Improved Lysosomal Trafficking Can Modulate the Potency of

Feb 2, 2017 - ... toxins, and conjugation sites. Many studies have indicated that release of toxins bound to antibodies via noncleavable linker chemis...
3 downloads 17 Views 3MB Size
Download PDF
Article pubs.acs.org/bc

Improved Lysosomal Trafficking Can Modulate the Potency of Antibody Drug Conjugates Rachel M. DeVay,†,§ Kathy Delaria,† Guoyun Zhu,† Charles Holz,† Davide Foletti,†,∥ Janette Sutton,† Gary Bolton,† Russell Dushin,‡ Christine Bee,†,# Jaume Pons,†,⊥ Arvind Rajpal,†,# Hong Liang,†,§ David Shelton,† Shu-Hui Liu,† and Pavel Strop*,†,# †

Rinat Laboratories, Pfizer Inc., 230 East Grand Avenue, South San Francisco, California 94080, United States Worldwide Medicinal Chemistry, Pfizer Inc., 445 Eastern Point Road, Groton, Connecticut 06340, United States

‡

S Supporting Information *

ABSTRACT: Antibody drug conjugates (ADCs) provide an efficacious and relatively safe means by which chemotherapeutic agents can be specifically targeted to cancer cells. In addition to the selection of antibody targets, ADCs offer a modular design that allows selection of ADC characteristics through the choice of linker chemistries, toxins, and conjugation sites. Many studies have indicated that release of toxins bound to antibodies via noncleavable linker chemistries relies on the internalization and intracellular trafficking of the ADC. While this can make noncleavable ADCs more stable in the serum, it can also result in lower efficacy when their respective targets are not internalized efficiently or are recycled back to the cell surface following internalization. Here, we show that a lysosomally targeted ADC against the protein APLP2 mediates cell killing, both in vitro and in vivo, more effectively than an ADC against Trop2, a protein with less efficient lysosomal targeting. We also engineered a bispecific ADC with one arm targeting HER2 for the purpose of directing the ADC to tumors, and the other arm targeting APLP2, whose purpose is to direct the ADC to lysosomes for toxin release. This proof-of-concept bispecific ADC demonstrates that this technology can be used to shift the intracellular trafficking of a constitutively recycled target by directing one arm of the antibody against a lysosomally delivered protein. Our data also show limitations of this approach and potential future directions for development.

■

INTRODUCTION Antibody drug conjugates (ADCs) have emerged as a promising class of cancer therapeutics due to their ability to specifically deliver highly potent cytotoxic agents to tumor cells.1−3 Most commonly, ADCs rely on binding to receptors on the cancer cell surface, followed by internalization and cytotoxic payload release. Recent advances in ADC engineering have led to the development of multiple technologies and payloads that have diversified ADCs beyond just target selection. As a result, ADC design can also vary by the type of toxin, the linker chemistry used for its conjugation, the number of toxins conjugated to an antibody molecule, and finally the location of the conjugation site(s).4−11 The linkers that are used to attach toxins to the antibodies can significantly affect ADC potency and toxin release.12 Two main classes of linkers are currently utilized in the clinic: cleavable and noncleavable. Cleavable linkers are those that can be cleaved chemically or enzymatically. They may release their toxins more effectively than their noncleavable counterparts, but their enhanced activity can come with an increase in nonspecific cleavage and release of toxin in the circulation. On the other hand, noncleavable linkers are hypothesized to be more stable in circulation since ADCs utilizing such chemistries © 2017 American Chemical Society

likely release their toxins following internalization into their target cells and ADC degradation.13−15 Presumably, the deeper into the endolysosomal pathway they are delivered, the more effectively noncleavable ADC toxins are released. However, the potency of these ADCs likely relies on the expression and intracellular trafficking of their targets to internalize and release a minimum threshold concentration of toxin needed to mediate cell death. Therefore, it might be possible that a highly expressed target that is recycled or retained on the cell surface requires relatively high concentrations of ADC or higher drug antibody ratios (DARs),16,17 while an ADC against a target that has relatively low expression may be comparably potent if it is trafficked more efficiently to lysosomes. Cell surface localized receptors are internalized via a variety of mechanisms. For example, Amyloid Precursor Like Protein 2 (APLP2) has been shown to travel directly to lysosomes following endocytosis.18−21 APLP2 is ubiquitously expressed, and it has been shown to have increased levels in a variety of cancer cell lines.22 Interestingly, APLP2 efficiently delivers its Received: January 7, 2017 Revised: February 1, 2017 Published: February 2, 2017 1102

DOI: 10.1021/acs.bioconjchem.7b00013 Bioconjugate Chem. 2017, 28, 1102−1114

Article

Bioconjugate Chemistry

Figure 1. Comparison of Trop2 and APLP2 expression and trafficking. (A) Relative ratio of cell surface localized receptor numbers of Trop2:APLP2 in SKOV3, N87, BxPC3, and Colo205 cell lines. (B) Internalization kinetics for Trop2 and APLP2 in SKOV3 cells. Background signal calculated as measured fluorescence from noninternalized control cells. (C,D) Trafficking of internalized anti-Trop2 or anti-APLP2 (green) in (C) SKOV3 cells or (D) Colo205 cells after 4 h. Colocalization colormap showing positive colocalization (yellow to red on color bar) vs negative correlation (green to blue on color bar). Scale bar = 10 μm; Lamp2 staining for lysosomes and late endosomes shown in red; DAPI staining shown in blue.

hypothesized that by linking HER2 to APLP2 via a bispecific ADC, we could redirect HER2 to lysosomes and potentially improve its efficacy.

interacting partners, and complexes thereof, from the cell surface along its endocytic pathway to lysosomes.18,20,21 Indeed, APLP2 binds and delivers PCSK9 and proteins that are in complex with PCSK9 (such as monoclonal anti-PCSK9 antibodies or the constitutively recycled LDL receptor) to lysosomes.18,19 Although the details of HER2 endocytosis and trafficking are complex and involve many different players (reviewed23), it is theorized that HER2 is recycled following internalization.24,25 HER2 is a target of several antibody-based therapeutics, including ADCs, due to its high expression in a subset of breast cancers and other cancer types.26−32 Anti-HER2 efficacy might be confounded, however, by its constitutive recycling, and attempts to modify its trafficking via clustering have been made to abrogate this.33,34 Dimerization caused by αHER2 antibodies has been proposed as a potential mechanisms of action of anti-Her2 monoclonal antibodies.35,36 Another study showed that inhibition of Hsp90 alters HER2 trafficking by shifting its endocytic trafficking route toward lysosomes.37 In this study, we first assessed whether intracellular trafficking affects noncleavable ADC mediated cell killing by comparing an ADC with efficient lysosomal targeting (αAPLP2 ADC) to an ADC targeting Trop2 (αTrop2-ADC, tumorassociated calcium signal transducer 2), an archetypal ADC tumor cell target. Trop-2, also known as TACSTD2, EGP-1, GA733-1, and M1S1, is a type I transmembrane protein frequently expressed on a variety of human carcinomas and its expression is often associated with poor prognosis of the diseases. When bound to an antibody, Trop2 is internalized with an internalization t1/2 of ∼30 min.38 Furthermore, we explored the possibility of using bispecific ADCs to manipulate the intracellular trafficking pathways of promising oncology targets, such as HER2, in an effort to increase their efficacy. We

■

RESULTS

APLP2 and Trop2 Follow Distinct Postendocytic Trafficking Routes. Noncleavable ADCs are thought to require antibody degradation in the lysosomal compartment in order to release the toxin and exhibit cytotoxic activity. Given this prerequisite, it is likely that slow lysosomal trafficking kinetics or high rates of target:ADC complex recycling would lead to a lower level of efficacy. To investigate this hypothesis, we wanted to test whether an antibody targeting a protein that is routed to lysosomes directly from the cell surface could mediate cell killing more efficiently than an antibody against a typical target that might show various degrees of recycling and lysosomal trafficking. We selected APLP2 protein (a member of the amyloid precursor protein family), that has been shown to be efficiently routed from the cell surface to lysosomes.18,20 Additionally, APLP2 is able to concomitantly transport its binding partners (such as antibodies) to lysosomes. Here, APLP2 trafficking was compared to that of Trop2, a protein that is highly expressed on many tumor cells with Trop2 ADCs currently undergoing investigation in the clinic.39−41 We have previously shown that in low-expressing cell lines where target numbers are limiting, the potency of ADCs can be improved by increasing the number of conjugated toxins per antibody.16 Another way to improve the potency of noncleavable ADCs might be to improve their trafficking to lysosomes. The internalization of Trop2 is comparable to other targets; however, APLP2 trafficking to lysosomes is much more efficient.38 We hypothesized that in cells that express comparable levels of APLP2 to Trop2, APLP2-ADC molecules 1103

DOI: 10.1021/acs.bioconjchem.7b00013 Bioconjugate Chem. 2017, 28, 1102−1114

Article

Bioconjugate Chemistry

Figure 2. Lysosomal delivery of antibody drug conjugates increases cell killing efficiency in vitro. Cytotoxicity of negative control ADC (NCPEG6MMAD), αTrop2-PEG6MMAD, and αAPLP2-PEG6MMAD was evaluated on target-expressing (A) Colo205 and (B) SKOV3 cells after 96 h as described in the Experimental Section. (C) Summary table of EC50 values (nM) and maximal cell killing (%) from representative experiments of all conjugates tested on the SKOV3 and Colo205 cell lines.

route in SKOV3 cells (isotype control, NC, shown in Figure S1B), as evidenced by the punctate localization of internalized monoclonal antibodies clustered in and around lysosomes (Figure 1C; Pearson’s correlation above threshold of 0.42). In contrast, Trop2 monoclonal antibodies remain diffusely localized on the cell surface or endocytic vesicles, but do not reach lysosomes efficiently (Figure 1C; Pearson’s correlation above threshold of 0.08). Colocalization of Trop2 and APLP2 antibodies with Lamp2 positive compartments is depicted in the far right panels using colocalization heat maps. The color bar above the images indicates the color spectrum associated with positive or negative correlation between the antibodies and Lamp2. The same trafficking pattern for both Trop2 and APLP2, but importantly not isotype control (NC), was also observed in Colo205 (Figure 1D; Figure S1A) and BxPC3 cells (Figure S1C), indicating that these trafficking routes are consistent across multiple cell lines with varying relative expression levels of Trop2 and APLP2. Together, these data show that while the endocytic internalization rates of Trop2 and APLP2 are similar, trafficking of these proteins diverges post-endocytosis where APLP2 travels more efficiently to lysosomes. Lysosomal Delivery of Antibody Drug Conjugates Increases Cell Killing Efficiency. Previously, we described site-specific conjugations that allowed for controlled higher loading of toxin to antibody, which in turn increased the efficacy of ADCs.16 In addition to site-specific conjugation, the stability of the linkers used to attach the toxins varies both in the blood and post-endocytosis. Noncleavable linkers can

would more efficiently mediate cell killing than Trop2-ADC molecules due to their more efficient lysosomal delivery. In order to directly compare recycling vs lysosomal transport, the cell surface expression levels and rates of internalization would need to be consistent between Trop2 and APLP2. We therefore sought to identify cell lines with comparable cell surface expression and internalization rates for these targets. We first determined cell surface expression of Trop2 and APLP2 on a variety of cell lines using a FACS based binding method (Figure 1A), and found that SKOV3 cells express the most comparable levels of Trop2 and APLP2 (2:1 ratio) out of the cell lines tested. Internalization rates of Trop2 and APLP2 in SKOV3 cells were then measured to determine whether uptake of ADCs against these targets would be similar. To assess this, monoclonal antibodies directly conjugated with an Alexa488 dye identifying Trop2 (αTrop2) or APLP2 (αAPLP2) were incubated and allowed to internalize in SKOV3 cells. A quenching antibody against Alexa488 was added to eliminate signal from the cell surface, and the resulting fluorescence intensities per cell of αTrop2 and αAPLP2 were measured over time and normalized to background signals to calculate their respective linear rates of internalization. Importantly, Trop2 and APLP2 internalization rates were similar (0.14/min ± 0.01 vs 0.10/min ± 0.01, respectively) in SKOV3 cells (Figure 1B). Thus, differences in post-endocytic trafficking could be directly compared since differences in cell surface turnover were negligible. As described previously, 22 APLP2 is transported to lysosomes following endocytosis. APLP2 follows this same 1104

DOI: 10.1021/acs.bioconjchem.7b00013 Bioconjugate Chem. 2017, 28, 1102−1114

Article

Bioconjugate Chemistry

Figure 3. αAPLP2-ADC is more efficacious than αTrop2-ADC on Colo205 xenograft model. (A) In vivo efficacy comparison of the αTrop2-ADC and αAPLP2-ADC conjugates in the Colo205 xenograft model, along with a negative control conjugate NC-PEG6MMAD. All compounds were dosed at 3, 6, or 10 mpk. Once tumors were established, mice were dosed intravenously with αAPLP2-ADC or αTrop2-ADC at 3, 6, or 10 mpk. Tumors were measured biweekly over the course of 55 days. (B) Total antibody pharmacokinetic profiles in mice of the αTrop2-ADC and αAPLP2ADC. (C) ADC pharmacokinetic profiles in mice of the αTrop2-ADC and αAPLP2-ADC. Pharmacokinetics of total antibody as well as ADC were measured by ELISA as described in the Experimental Section. Both compounds were given as a single dose at 3 mpk.

provide higher stability in circulation, but they require antibody degradation by lysosomal enzymes to release the toxins and kill cells.14,42−44 One method to improve release of noncleavable toxins from ADCs once they are internalized might be to increase their lysosomal delivery. To test this, we conjugated Trop2 and APLP2 antibodies to the MMAD45,46 toxin via a noncleavable AmPEG6 linker.44 We utilized site-specific conjugation via transglutaminase to minimize instability issues that were reported for maleimides.10,11,16,43,44,47−51 We generated Trop2 and APLP2 conjugates with either two (DAR2) or four drugs per antibody (DAR4), as described in Experimental Section. To test our hypothesis that lysosomal delivery of ADCs increases their efficacy, we tested the ability of αTrop2-ADC and αAPLP2-ADC to kill SKOV3 cells. Interestingly, SKOV3 were significantly more sensitive to αAPLP2-ADC than they were to αTrop2-ADC (Figure 2B), although they express approximately twice the amount of Trop2 relative to APLP2 (Trop2:APLP2 ratio of 2:1). Indeed, the EC50 of αAPLP2ADC was 0.11 ± 0.04 nM versus 5.95 ± 1.6 nM for αTrop2ADC (Figure 2C). In addition, the maximal killing of αAPLP2ADC reached 79% compared to 57% for αTrop2-ADC. The isotype control antibody induced some cell killing at the highest

concentration tested (Figure 2B). Together, these data suggest that lysosomal targeting increases the efficiency with which ADCs are able to mediate cell death. We next looked at the effect of these ADCs in Colo205 cells, where Trop2 expression exceeds that of APLP2 by a ratio of approximately 13:1 (Figure 1A). Interestingly, αAPLP2-ADC was consistently more efficacious in Colo205 cells than αTrop2-ADC (Figure 2A) with an EC50 of 0.23 ± 0.01 nM vs 0.94 ± 0.08 nM for αTrop2-ADC (Figure 2C). In addition, the maximal killing of αAPLP2-ADC reached 85% compared to 72% for αTrop2-ADC. Thus, efficient lysosomal targeting may be able to compensate for relatively low target expression, and the intracellular trafficking of ADC targets may greatly affect their efficacy. To determine whether these observed effects are translated in vivo, we tested these molecules in a Colo205 xenograft model in athymic nude mice. This model was successfully used for previous ADC studies16 and represents a model that is moderately sensitive to noncleavable αTrop2-ADC treatment. At all doses, αAPLP2-ADC outperformed the αTrop2-ADC, even though Trop2 expression is significantly higher. Notably, even at the lowest dose (3mpk), αAPLP2-ADC treatments resulted in more pronounced tumor regression than the 10mpk 1105

DOI: 10.1021/acs.bioconjchem.7b00013 Bioconjugate Chem. 2017, 28, 1102−1114

Article

Bioconjugate Chemistry

Figure 4. αAPLP2/αHER2 bispecific alters HER2 localization to lysosomes. (A−C) Relative expression of HER2 and APLP2 in JIMT-1, MCF7, and N87 cells, as assessed by Western blot using primary antibodies (A) αHER2 or (B) αAPLP2. (C) Summary of relative HER2:APLP2 results. (D) HER2 localization (green) relative to lysosomes (Lamp2; red) after 4 h antibody treatment in JIMT-1 cells. DAPI is shown in blue. Scale bar = 10 μm.

our hypothesis, we first determined the relative expression levels of HER2 and APLP2 in a variety of cell lines by quantitative Western blot (Figure 4A−C). Standard curves were first created to calculate the linear relationship between the Western blot signal and corresponding protein concentration for both αAPLP2 and αHER2 antibodies. Notably, Western detection by αAPLP2 has a higher signal for its associated protein concentrations than the HER2 antibody. JIMT-1 and MCF7 cell lines each showed comparable expression levels of HER2 and APLP2. Herceptin resistant cell lines were selected to isolate the effect of toxin delivery to lysosome from functional inhibition of Her2 signaling and to better mimic potential clinical use of αHer2-ADC. HER2 expression in MCF7 cells was generally of very low expression, but the signal was still within the dynamic range of the Licor Odyssey imaging system and the linear range of our standard curves, for both proteins. HER2 receptor numbers were previously reported for MCF7, JIMT-1, and N87 to be 6000, 65 000, and 1 300 000 copies of HER2 per cell, respectively.34 We next looked at subcellular localization of HER2 following treatment with negative control antibody (NC), αAPLP2, Herceptin, and the αAPLP2 /αHER2 bispecific antibody in JIMT-1, MCF7, and N87 cells. After a 4 h incubation no change in HER2 localization was observed for αAPLP2 or NC antibody treatment in JIMT-1 cells (Figure 4D; Pearson’s correlation −0.32 and −0.61, respectively). αHER2 treatment showed some shift from its typical surface distribution toward lysosomes (Pearson’s correlation −0.05), with further increase in the presence of αAPLP2 /αHER2 bispecific antibody (Figure 4D; Pearson’s correlation 0.36). Colocalization colormaps visually depicting the correlation between HER2 and Lamp2 are shown in Figure S2A This effect was also observed in MCF7 cells and N87 cells (Figures S3 and S4, respectively). Notably, N87 cells have very high expression of HER2 and

αTrop2-ADC (Figure 3A). The pharmacokinetics of these two ADCs were nearly identical, and therefore this effect was not due to differences in exposure of the two different ADCs (Figure 3B,C). Thus, αAPLP2-ADC is more efficacious in a Colo205 xenograft model than αTrop2-ADC, despite its dramatically lower expression levels. We hypothesize that this effect is due to a more efficient lysosomal trafficking route of APLP2 that enables increased efficiency of drug release compared to Trop2. Bispecific Antibodies Redirect Recycled ADC Targets to Lysosomes. Our data thus far suggest that lysosomal targeting of noncleavable ADC via APLP2 results in increased cell killing relative to Trop2 ADCs, presumably due to more efficient toxin release. We hoped to apply this theory to the established ADC therapeutic Kadcyla in an effort to improve its efficacy. Herceptin targets HER2, a protein that is highly expressed in a subset of breast and ovarian cancers, but which is recycled back to the cell surface following endocytosis.24,25,34,52−54 In an effort to take advantage of the ability that APLP2 has to deliver its interacting partners to lysosomes for the purpose of killing tumor cells, we constructed bispecific ADC molecules to simultaneously bind APLP2 and HER2. We hypothesized that in this scenario, the ADC could be homed to tumor cells efficiently via the HER2 arm, and post-endocytosis, the APLP2 arm could facilitate delivery of the entire complex (HER2:ADC:APLP2) to lysosomes. Thus, HER2 trafficking would be altered in the presence of the bispecific ADC molecule, allowing the ADC to be directed to lysosomes for toxin release. To achieve this, we utilized our full-length bispecific antibody technology55 and created antibodies with one αHER2 (Herceptin) arm and one arm of a αAPLP2 antibody. Importantly, the bispecific antibodies bound to recombinant extracellular domains (ECD) of both APLP2 and HER2 simultaneously by a sandwiching ELISA (Figure S2B). To test 1106

DOI: 10.1021/acs.bioconjchem.7b00013 Bioconjugate Chem. 2017, 28, 1102−1114

Article

Bioconjugate Chemistry

Figure 5. In vitro comparison of αHER2, αAPLP2, and αAPLP2/αHER2-ADC mediated cell killing. (A−F) Cytotoxicity of NC-PEG6MMAD, and bispecific ADCs with either DAR2 or DAR4 were evaluated on target-expressing cell lines: MCF7 DAR2 (A) and DAR4 (B), JIMT-1 DAR2 (C) and DAR4 (D), and N87 DAR2 (E) and DAR4 (F) as described in the Experimental Section. Shown is an average of replicate experiments with standard deviation (SD). (G) Summary table of average ± SD of EC50 (nM) and maximal killing (%) of all conjugates tested on the three cell lines. Significance determined from Student’s t test. *p < 0.05; **p < 0.005.

APLP2 and HER2. MCF7 cells were not sensitive to αHER2ADC alone, and the bispecific αAPLP2/αHER2-ADC performed equivalently to the αAPLP2/NC-ADC monospecific ADCs for the DAR2 and DAR4 loaded antibodies, indicating that due to its relatively low expression, the HER2 targeting arm did not have an effect on the APLP2 arm (Figure 5A,B,G). JIMT-1 cells, which have comparable expression levels of HER2 and APLP2, were more sensitive to αAPLP2/αHER2ADC and had significantly lower EC50 values than the monovalent HER2-ADC antibody (DAR2: 0.13 nM vs 0.37 nM; Student’s t test