACE- amount of the delivered agent should be therapeutically directed lung targeting and can fill the gap between rat relevant. The results of this study demonstrate that mAb models and future human clinical trials. i2H5 bind to monkey ACE and accumulated in the lung after systemic injection. We have previously demon- Materials and methods strated that radiolabeled mAb 9B9 accumulates selectively (and safely) in the lungs of macaque rrhesus The monoclonal antibodies (mAbs) against human lung (Macaca mulatta)7,8 and human volunteers10 after systemic ACE (9B9, 3A5, i1A8, 3G8, i2H5, 1G12, 6A12, 5F1) used injections. Thus, previous work, as well as the results of in this study were described previously.23 Two new this study, suggest that mAb i2H5 and 5C5 will also mAbs – 7A2 and 5C5 were raised against human kidney accumulate selectively in human lung after systemic ACE by Pocard Ltd (Moscow, Russia). injection. These findings may also allow for improved antibody ACE activity assay redirecting of viral vectors. For example, we recently ach- ACE activity in heparinized plasma (diluted 1:10–1:40 ieved substantial improvements in gene delivery to the with PBS) was determined fluorimetrically with 5 mM rat pulmonary endothelial cells in vivo by using a bispe- Hip-His-Leu36 or 2 mM of Z-Phe-His37 as a substrate. The cific conjugate (Fab-9B9) that redirected adenovirus infec- kidney and lung tissue samples were prepared by homo- tion away from its native receptor, CAR, towards ACE.15 genization in 10 volumes of 50 mM Tris-HCL (pH 7.5) However, although this tropism modification signifi- containing 150 mM NaCl and 0.5% Triton X-100. The cantly increased lung transgene expression (20-fold) and homogenate was centrifuged 10 000 g for 10 min. The reduced liver and spleen transgene expression, residual solubilized ACE remains in the supernatant. transgene expression in these organs remained high in absolute terms, indicating the need for further improve- ACE plate precipitation assay (cross-reactivity of anti- ments. The combination of transductional retargeting ACE mabs) adenoviruses (via ACE) and transcriptional (with the use Ninety-six-well plates were coated with anti-ACE mAbs of an endothelial specific promoter) dramatically via anti-mouse IgG.23 Wells were washed and then incu- improved the selectivity of transgene expression for lung bated with serum/plasma or tissue homogenates

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 288 Animal care protocol Animal procedures were approved by the University of Illinois at Chicago Institutional Animal Care and Use Committees.

Antibody administration and in vivo biodistribution The in vivo biodistribution of anti-ACE antibodies was performed as following. Male stump-tail/Bear macaques (Macaca arctoides) were anesthetized with ketamine (10 mg per kg) followed by pentobarbital injection (35 mg/kg). Subsequently mAbs to ACE (9B9 or i2H5) and control mouse IgG (1 mg/0.5 ml) were injected into cephalic vein. After 2 h, the animals were killed and blood and internal organs (washed in saline) were collected. Kidney and lung tissues were further processed for tissue homogenate preparation as described above. Quantiﬁcation of tissue distribution of non-radiolabeled mAbs to ACE was performed using modiﬁed plate precipitation assay (see above and Ref. 23). The amount of mAbs accumulated in the monkey’s organs was estimated indirectly, by assaying the level of ACE activity, precipitated from tissue homogenates or plasma of studied animals to the plate, covered by anti-mouse IgG. ACE from plasma or solubilized from the membrane of kidney or lung of the studied monkeys, binds to mAbs to ACE, accumulated in this particular organ after intravenous injection, and then quantitatively precipitated by polyclonal antibodies to mouse immunoglobulins, which where previously adsorbed to the microtiter plate.

ELISA Ninety-six-well microtiter plates were coated with 100 ␮l of 1/50 dilution of puriﬁed human ACE (Chemicon, Temecula, CA, USA). mAbs (9B9 or i2H5) bound to the plate were quantiﬁed by goat-anti-mouse IgG, conjugated to alkaline phosphatase.24

Cell ELISA CHO cells, transfected to express ACE-clone 2C224 or human umbilical vein endothelial cells (HUVEC, third passage) cultivated as described previously,35 were grown in 96-well microtiter plates. These cell lines have been demonstrated to be a suitable in vitro model system to study anti-ACE antibody-dependent gene and drug targeting to the pulmonary endothelium.15,24 Cells were washed several times with PBS. Control mouse IgG or Figure 5 The binding of anti-ACE mAbs 9B9 and i2H5 to ACE on the anti-ACE mAbs diluted in PBS/casein (0.2%) were added plastic (a, b) and on the cell surface (c). Binding of mAbs to ACE was to the wells at concentrations of 0.1–10 ␮g/ml and incu- estimated by ELISA. (a) Dilutions of mAbs in 2% dry powder milk in ° PBS were incubated with ACE-coated plate in a volume of incubation (as bated for 1 h 37 C. Plates were washed and cells fixed well as volume of ACE coating) of 50 ␮l. (b) mAb 9B9 (1 ␮g/ml) and with 4% buffered paraformaldehyde. Quantification of mAb i2H5 (3 ␮g/ml) were incubated with ACE-coated plates in 100 ␮l, bound and unbound antibodies was determined by incu- whereas their mixture contains 50 ␮l of each. (c) The CHO-ACE cell line bating the plates with alkaline phosphatase conjugated (clone 2C2) was grown on 96-well plates until confluence. mAb 9B9 (1 goat-anti-mouse IgG.24 ␮g/ml) and mAb i2H5 (3 ␮g/ml), their mixture or control mouse IgG (at the same concentration) were added to the cell in serum-free medium in ACE shedding assay a volume of 100 ␮l and incubated 1 h at 4°C. Anti-ACE mAbs bound to ACE on the plastic (a, b) or on the surface of CHO-ACE cells (c) were CHO-ACE cells (clone 2C2) or HUVEC, third passage, revealed similarly with goat-anti-mouse IgG, conjugated with alkaline were grown in 96-wells plate until confluence. Then cells phosphatase.24 The data represent the results of at least three independent were washed and incubated with mAbs 9B9, i2H5, 5C5 experiments and expressed as mean ± s.d. or control mouse IgG (10 ug/ml in serum-free culture medium) for 4 h at 37°C. After that period, culture medium was harvested for ACE activity measurement. Cells, remaining on the plate after harvesting culture (sources of ACE) diluted in PBS/BSA, washed and fluid were lysed with a detergent (CHAPS). ACE activity assayed for plate-bound ACE activity using Hip-His-Leu on the cell surface (lysates), as well as in the culture as a substrate.23 medium (released) was determined by the fluorimetric

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 289 assay (see above) with Hip-His-Leu as a substrate. To enzyme (ACE) with antigen in vitro and in vivo: antibody tar- determine correctly the effect of anti-ACE mAbs on the geting to the lung induces ACE antigenic modulation. Int Immu- rate of ACE shedding (by referring ACE level in culture nol 1994; 6: 1153–1160. medium to that on the cell surface), we performed the 12 Muzykantov VR et al. Immunotargeting of anti-oxidant enzymes following calculations. ACE activity which appeared in to the pulmonary endothelium. Proc Natl Acad Sci USA 1996; 93: 5213–5218. the culture medium of the cells incubated with non- 13 Muzykantov VR et al. Targeting of antibody-conjugated plas- immune mouse IgG (basal, constitutive ACE shedding) minogen activators to the pulmonary vasculature. J Pharm Exp was subtracted from that for the cells incubated with Ther 1996; 279: 1026–1034. anti-ACE antibodies. The result of subtraction reflects 14 Atochina EN et al. Immunotargeting of catalase to endothelial antibody-induced appearance of ACE activity in the cul- surface antigens ACE or ICAM-1 in rat lung protects the pul- ture fluid, ie antibody-induced ACE shedding. monary vasculature against oxidative stress. Am J Physiol (Lung Neither mAb 9B9 nor mAb i2H5, which are directed Cell Mol Physiol) 1998; 275: L806–L817. to the N-terminal domain of ACE, inhibited ACE activity 15 Reynolds PN et al. A targetable injectable adenoviral vector for with substrate (Hip-His-Leu) which we used and which selective gene delivery to pulmonary endothelium in vivo. Mol hydrolyzed mainly by C-terminal domain of ACE.23 Ther 2000; 2: 562–578. Therefore, mAbs themselves did not affect the shed- 16 Reynolds PN et al. Combined transductional and transcriptional ding assay. targeting improves the specificity of transgene expression in vivo. Nature Biotech 2001; 19: 838–842. 17 Donahue RE, Dunbar CE. Update on the use of non-human pri- Acknowledgements mate models for preclinical testing of gene therapy approaches targeting hematopoetic cells. Hum Gene Ther 2001; 12: 607–617. The authors thank Professor Ronald F Albrecht, Head of 18 Chan AWS et al. Transgenic monkeys produced by retroviral the Department of Anesthesiology, University of Illinois gene transfer into mature oocytes. Science 2001; 291: 309–312. at Chicago, for his continuous encouragement and sup- 19 Danilov SM et al. Immunohistochemical study of angiotensin- port of this project. We acknowledge the technical assist- converting enzyme in human tissues using monoclonal anti- ance of David J Visintine. We also thank Dr T Hewett bodies. Histochemistry 1987; 87: 487–490. (University of Illinois at Chicago), Dr TJ Rowell 20 Fishman A. Dynamics of the pulmonary circulation. In: (University of Louisiana at Lafayette) and Dr B Lapin Hamilton WF, Dow P (eds). Handbook of Physiology Circulation, (Institute of Medical Primatology, Sochi, Russia) for pro- vol. 2. Am Physiol Soc: Washington DC, 1963, pp 1667–1774. viding plasma and tissue samples of the monkeys men- 21 Crapo JD et al. Cell numbers and cell characteristics of the nor- tioned in this paper. This study was supported in part mal human lung. Am Rev Respir Dis 1982; 125: 332–337. 22 Panes J et al. Regional differences in constitutive and induced by grants DA08256 and DA12672-NIDA-NIH (to DCY). ICAM-1 expression in vivo. Am J Physiol 1995; 269: H1955– Parts of this work were presented at the Annual Meetings H1964. of American Thoracic Society (ATS), San Diego, 1999, and 23 Danilov S et al. Structure–function analysis of angiotensin I-con- San Francisco, 2001. verting enzyme using monoclonal antibodies. J Biol Chem 1994; 269: 26806–26814. References 24 Balyasnikova IV et al. Antibody-mediated lung endothelium targeting: in vitro model of lung endothelium targeting using a cell 1 Soffer RL. Angiotensin-converting enzyme and the regulation line expressing angiotensin-converting enzyme. Tumor Targeting of vasoactive peptides. Ann Rev Biochem 1976; 45:73–94. 1999; 4:70–83. 2 Corvol P, Michaud A, Soubrier F, Williams TA. Recent advances 25 Murciano JC et al. Vascular immunotargeting to endothelial sur- in knowledge of the structure and function of the angiotensin I face in a specific macrodomain in alveolar capillaries. Am J Resp converting enzyme. J Hypertens 1995; 13: (Suppl. 3) S3–10. Crit Care Med 2001; 164: 1295–1302. 3 Caldwell PR et al. Angiotensin-converting enzyme: vascular 26 Stevens J, Danilov SM, Fanburg BL, Lanzillo J. A sensitive two- endothelial localization. Science 1976; 191: 1050–1051. site sandwich enzyme immunoassay for human angiotensin- 4 Ryan US, Ryan JW, Whitaker C, Chiu A. Localization of angiot- converting enzyme utilizing monoclonal antibodies. J Immunol ensin converting enzyme (kininase II). II. Immunocytochemistry Meth 1990; 132: 263–273. and immunofluorescence. Tissue Cell 1976; 8: 125–145. 27 Muzykantov VR. Immunotargeting of drugs to the pulmonary 5 Franke FE. Angiotensin-I-converting enzyme (CD143) on endo- vascular endothelium as a therapeutic strategy. Pathophysiology thelial cells in normal and in pathological conditions. In: Kishi- 1998; 5:15–33. moto T (ed.). Leukocyte Typing VI. New York: Garland Pub- 28 Muzykantov VR et al. Streptavidin facilitates internalization and lishing, 1997, pp 749–751. pulmonary targeting of an anti-endothelial cell antibody 6 Danilov SM et al. Lung uptake of antibodies to endothelial anti- (platelet-endothelial cell adhesion molecule. (1) A strategy for gens: key determinants of vascular immunotargeting. Am J Phy- vascular immunotargeting of drugs. Proc Natl Acad Sci USA siol (Lung Cell Mol Physiol) 2001; 280: L1335–L1347. 96 7 Danilov SM et al. Radioimmunoimaging of lung vessels: an 1999; : 2379–2384. approach using Indium-111-labeled monoclonal antibody to 29 Muzykantov VR et al. Epitope-dependent selective targeting of angiotensin-converting enzyme. J Nucl Med 1989; 30: 1686–1692. thrombomodulin monoclonal antibodies to either surface or 8 Danilov SM et al. Lung is the target organ for the monoclonal intracellular compartment of endothelial cells. Drug Deliv 1998; antibodies to angiotensin-converting enzyme. Lab Invest 1991; 5: 197–205. 64: 118–124. 30 Muzykantov VR et al. Endothelial cells internalize monoclonal 9 Muzykantov VR et al. Immuno-targeting of streptavidin to the antibody to angiotensin-converting enzyme. Am J Physiol (Lung pulmonary endothelium. J Nucl Med 1994; 35: 1358–1365. Cell Mol Physiol) 1998; 270: L704–L713. 10 Muzykantov VR, Danilov SM. Targeting of radiolabelled mono- 31 Hooper NM, Karran EH, Turner AJ. Membrane protein secret- clonal antibody against angiotensin-converting enzyme to the ases. Biochem J 1997; 321: 265–279. pulmonary vasculature. In: Torchilin VP (ed.). Handbook of Tar- 32 Andres G et al. Cellular mechanisms of adaptation of grafts to geting Delivery of Imaging Agents. CRC: Boca Raton, FL, 1995, pp antibody. Transplant Immunol 1996; 4:1–17. 465–486. 33 Woolkalis MJ et al. Regulation of thrombin receptors on human 11 Danilov S et al. Interaction of mAb to angiotensin-converting umbilical vein endothelial cells. J Biol Chem 1995; 270: 9868–9875.

Gene Therapy Lung endothelium targeting in primates IV Balyasnikova et al 290 34 Chu M, Bird CH, Teasdale M, Bird PI. Turnover of thrombomo- 36 Friedland J, Silverstein E. A sensitive ﬂuorometric assay for dulin at the cell surface occurs at a similar rate to receptors that serum angiotensin-converting enzyme. Am J Pathol 1976; 66: are not actively intermalized. Thromb Haemost 1998; 90: 119–127. 416–424. 35 Balyasnikova IV, Danilov SM, Muzykantov VR, Fisher AB. 37 Piquilloud Y, Reinharz A, Roth M. Studies on the angiotensin Modulation of angiotensin-converting enzyme in cultured converting enzyme with different substrates. Biochim Biophys human vascular endothelial cells. In Vitro Cell Dev Biol 1998; 34: Acta 1970; 206: 136–142. 545–554.

Gene Therapy