Cancer Immunol Immunother (1991 ) 33: 367 - 374 034070049100078Q ~ ancer mmunology mmunotherapy © Springer-Verlag 1991 Adriamycin(hydrazone)-antibody conjugates require internalization and intracellular acid hydrolysis for antitumor activity Gary R. Braslawsky 1 , Kathleen Kadowl, Jay Knipe~, Kerry McGoff 1, Mary Edson 1, Takushi Kaneko 2, and Robert S. Greenfield I 1 Bristol-Myers Squibb, 5 Research Parkway, E O. Box 5100, Wallingford, CT 06492-7600, USA 2 Pfizer Inc., Central Chemistry, Groton, CT 06340, USA Received 16 April 1991/Accepted 22 May 1991 Summary. Adriamycin hydrazone (ADM-Hzn) inmmno- conjugates have previously been shown to exhibit anti- body-directed antitumor acUvity in vitro and in vivo. In this report, the biological and biocherrlical properties of the mAb and linker were investigated. Conjugates prepared with two antibodies 5E9 [anti-(transferrin receptor)] and G28.1 (anti-CD37), (which internalize from the surface of target cells following binding) were more cytotoxic in vitro and had greater antitumor activity against Daudi B lym- phoma tumor xenografts than a non-internalizing immuno- conjugate prepared witb mAb 2H7 (anti-CD20). In addi- tion, the 13-acylhydrazone bond linking the drug to the mAb was labile at pH 5 and released unmodified ADM at a rapid rate (tl/2 -- 2.5 h). Immunoconjugates prepared with an oxime linkage at the C-13 position were stable to acid and were not cytotoxic. These findings suggest that internalization of ADM-Hzn immunoconjugates and re- lease of free ADM from the mAb in acidic intracellular compartments were important steps in the mechanism of action of ADM-Hzn immunoconjugates. Key words: Monoclonal antibodies - Immunoconjugates - Adriamycin Introduetion To be effective, a cytotoxic molecule must be transported from the extracellular space across the plasma membrane and into the nucleus or cytoplasm of the target cell. For a drug to be clinically useful, it must therefore täke advan- tage of biochemical or uptake differences between the neo- plastic and normal cell populations. Yet few exploitable differences have been found. The current interest in mAb- based serotherapies for the treatment of malignant disease is based on their potential to discriminate neoplastic cell populations from normal tissues. Major therapeutic ap- Offprint requests to: G. R. Braslawsky proaches using mAbs are (a) the use of unconjugated mAb to activate or enhance endogenous defense mechanisms following antigen binding [13, 18], and (b) their use as carriers of radioisotopes [17, 21], toxins [19, 22-24], or chemotherapeutic drugs [7, 9, 11, 12, 32]. mAbs that bind to cell-surface receptors can enter the cell by a process similar to receptor-mediated endocytosis [10, 25, 26]. mAbs can also be internalized during mem- brane utilization or membrane biosynthesis [27]. Presum- ably, mAb-drug immunoconjugates would enter the cell via this same type of uptake mechanism. Receptor-medi- ated internalization of immunotoxins has been shown to be a necessary requirement for mAb-directed cytotoxicity [6, 20, 22, 29]. Antitumor activity may be associated with the ability of the mAb to substitute for the B (binding) chain of the toxin, and mimic B chain function for both binding and intracellular delivery of the A (toxic) chain. Those parame- ters that influenced internalization, such as the ability of the antibody to cross-link the receptor [24], the rate of intracellular delivery [23], antigen density [23], and the amount of surface antibody internalized [20], all affected cytotoxic efficiency. The requirement for drug-antibody conjugates to be internalized for biological activity remains unclear. Evi- dence has been reported that methotrexate-mAb conjugates were processes in lysosomes [8, 9], and it has been sug- gested that delivery of cis-aconityl-linked anthracycline into acidic intracellular compartments via endocytic processes [7, 32] mediated cell tdlling. In contrast, mAb- Vinca-alkaloid conjugates may not require internalization for biological activity [26]. Recently our laboratories have described the synthesis of an Adriamycin (ADM) immuno- conjugate in which the drug was attached to the antibody via an acid-sensitive 13-acylhydrazone (Hzn) linker [12]. Conjugates retained mAb-binding activity following drug conjugation and showed antibody-dependent (immu- nospecific) cytotoxicity in vitro [1, 12]. More important, when tested in human tumor xenograft models, ADM-Hzn conjugates demonstrated greater inhibition of tumor growth, increased potency and less systemic toxicity than free ADM [5, 12a].
`
`IMMUNOGEN 2090, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`368 O o où ,ù,-ùù'%'s's'~-"l~ ~ CH2OH CH30 O OH O CH3 NH~20 HO 1 o OH "'°~/~/~s'S'~N~ ~ CH2OH 02 N ~ß~'~j CHaO O OH O CH3 NH~20 HO 3 O O N. N H~./'~ S _ S."%'~ N_ MAb H2OH CH30 0 OH 0 CH3 NH~2 HO O H N .O~ S- S/'''/~ N - MAb CH2OH CH~O O OH O CH3 N~HO2 HO 2 4 Fig. 1. Chemical structure of Adriamycin hydrazone (ADM-Hzn) (1) and its immunoconjugate (2). Chemical smacture of ADM oxime (3) and its immunoconjugate (4) The method of synthesising ADM-Hzn conjugates (condensation of ADM-Hzn with thiolated mAb) leads to the formation of a disulfide bond in the middle of the linker arm connecting mAb to drug. Reduction of the conjugate resulted in release from the protein of a 13-acylhydrazone thiol derivative of ADM while acid hydrolysis led to re- lease of unmodified ADM [12]. The importance of either of these reactions for cytolytic activity has not been fully investigated. We speculated that internalization of conju- gates into acidic intracellular compartments resulted in hydrolysis and intracellular release of active drug from the mAb protein. However, it has been suggested that unconju- gated ADM does not necessarily have to enter the cell to have cytolytic activity [28], which may make it possible for anthracycline immunoconjugates to mediate activity through membrane binding only. In the present study, we examined the intemalization of ADM-Hzn immunoconjugates prepared with different mAbs. Conjugates that bound to cell-surface antigens that were internalized exhibited better antitumor activity in vi- tro and in vivo than non-internalizing conjugates. In addi- tion, evidence is presented that shows that acid hydrolysis of the 13-acylhydrazone bond was also required for cyto- lytic activity. The proposed mechanism of action of killing mediated by internalization of mAb-ADM-Hzn conjugates is discussed. Materials and methods Monoclonal antibodies, mAb G28.1 (IgG1, anti-CD37 B cell antigen from J. Ledbetter, Oncogen Division) and mAb L6 [IgG2ä, anti-(lung carcinoma), from I. Hellstrom, Oncogen] were produced and purified from tissue-culture supernatants by InVitron Corp. (St. Louis, Mo.). mAb 5E9 [IgG1, anti-(human transferrin receptor), ATCC no. HB21], mab 3A1 (IgG, pan T-cell antigen, ATCC no. HB2) and mAb 2H7 (IgG1, anti-CD20 B cell antigen, J. Ledbetter) were purified from ascites from BALB/c mice (Charles River, Inc., Ma.). Cell lines. Daudi and Raji Burkitt's lymphoma cell lines were obtained from the ATCC. Cells were maintained in eulture in RPMI- 1640 medium containing 10% fetal calf serum, penicillin (100 U/ml) and streptomycin (100 p.g/ml). Epitope density on Daudi cells was previously determined by Scatchard analysis using radiolabeled antibodies [11]. The Daudi cell line bound 2.8x 105 molecules/cell of mAb 2H7, and 1.2x 105 molecules/cell of mAb 5E9. mAb L6 and mAb 3A1 did not bind to the Daudi or Raji cell lines. Preparation of ADM-Hzn immunoconjugates. ADM-Hzn immunoconju- gates were prepared by attaching ADM Hzn, a 13-acylhydrazone deriva- tive of ADM, to N-succinimidyl 3-(pyridyldithio) propionate (SPDP)- thiolated rnAbs. The derivative, which contains a pyridyl-protected disul- fide attached to the C-13 position of ADM via a hydrazone linkage (Fig. 1), was prepared as described previously [12]. Briefly, mAbs were reacted with SPDP (Pierce Chemical), dialyzed against phospate- buffered saline (PBS), and reduced with dithiothreitol (Sigma) to gener- ate free thiol groups. Thiolated mAbs were reacted overnight at 4 ° C with ADM-Hzn and centrifuged at 5000 g to remove any precipitate formed during the coupling procedure. To remove unbound drng, immunoconju- gates were exhaustively dialyzed against PBS, and passed over SM2 Biobead columns (BioRad Dir., Richmond, Calif.). Conjugates used in thJs study contained less than 5% free drug as determined by HPLC analysis, and retained more than 95% of the original binding activity [12]. Preparation of ADM-oxime immunoconjugates. An ADM derivative having an oxime linkage at the C-13 position of the ADM molecule was prepared as described by Kaneko and Webb [16]. Briefly, ADM-HC1 was condensed with 1-aminooxy-4-[(3-nitropyridin-2-yl)dithio]butane-
`
`IMMUNOGEN 2090, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`369 HC1 to form the ADM derivative 13-[4-(3-nitropyridin-2-yl)- dithio]butyloximino-Adriamycin-HC1 (ADM-oxime, Fig. 1). Immuno- conjugates were prepared by reacting ADM-oxime with SPDP-thiolated mAbs and purified as described above. High-pressure liquid chromatography (HPLC). The release of ADM under acidic conditions from either the hydrazone or oxime derivatives, as well as from the immunoconjugates was monitored and quantified by HPLC analysis. The derivatives were incubated at a concentration of 2gM in 0.1 M sodium acetate buffer, pH 5, at 37°C. At various times aliquots were removed and chromatographed on a Waters HPLC system (model 510 pumps and a 680 gradient controller) using a g-Bondapak C-18 10-gm column (3.8 x 300 mm). The mobile phase for elution of ADM-Hzn consisted of 68% triethylammonium formate buffer (0.05 M, pH 2.8) and 32% acetonitrile at a flow rate of 1.2 ml/min. ADM-oxime was chromatographed using a linear gradient from 68% buffer/32% acetonitrile to 30% buffer/70% acetonitrile over 13 min, after a hold for 7 min, at the initial conditions (flow rate 1.4 tal/min). Column effluents were monitored with an ABI mode1980 fluorescence detector (excitation wavelength at 240 nm and emission wavelength of 550 nm). Data acqui- sition and integration were performed using Turbochrome software (PE Nelson) running on an IBM PS/2 model 70 computer. Colony-formation assay. Cytotoxicity was determined by inhibition of colony formation in soft agar as previously described [12]. Daudi cells (1 x 105 cells/ml) were incubated for 2 h at 37 ° C with immunoconjugate or drug diluted in culture medium. Cells were washed and suspended at 5 x 103 cells/ml in RPMl-1640 medium containing 15% fetal bovine serum and 0.3% agarose (Marine Colloid). Suspensions of 1 ml were overlayed onto 0.4% agarose in six-well microtiter plates. Samples were incubated for 7-10 days at 37 ° C at 5% CO2. Colonies were visualized by staining for 48 h at room temperature with 0.5 tal, 1 mg/ml p-io- donitrotetrazolium violet (Sigma). Colonies were counted using an Opti- max 40-10 image analyzer. 6° 1 oc 30 c Œ o r- 03 o Pc o =~ 20 ŒE 0o 0 ; ;0 1'~ 2'0 Time (Hours at 37 °) Fig. 2. Flow-cytometric analysis of internalization rates for mAb 5E9 (•) or 5E9-ADM-Hzn (©) conjugates. Cells were incubated with suffi- cient antibody protein (20 #g/ml) to saturate all available antigenic sites on Daudi cells. After 1 h incubation at 4 ° C, cells were removed from the cold, washed, and warmed to 37 ° C. At various times after warming, cells were removed and stained by indirect immunofluorescence for cell-sur- face antibody. Non-specific immunofluorescence was determined by in- cubation with non-binding mAb L6 before staining. Non-specific fluorescence (mean channel number) was subtracted from the mean channel number obtained using mAb 5E9 or conjugate Tumor models. Daudi solid tumor xenografts were serially passaged in 4- to 6-week-old BALB/c nu/nu mice (Harlan-Sprague Dawley) using 2 × 107 tumor ceils implanted s.c. Tumors showed linear growth rates between 200 mm 3 and 4000 mm 3. The median tumor volume doubling time during exponential growth was 6.9+0.8 days. When tumor volumes reached 400-600 mm 3, mice were randomized into treatment groups of eight to ten animals each. Tumor volumes (10 were calculated using the formula: L×W 2 V= -- 2 [where L = length (mm), W = width (mm)]. Modulation experiments - fluorescence assay. Antigenic modulation induced by mAb or mAb-ADM conjugate was assayed by indirect im- munofluorescence using the Daudi B cell lymphoma cell line. Target cells (1 × 106 cells/ml) used to determine mAb-induced modulation were pulsed-labeled with the appropriate mAb (40-100 gg/tal) for 1 h at 4 ° C. Non-modulated cells were incubated under identical conditions with a similar concentration of a non-binding mAb (e. g. mAb L6). Cells were washed free from all unbound mAb, resuspended in culture medium and, at various times after incubation at 37 ° C, stained for residual cell-surface antibody by indirect immunofluorescence as described below. In certain experiments (Fig. 3), target cells were continuously exposed to mAb for 16 h at 37 ° C before being stained for residual cell-surface antibody. mAb bound to the target cell population was determined using in- direct immunofluorescence and quantified on an Epics V model 753 Flow Cytometer (Coulter Corp.). Cells, at various times during modula- tion at 37 ° C, were quickly cooled to 4 ° C, washed in ice-cold RPMI medium containing 2% fetal calf serum and incubated with 1 : 40 diluted fluorescein-isothiocyanate-conjugated goat F(ab)'2 anti-(mouse IgG) (Boehringer) for 30 min at 4 ° C. Cells were washed and the amount of cell-surface fluorescence quantified by flow cytotometry. Background binding levels were deterrnined using a non-modulated cell population (e. g. those cells incubated with mAb L6) and stained as described above. Modulation experiments-isotope assay. The amount of modulating anti- body incorporated into the cell population during mAb modulation was determined using 125I-labeled mAb. mAbs were radioiodinated with Na~25I (New England Nuclear, Ma.) using chloramine T. mAb-bound iodine was separated from free iodine by PD- 10 column chromatography (Pharmacia Fine Chem.). Specific activities of the labeled preparations ranged from 2 x 105 to 8 x 10 » cpm/gg protein. Total antibody incorporated during modulation was determined by incubating Daudi cells (1 × 106 cells ml) with 40 gg/tal ~25I-labelled mAb (triplicates) in RPMI culture medium for 16 h at 37 ° C. Cell-bound radioactivity was separated from unbound radioactivity by overlaying the cell population onto 0.15 ml ice-cold 1 : 1 mixture of n-butyl phthalate (Fisher) and dinonyl phthalate (Kodak Chem.). Samples were then cen- trifuged in a Fisher model 235B microcentrifuge at 4 ° C for 1 min. Ra- dioactivity in the pelleted cells was determined using a LKB model 1272 gamma counter. Non-specific binding was determined for each dilution by incubation with radioiodinated L6 antibody, and results (cpm) were subtracted from the values with the corresponding antibody. Results Modulation and internalization of mAb-ADM-Hzn conjugates The ability of mAb or mAb-ADM-Hzn conjugates to in- duce antigenic modulation was demonstrated by staining for cell-surface antibody using indirect immunofluores- cence. Daudi cells were incubated at 4°C with saturating concentrations of either mAb 5E9 or 5E9-ADM-Hzn con- jugate. As shown in Fig. 2, warming the cells to 37°C resulted in a rapid loss of 5E9 antibody from the cell surface. The rate of disappearance of mAb from the cell
`
`IMMUNOGEN 2090, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`370 5E9 ~~.~z_.,& '~e } O e) => fr" 2H7 Pre Post I ,z,, 0 3 6 9 12 15 18 21 24 27 Fluorescence intensity Fig. 3. Flow-cytometric analysis of Daudi lymphoma teils incubated with either mAb 5E9 (upper panel) or 2H7 (lower panel) before (pre- modulation) and after (post-modulation) incubation at 37 ° C for 16 h Table 1. Binding of 125[-labeled mAb to Daudi lymphoma cells before (premodulated) and after incubation at 37 ° C for 16 h (modulated) mAb Cell-bound radioactivity (cpm) Premodulated a Modulated b 5E9 9 876 12174 2H7 o 20 569 24 510 L6 841 745 a Cells were incubated with lzSI-labeled mAb (40 I-tg/ml) for 1 h at 4 ° C and washed free of unreacted mAb; cell-bound radioactivity was deter- mined as described in Materials and methods b Cells were continuously exposed to 125I-labeled mAb for 16 h at 37 ° C; cell-bound radioactivity was determined as described in Materials and methods c Results for mAb 2H7 were obtained separately from those for 5E9 and L6 2OO0 1500 E g lo00 ò E 500 Binding of Immunoconjugates o 60 • 40 0 10 20 0 20 3'o 4'o ~'o Days post implant Fig. 4. Antitumor activity of DaudJ tumor xenografts after inoculation of either internalizing, 5E9-ADM-Hzn or non-internalizing 2H7-ADM- Hzn, immunoconjugates. Mice (eight per group) were inoculated (i. p.) on days 21 and 26 after implant. 2H7-ADM conjugate (•) was inocu- lated at 600 mg/kg protein and 3.4 mg/kg drug. The 5E9 conjugate was inoculated at either a matching antibody dose (0): 600 mg/kg protein and 9 mg/kg ding, or at a matching drug dose ( • ): 250 mg/kg protein and 4 mg/kg drug. Control mice (O) were untreated. Inset: antibody titration curves of the 2H7-ADM-Hzn (•) or 5E9-ADM-Hzn (O) con- jugate on Daudi cells surface was estimated to be approximately 2%-8%/min. In order to rule out the possibility that the loss of cell-sur- face antibody was not due to capping and subsequent sloughing of antibody from the cell surface, Daudi cells were incubated with saturadng amounts of 125I-labeled 5E9, and cell-bound radioactivity (Table 1) as well as cell- surface fluorescence (Fig. 3) was determined after incuba- tion at 37 ° C. Essentially all of the radiolabeled antibody remained associated with cells, although cell-surface anti- body could not be detected by immunofluorescence. Similar results were obtained when saturating amounts of mAb 5E9-ADM-Hzn conjugates were used instead of mAb 5E9 (Fig. 2). The kinetics of antibody internalization was essentially the same for conjugated and unconjugated 5E9 mAb on the Daudi cell line. Not all mAbs were inter- nalized following binding to a cell-surface antigen. This is illustrated in Fig. 3 and Table 1, where mAbs 2H7 and 5E9 (40 pg/ml) were incubated with Daudi cells and stained for cell-surface fluorescence before (premodulated) and after (post-modulated) incubation at 37°C for 16 h. In these experiments no loss of cell-surface antibody nor cell- bound radioactivity was observed when antigenic sites on the Daudi cell line were saturated using mAb 2H7. In contrast to mAb 2H7, mAb 5E9 showed a significant loss of cell-surface antibody (Fig. 3) without a comparable loss of cell-associated antibody (Table 1). These findings indi- cate that almost all of the membrane-associated 5E9 anti-
`
`IMMUNOGEN 2090, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`E 7o] 6~ 3~ 2~ o] 70 6~ J J 3~ 2~ I0- o] ADM-HZN l!llllll]IIlllil''ll~~J~lllmlH''llll{l" l'lll fl~T[l' HII' U ll'l~llmllm]]lil]'1]illlL~[~ ADM , 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Retention Time (minutes) 371 Fig. 5. Chromatogram of the ADM-Hzn before (top) and after (bottom) incubation for 4 h at 37 ° C, pH 5 sodium acetate buffer. Positions of ADM-Hzn and ADM are indicated on the chromatograms Table 2. Cytotoxicity of Adriamycin hydrazone (ADM-Hzn) immuno- conjugates in vitro ~ Dose 2H7-ADM-Hzn 5E9-ADM-Hzn 3A1 ADM-Hzn (~tg/ml) Col./well b I c (%) Col./well I (%) Col./well I (%) 6 64 __+ 11 60 37 ± 8 75 84_+ 2 43 3 110± 2 25 85_+ 0.8 42 114_+13 22 1.5 149± 3.5 0 127±13 14 149_+11 0 0.7 154+27 -5.4 131_+ 8 11 155_+ 9 -6 a Immunoconjugates were incubated with Daudi cells for 1.6 h at 37 ° C prior to cloning in soft agar u Colonies per well c Inhibition (%) = [l-(treated/control)] x 100 body was internalized by the cell population, whereas the 2H7 antibody remained membrane-associated. In vitro and in vivo antitumor activity The necessity of internalization of ADM-Hzn immuno- conjugates for cytotoxic or antitumor acitivity was investigated using conjugates prepared with either the internalizing 5E9 or non-internalizing 2H7 mAbs. The im- munoconjugates bind to different antigens on the Daudi cell line, and there were approximately 2.5 times as many 2H7 binding sites as 5E9 binding sites (see insert, Fig. 4). In vitro cytotoxic activity was investigated by assaying the inhibition of colony formation. The 3A1-ADM-Hzn conju- gate was used as a non-binding control. As shown in Table 2, cytotoxic activity of the 2H7 conjugate was com- parable to that obtained with the non-binding 3A1 im- munoconjugate. In contrast, 5E9-ADM-Hzn showed in- creased cytotoxic activity at each dose level. While differ- ences in colony inhibition were relatively small, this find- ing was consistent over the course of many experiments. Fig. 4 compares antitumor activity of the non-internal- izing (2H7) and internalizing (5E9) ADM-Hzn immuno- conjugates on progressively growing Daudi tumor xeno- grafts. Tumor growth inhibition mediated by the 5E9 conjugate was compared to that with the 2H7 conjugate at both matching drug (4 mg/kg) and at matching antibody (600 mg/kg) doses. Antitumor activity was significantly greater with the 5E9 conjugate at both conjugate doses, as compared to the non-modulating 2H7-ADM conjugate. Hydrolysis of ADM-Hzn and ADM-oxime derivatives mAb-ADM-Hzn conjugates were found to be stable at pH />7.4, but rapidly released ADM from the mAb at pH ~<6.5 [12]. As shown by the HPLC chromatogram in Fig. 5, ADM was the only product detected after acid hy- drolysis of ADM-Hzn following a 4-h incubation at 37 ° C, pH 5. Comparison of the release of ADM from ADM-Hzn and ADM-oxime as a function of time at 37 ° C, pH 5, is shown in Fig. 6. The hydrazone derivative was rapidly converted to ADM (tl/2 = 2.5 h), while release of ADM from the ADM-oxime derivative was not detected at 6 h or even over a 24-h observation period. Immunoconjugates prepared from these derivadves demonstrated similar rates of hydrolysis (data not shown). Cytotoxic activity of ADM-oxime-mAb conjugates ADM-Hzn and ADM-oxime were linked to mAb G28.1, and both conjugates were shown to be internalized after binding to cell-surface antigens (Fig. 7). When tested for cytolytic activity, only the acid-cleavable ADM-Hzn con-
`
`IMMUNOGEN 2090, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`372 1800 1600 " "~ 1400 120o cz 1000 " c 800 " õ B 600 " B 400 " 200 " 0 A f B -- . • . _- _- . . 1 ~ ; 4 a Time (h) Fig. 6. Appearance of ADM from either ADM-Hzn derivative (.4, []) or ADM-oxime derivative (B, [] ) as a function of time after incubation at 37 ° C, pH 5 sodium acetate buffer jugate was cytotoxic (Fig. 8). In vitro cell killing by G28.1- ADM-oxime was similar to background levels determined by using non-binding L6-ADM-Hzn and L6-ADM-oxime conjugates. Thus, both the oxime- and hydrazone-attached G28.1-ADM conjugates were internalized, only the acid- sensitive hydrazone-linked immunoconjugate was capable of mediating cytotoxic activity in vitro. Discussion The concept of mAb-mediated drug delivery requires not only an understanding of how an antibody delivers the drug to the tumor cell, but how the drug is activated and able to reach the appropriate intracellular target sites. Previous studies showed that ADM-Hzn immunoconjugates were 100- Daudi cells phenotype : / c 80 G28,1 + / o L6 - o 6O >, t- O 4o "Õ c o 20 .... 4m" ,A 0 le-9 le-8 le-7 le-6 ADM concentration (molar) Fig. 8. Cytotoxic activity of ADM immunoconjugates prepared with the acid-sensitive (hydrazone) and resistant (oxime) linker. Daudi cells were incubated with two-fold serially diluted G28.1-ADM-Hzn ( • ), G28.1- ADM-oxime (O), L6-ADM-Hzn ([]) and L6-ADM-oxime (•) conju- gates. After 2 h exposure at 37 ° C, cells were washed and cytotoxic activity was determined using a colony-formation assay in soft agar highly effective antitumor agents and demonstrated immu- nospecific cell killing using both in vitro and in vivo mod- els [4, 1, 5, 12]. In this report, we showed that the cytolytic activity of ADM-Hzn immunoconjugate relied upon the appropriate combination of mAb, linker, and drug. The linker used to attach ADM to the mAb contained both an acid-labile acylhydrazone linkage at the C-13 position of the molecule, and a reducible disulfide bridge in the linker arm. By comparing conjugates prepared with a 13-hydra- zone-linked drug to a 13-oxime-linked drug, we demon- strated that hydrolysis of the drug at the C-13 position and release of free ADM was required for cytolytic activity. Since both conjugates contained the disulfide bond in the LU m Z _J ùJ uJ (.3 W > H F-- < ..J LU rr t G2O. | (HYD) AOH l [ ~ G28. ! (OXIHE) ADH t I ,t~;/Po.. ù-,l, / =/POST COHJ. I POST ~AB ',!( ,I :~i[ù s' ù~,', '/i!;~r,,ii!b., , ~. , 3 6" ~ " J. 2 ** J. 5 ~~: 21 2:~ ~ 3 6 ~ I. IL Iß FLUORESCENCE Fig. 7. Modulation of G28.1-ADM-Hzn (left partei) and G28.1-ADM- Oxime (right panel) from cell-surface membrane, Raji cells were incu- bated with either G28.1 mAb or immunoconjugate for 16 h at 37 ° C, and INTENSITY cell-surface antibody was determined b E flow cytometry. PRE, the amount of cell-surface antibody before internalization
`
`IMMUNOGEN 2090, pg. 6
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`373 linker arm, these findings also suggest that the disulfide bond in the middle of the linker arm may not play a major role in the activity of the conjugates. This has been sup- ported by other studies in which conjugates prepared with a thioether linkage instead of the disulfide showed similar in vitro cytolytic activity (European Patent Appl. no. 0328 157, 1989). The biology of the mAb (i. e. its ability to deliver the conjugated drug to an acidic compartment) also played a major role in the activity of ADM-Hzn immunoconjugates. A correlation between conjugate internalization and cyto- lytic or antitumor activity was demonstrated using internal- izing and non-internalizing immunoconjugates, mAb 5E9, which binds to human transferrin receptor, was rapidly internalized into target cells following binding. These find- ings agree with other reports that show rapid internaliza- tion of transferrin receptor following binding of mAb or transferrin [14]. Conjugation of ADM-Hzn to 5E9 did not alter the internalization kinetics of the antibody. Similar results were obtained using mAb G28.1, which binds to the CD37 surface antigen on B cells. The requirement of antibody-induced internalization was demonstrated by comparing antitumor acitivity of ADM-Hzn conjugates attached to internalizing mAbs 5E9 or G28.1 with that obtained using mAb 2H7, which recog- nizes the CD20 antigen on B cells. In contrast to the mAbs against transferrin receptor and CD37, mAb 2H7 did not internalize following binding. In vitro cytotoxicity of ADM-Hzn conjugated to mAbs 5E9 and 2H7 was com- pared using colony-formation assays. At all concentrations tested, the cytotoxicity obtained using a 5E9-ADM conju- gate was greater than that of 2H7-ADM. Conjugates pre- pared with mAb G28.1 also showed similar cytotoxicity to the 5E9-ADM conjugate in an in vitro assay. It has been difficult to demonstrate antibody-directed cytotoxicity of drug mAb conjugates in vitro often because of high non-specific cytotoxic activity, especially at high input doses. This was also evident in this study in which the cytotoxic activity of 5E9 to 2H7-ADM conjugates was compared to that of a non-binding immunoconjugate (Table 2). Studies have shown that the length of exposure of the drug-rnAb conjugate to the target cell can influence both non-specific and specific cell killing by the conjugate. Non-specific cell killing can often be associated with up- take of free drug, either as a contaminant in the preparation or released from the conjugate during exposure. Non-spe- cific cytotoxicity may also be due to alternative mecha- nisms by which antibody can be endocytosed, such as through membrane utilization [27] or receptors on the tumor cell surface membrane (FcR) that bind to the Fc region of the antibody molecules [31]. FcR have been found on a variety of tumors, including those of non-lym- phoid origin [2-4]. The ability of human IgG-melphalan conjugates to internalize and kill FcR-bearing human B-lymphoid tumors in vitro may have been associated with this delivery mechanism [20]. These findings have made it more difficult to analyze the mechanism of action of drug conjugates using in vitro assays alone. Internalization of methotrexate-immunocon- jugates has been suggested by Garnett and co-workers [9], who showed that ammonium chloride and leupeptin, two inhibitors of lysosomal functions, reduce conjugate activi- ty. The mechanism of action of cis-aconityl-linked an- thracycline conjugates has also been suggested to occur through internalization into acidic intracellular vesicles [7, 32]. However, internalization of immunoconjugates may not be a requirement for all drug-mAb combinations. Starling et al. [26] provided evidence that Vinca-alkaloid immunoconjugates may act through mechanisms other than receptor-mediated endocytosis. In this study, conju- gate internalization was confirmed by comparing anti- tumor activity of the non-modulating 2H7 ADM-Hzn con- jugate to that of the 5E9-ADM-Hzn conjugate on progres- sively growing Daudi tumor xenografts. Despite the fact that the 2H7-ADM conjugate has approximately 2.5 times the antigen-binding sites compared to the 5E9 conjugate, antitumor activity was obtained only with the 5E9 im- munoconjugate. Since internalization was required for cytotoxicity, it is likely that ADM-Hzn immunoconjugates enter into acidic intracellular vesicles such as endosomes and lysosomes. Binding of anti-(transferrin receptor) antibody mediates the delivery of the receptor to the lysosome [14, 30]. The 13-acylhydrazone bond in the conjugate was shown to be acid-sensitive and rapidly (tl/2 = 2.5 h) released unmod- ified ADM from the mAb protein at pH5, a value typically found in intracellular and lysosomal vesicles [27]. The chemical properties of the 13-acylhydrazone bond make it highly compatible with release of ADM following delivery of conjugates into the acidic environment of the cell. The importance of the acid lability of the 13-hydrazone bond to produce cytolytic activity was clearly shown using the oxime-linked conjugate. The ADM-oxime analog was found to be highly resistant to acid conditions and did not release free ADM even at apH value as low as 2.8 (data not shown). The ADM-oxime conjugate was found to be completely inactive in cytotoxicity tests, whereas in the same assay the ADM-Hzn conjugate was active. The data suggest that proteolytic cleavage of mAb protein or reduc- tion of the disulfide bond in the linker may not play a significant role in the cytotoxicity of the conjugates and provide evidence that cytotoxicity was mediated through internalization into acidic compartments with subsequent release of ADM. ADM-Hzn and ADM-Hzn immunoconjugates may be considered acid-sensitive prodrugs of ADM. Free ADM- Hzn had an approximately 10 times decrease in cytotoxic activity in vitro, and was inactive in vivo [5]. Exposure of non-conjugated ADM-Hzn to acidic conditions prior to inoculation in vivo (e.g. conversion to ADM) restored some of the original antitumor activity (results not shown). The lack of in vivo antitumor activity of free ADM-Hzn was probably because it does not efficiently reach acidic compartments within tumor cells. ADM-Hzn conjugates were active because the mAb directed the prodrug to acidic compartments within the tumor cells where it can be acti- vated and released from the rnAb. This type of delivery mechanism also required that the released drug be stable to acid environments. ADM is stable under mild acidic conditions that are present in intra- cellular vesicles. In contrast, mAb-mediated delivery of mitomycin C (MMC) into acidic intracellular compart-
`
`IMMUNOGEN 2090, pg. 7
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`374 ments may result in acid degradation of the MMC. This may partly explain the poor activity of MMC conjugates at low molar ratios and the somewhat better activity of MMC conjugates with high molar ratios achieved by conjugation to polyglutamic acid carrier molecules [11]. Similarly, Huang et al. [15] demonstrated that liposomes containing the acid-sensitive drug cytosine [3-D-arabinofuranoside were significantly less toxic after endocytosis when com- pared to free drug. In contrast, liposomes containing acid- stable methotrexate were more toxic after endocytosis than those containing free methotrexate. In conclusion, we have demonstrated the dual require- ment of