2.1. Gemtuzumab Ozogamicin
Gemtuzumab ozogamicin (GO; Mylotarg™) is composed of a humanized monoclonal anti-CD33 IgG4 (immunoglobulin G4) antibody conjugated to the payload calicheamicin at an average drug-antibody ratio of 2 to 3. The unarmed antibody has been shown to have no cytotoxic potency against acute myeloid leukemia (AML) cells in vitro and to mediate neither complement-binding nor antibody-dependent cell-mediated cytotoxicity (ADCC) in vivo [4
]. The bifunctional 4-(4-acetylphenoxy) butanoic acid (AcBut) linker of GO attaches on the one end via amid bonds to surface-exposed lysine residues of the antibody and forms on the other end an acyl hydrazone linkage with the payload. Upon internalization of GO, acidification in the lysosomal pathway releases calicheamicin as a prodrug that is activated by undergoing spontaneous reaction with reduced glutathione within the cytosol [5
]. The CD33 surface antigen is an inhibitory receptor and adhesion molecule, whose normal expression pattern is largely restricted to cells from the myeloid lineage, including Kupffer cells and circulating macrophages [6
]. CD33-positive leukemia is defined as either presence of CD33 on greater than 20–25% of the leukemic blasts or by CD33 immunofluorescence staining greater than fourfold above background or by 80% CD33-positive cells by flow cytometry [7
]. Despite the heterogeneity of CD33 expression on AML cells, more than 80% of patients fulfill at least one of these criteria for CD33-positive disease. In May 2000, Gemtuzumab ozogamicin (GO) received accelerated FDA approval as a stand-alone treatment for CD33-positive AML patients of 60 years or older in first relapse who are not candidates for other chemotherapies. However, Pfizer voluntarily withdrew the ADC from the U.S. market in 2010, after subsequent trials failed to verify its clinical benefit and, even worse, observed a high number of early deaths due to veno-occlusive liver disease. In 2014, a meta-analysis of five randomized trials spurred renewed interest in the therapeutic potential of GO by showing that 5-year survival rates for patients with favorable cytogenetics and intermediate risk were improved by adding GO to induction chemotherapy [8
]. In follow-up studies, GO at reduced doses was confirmed to improve survival compared to best supportive care with a favorable risk/benefit profile [9
]. In September 2017, the FDA then re-approved GO at a 3-fold lower recommended dose and with a different dosing schedule as either monotherapy or in combination with chemotherapy for the treatment of adults with newly diagnosed CD33-positive AML as well as for the treatment of relapsed/refractory AML patients.
Currently, GO is the only FDA-approved, CD33-targeted therapy, since clinical development of the next most advanced CD33-targeted ADC, which used a PBD payload, was stopped due to increased rates of death in the treatment versus control arm [10
]. Although CD33 surface levels are a key determinant of how much GO can bind to AML cells, assessing whether patients with higher CD33 expression levels benefit more from GO therapy is complicated by the fact that CD33 levels are associated with established prognostic factors, including genetic subgroups. However, a recent study demonstrated CD33-positivity to independently correlate with GO benefit for younger and older adults with AML except for AML forms characterized by cytogenetic rearrangements that disrupt genes encoding subunits of a transcription factor known as a core-binding factor [11
]. Another recent study determined the impact of single nucleotide polymorphism (SNP) in the splice enhancer region of the CD33 gene that regulates the expression of an alternatively spliced CD33 isoform without exon 2, which contains the antibody binding site for GO [12
]. Patients with the CC genotype of this SNP had significantly lower levels of the CD33 isoform lacking exon 2 and, in contrast to the other SNP genotypes, their risk of relapse was almost halved by adding two doses of GO to a standard five-course chemotherapy regimen.
Acquired resistance to GO was frequently observed to be mediated by multidrug resistance mechanisms [13
]. In vitro treatment with inhibitors of p-glycoprotein (MDR1/multidrug resistance protein 1/ABCB1) and other multidrug resistance proteins increased cytotoxic potency of GO on resistant cell lines and patient samples [14
]. In HL-60 cells made GO resistant by chronic exposure, it was shown that MDR-1 is highly expressed. While MDR-1 up-regulation was reversible upon GO withdrawal, the resistant cells, in contrast to the parental line, retained the ability to rapidly re-induce MDR-1 upon re-exposure [16
]. Activated signaling via the phosphatidylinositol-3-kinase (PI3K)/AKT pathway is another mechanism associated with GO resistance in vitro in primary AML cells. MK-2206, a selective AKT inhibitor, significantly sensitized resistant human AML cells to GO and free calicheamicin [17
]. These findings support the notion that DNA damage is necessary but not sufficient for cell killing by ADCs with a calicheamicin payload since the ability of cells to repair DNA damage and to activate downstream anti-apoptotic factors can critically modulate GO cytotoxicity. The mechanism underlying PI3K/AKT-mediated GO resistance appears to increase survival signaling via activation of anti-apoptotic factors. In agreement with this, GO treatment was found to induce p38 stress kinase activation and to cause the cell death mediators, Bak and Bax, to adopt a pro-apoptotic conformation in two sensitive AML cell lines but not in a resistant one [18
]. Maimaitili et al. found that in six of eight tested AML cell lines, the cytotoxicity of GO could be synergistically enhanced by concurrent treatment with the mTORC1/2 inhibitor, PP242 [19
]. GO sensitization of cells by PP242 is mediated by a dual mechanism that combines enhanced lysosomal function with blockage of activation of the checkpoint kinase Chk1, a key regulator of DNA damage-induced cell cycle arrest.
2.2. Brentuximab Vedotin
In 2011, Brentuximab vedotin (BV, Adcetris™) was FDA-approved for the treatment of two blood cancer indications: in Hodgkin’s lymphoma (HL) for patients that have failed an autologous stem cell transplant (ASCT) and in systemic anaplastic large cell lymphoma (sALCL) for patients that have relapsed after multi-agent chemotherapy. BV consists of SGN-30, a chimeric IgG1 antibody against CD30, conjugated by a protease-cleavable valine-citrulline peptide linker to the membrane permeable tubulin polymerization inhibitor, monomethyl auristatin E (MMAE). CD30, as a surface marker, is normally restricted to activated T- and B-cells and is implicated in autoimmune regulation, but on certain lymphoid cancers, its surface levels are substantially elevated. While BV treatment has generated compelling response rates in both indications, treatment with unarmed SGN-30 at up to 7-fold higher doses than BV yielded no objective responses in HL and only a 17% response rate in ALCL [20
]. This suggests that the antibody component of BV does not significantly contribute to overall efficacy per se
neither by functionally blocking CD30 nor by recruiting immune effector cells. Long-term follow-up of the pivotal study that led to approval of BV treatment for HL patients demonstrated 5-year rates of 22% for progression-free survival (PFS) and of 41% for overall survival (OS) [21
]. Thirty-four percent of treated patients initially achieved a complete response indicating that their leukemic cells were not innately resistant to BV. The 5-year outcome of such complete responders was even more favorable with PFS and OS rates of 52% and 64%, respectively. But still, the fact that more than a third of the complete responders relapsed and died within 5 years of treatment shows that their leukemic cells acquired resistance to BV.
Chen et al. investigated resistance mechanisms in vitro by constant BV exposure of an ALCL cell line and pulsatile BV exposure of an HL cell line [22
]. The cell lines escaped from continued or repeated cytotoxic effects of BV by various mechanisms that included down-regulating CD30, up-regulating the p-glycoprotein drug transporter, and developing resistance to MMAE. While up-regulation of members of the drug transporter class was also confirmed by immunohistochemistry in clinical samples from BV resistant HL and ALCL patients, none of these samples showed down-regulation of CD30. Moreover, in vitro resistance was neither correlated with the percentage of CD30-positive cells nor with the median cell surface CD30 signal intensity for a given cell population. Thus, it appears that factors other than regulation of CD30 levels contribute substantially to the resistance phenotype. Drug transporters, like p-glycoprotein, preferentially export hydrophobic cargo out of cells. Therefore, one way to overcome the effect of their up-regulation would be to replace the non-charged MMAE payload and peptide linker of BV with a payload/linker combination that produces charged or strongly polar metabolites [23
]. The caveat of such a strategy is that it improves potency on multidrug-resistant tumor cells at the expense of limiting the so-called bystander effect. Bystander killing of surrounding tumor cells can occur when active metabolites of ADCs are either prematurely released outside of cells or diffuse out of tumor cells after ADC internalization and degradation or leak from dying tumor target cells. The ability of ADC payload metabolites to cross the biomembranes of surrounding cells determines the extent of bystander killing and this ability is decreased for charged or polar molecules.
While preclinical and clinical findings show that leukemic cells can adapt to the cytotoxic stress exerted by BV treatment, such resistance-mediating adaptations need not necessarily be stable after treatment stops. A Phase II study investigated BV retreatment of 20 HL patients and eight sALCL patients who previously had achieved complete or partial response with this drug. For HL patients, retreatment yielded an objective response rate of 60% and a complete response rate of 30%. For sALCL patients, response rates were even better with 88% and 63%, respectively [24
]. The median duration of renewed partial or complete responses was 9.2 and 12.3 months for the two patient populations, respectively. This demonstrates that retreatment after therapy discontinuation due to emerging acquired resistance is a therapeutic option for many patients. However, it should be noted that peripheral neuropathy is a known adverse effect of BV treatment that occurred more frequently in retreated patients (~30%). In fact, most patients discontinuing retreatment did so for this reason. Peripheral neuropathy must be clinically managed by dose reduction potentially limiting the efficacy of BV in these patients. The instability of the BV resistance phenotype after treatment discontinuation suggests that underlying stress adaptations come at a certain cost for the vitality of leukemic cells.
2.3. Trastuzumab Emtansine
In 2013, the FDA approved Trastuzumab emtansine (T-DM1, Kadcyla™) for the treatment of HER2-positive metastatic breast cancer. T-DM1 consists of the humanized anti-HER2 IgG1 antibody trastuzumab conjugated via a non-cleavable succinimidyl 4-(N-maleimidomethyl) cyclohexane-1-carboxylate (SMCC) linker to DM-1, a microtubule inhibiting maytansinoid. Two properties set T-DM1 apart from most other ADCs: (i) due to gene amplification, its target antigen HER2 is extremely abundant on the targeted tumors [25
] with more than one million molecules per cell (this is about 10–100 fold higher than other ADC targets) and (ii) trastuzumab, as an unarmed antibody (Herceptin™), was already FDA-approved for the same indication in 1998 [26
], based on its ability to functionally neutralize HER2 and to also exert antibody-dependent cell-mediated cytotoxicity (ADCC). As monotherapy, trastuzumab is given at a loading dose of 8 mg/kg in the first week and from week 4 onwards, at 6 mg/kg every 3 weeks, while T-DM1 is dosed continuously at 3.6 mg/kg every 3 weeks from start. Thus, despite somewhat lower antibody exposure for T-DM1 than for trastuzumab treatment, functional blocking of HER2 signaling and ADCC could conceivably contribute to the overall efficacy of T-DM1.
Two Phase III trials established T-DM1 as a standard of care in second or higher line therapy for HER2-positive, metastatic breast cancer patients. In the EMILIA trial [27
], T-DM1 was compared to oral combination therapy with capecitabine, a 5-fluorouracil prodrug, plus the HER2 inhibitor, lapatinib. It improved PFS by 3.2 months and OS by 4 months, while almost doubling the duration of response to 12.5 months. The TH3RESA trial compared T-DM1 versus treatment of physician’s choice in locally advanced or metastatic breast cancer patients who had received a taxane in any setting and had progressed after treatment with two or more HER2-directed regimens in the advanced setting [28
]. In this difficult to treat patient population, T-DM1 almost doubled PFS and increased OS by 7 months despite the fact that about half of the patients in the physician’s choice group were allowed to cross over to T-DM1 therapy. The fact that all treated patients had previously progressed on a microtubule inhibiting drug, as well as on at least two anti-HER2 regimens, clearly illustrates that the efficacy of an ADC is more than just the sum of its parts. Importantly, T-DM1 was also better tolerated than the control treatments in all Phase III trials.
The EMILIA trial also reported an interesting difference in therapeutic outcome between treatment arms for patients with PIK3CA mutations [30
]. Cancer cells with such mutations become resistant to HER2 signaling blockade due to constitutive activation of the alternative signaling pathway triggered by PI3K. While PFS was similar for T-DM1-treated patients with and without PIK3CA mutations (10.9 vs. 9.8 months, respectively), it was worse for patients treated with lapatinib and capecitabine (4.3 vs. 6.4 months). This indicates that for the overall clinical efficacy of T-DM1, the targeting function of trastuzumab might play a bigger role than its HER2 signaling inhibition functionality.
More recently, the Phase III MARIANNE study investigated T-DM1 as first-line HER2 therapy in patients with progressive/recurrent locally advanced breast cancer or untreated metastatic breast cancer. With regards to PFS as the primary endpoint, T-DM1 monotherapy, as well as its combination with pertuzumab, another FDA-approved HER-2 antibody, was non-inferior but also not superior to combination therapy with trastuzumab plus a taxane [31
]. Although the reasons for this are not fully understood, it should be noted that T-DM1 has a positively charged payload and hence there is no bystander effect, whereas taxanes readily cross biomembranes and are therefore also toxic for tumor cells with low or no HER2 expression. Under selective pressure by anti-HER2 treatment, subclones with limited or no HER2 expression can become dominant in a tumor that has ab initio heterogeneous HER2 expression [32
]. This will gradually result in increased resistance to T-DM1 but not to taxanes [33
] and adding pertuzumab to T-DM1 does not change this. Perhaps not too surprising then, the phase III KRISTINE trial [34
] found that for neoadjuvant treatment of HER2-positive patients with localized, operable breast cancer combining T-DM1 with pertuzumab was inferior to a combination of trastuzumab, pertuzumab, and two chemotherapeutics (docetaxel and carboplatin) with regards to pathologic complete response. However, T-DM1 was superior to trastuzumab in the recently published phase III KATHARINE trial comparing the adjuvant treatment of patients with HER2-positive early breast cancer who had residual invasive disease at surgery following neoadjuvant therapy containing a taxane and trastuzumab. After random assignment to 14 cycles of adjuvant therapy, the risk of recurrence of invasive breast cancer or death was cut in half by T-DM1 versus trastuzumab treatment [35
]. Given the clinical success of T-DM1, it’s not surprising that there are currently at least nine other HER2-based ADCs in clinical development [36
Due to its lack of bystander effect, T-DM1 tumor uptake and therapeutic response are expected to correlate closely with expression levels of HER2. Preclinical confirmation comes from microPET/CT imaging of zirconium-78 labeled T-DM1 in mouse xenograft studies [37
], while clinical proof is provided among others by a biomarker analysis of the TH3RESA trial demonstrating that patients with higher HER2 mRNA levels benefited more from receiving T-DM1 [38
]. In agreement with this, Loganzo et al. found by exposing HER2-overexpressing cell lines in vitro to multiple cycles of T-DM1 treatment at IC80
levels that HER2 down-regulation was the primary mechanism of acquired resistance besides the up-regulation of drug efflux transporters [39
]. In another preclinical study, Sung et al. also reported down-regulation of HER2 as a mechanism of acquired resistance in two breast carcinoma cell lines, while a gastric carcinoma cell line developed T-DM1 resistance by altered internalization and trafficking of T-DM1 instead [40
]. In the latter cell line, proteins that mediate caveolae formation and endocytosis were enriched, and T-DM1 was internalized into intracellular caveolin-1 positive endocytic compartments with neutral pH. These endosomes differ with regards to pH and proteinase activities from the increasingly acidic endo-lysosomal maturation pathway, for which T-DM1 is destined in sensitive cells. Not surprisingly, the same cell line was cross-resistant to trastuzumab-ADCs comprising auristatin analogs conjugated by a non–cleavable maleimide linker because of its inability to fully degrade the antibody moiety in endosomes which prevents payload release. Resistance was overcome, however, when the same ADCs contained a protease-cleavable valine-citrulline peptide linker instead. The payloads could then be released by linker cleavage and, unlike lysine-MCC-DM1, the active metabolite of T-DM1, the uncharged auristatin analogs are able to escape into the cytosol by crossing the endocytic membrane.
Diminished cytosolic accumulation of lysine-MCC-DM1 was also observed in another highly T-DM1 resistant gastric carcinoma cell line derivative [41
]. In that case, reduced activity of the vacuolar H+-ATPase (V-ATPase) led to diminished lysosome acidification, which, in turn, affected lysosomal proteinase activities and prevented efficient degradation of T-DM1. In an independent study, Rios-Luci et al. also reported that resistant cell clones with increased lysosomal pH had impaired proteolytic degradation of T-DM1 [42
]. Moreover, Hamblett et al. discovered that the active metabolite, lysine-MCC-DM1, requires the presence and activity of the lysosomal transporter, SLC46A3, in order to escape into the cytosol and that down-regulation of SLC46A3 can render cells T-DM1 resistant [43
]. A combined loss of SLC46A3 and PTEN functionality was reported to contribute to T-DM1 resistance in yet another breast cancer cell line [44
]. Silencing of SLC46A3 conferred partial resistance, and a PI3K inhibitor sensitized cells to T-DM1. Thus, the resistance observed in the reported cell line is likely a combined effect of the reduced cytosolic escape of lysine-MCC-DM1 from lysosomes and apoptosis resistance due to constitutive survival signaling by the PI3K pathway. Wang et al. reported, for a breast cancer cell line, another anti-apoptotic signaling mechanism resulting in T-DM1 resistance in vitro. They found that aberrant activation of STAT3 due to overexpression and ligand-induced signaling of the LIF (Leukemia Inhibitory Factor) receptor blocked apoptosis induction by T-DM1 [45
Moreover, PLK1, the mitotic Polo-like kinase 1, mediates T-DM1 resistance by overriding spindle assembly checkpoint-dependent mitotic arrest [46
]. Genomic or pharmacological inhibition of PLK1 restored T-DM1 sensitivity also by cdk1-dependent phosphorylation and by inactivation of the apoptosis guards Bcl-2/Bcl-xL. Cyclin B1, which together with cdk1 forms the so-called maturation promoting factor responsible for the switch-like commitment of cells to mitosis, has also been implicated in T-DM1 resistance. In fresh HER2-positive human breast tumor explants, induction of cyclin B1 by T-DM1 was reported to correlate with apoptosis induction. Also, in breast cancer cell lines made resistant to T-DM1 in vitro, the ADC failed to induce cyclin B1 [47
]. Finally, while in vitro tubulin gene mutations [48
] and down-regulation of βIII and βV-tubulins [49
] in tumor cells have been shown to play a role in acquired resistance to taxanes, there is currently no preclinical or clinical evidence for their relevance for ADC resistance. Nevertheless, it has become clear that under T-DM1-induced cellular stress, cancer cells can develop a multitude of resistance phenotypes by reducing the effectiveness of various steps in the cascade of events that lies between ADC uptake and cell death induction.
2.4. Inotuzumab Ozogamicin
In August 2017, Inotuzumab ozogamicin (INO, Besponsa™) gained FDA approval for the treatment of adults with relapsed or refractory B-cell precursor acute lymphoblastic leukemia (ALL). INO uses an IgG4 antibody directed against the B-lymphocyte cell adhesion molecule CD22 as the targeting moiety and a DNA double-strand break inducing calicheamicin analog as payload. As with GO, an acid-labile hydrazone linkage releases the payload in late endosomes/lysosomes from the antibody moiety. Over 90% of ALL patients express CD22 on leukemic blasts [50
]. CD22 is constitutively endocytosed and degraded [52
], which makes it an ideal target for payload delivery in general. Currently, there are four more CD22-based ADCs in various stages of preclinical and clinical development for different blood cancer indications [53
]. The first FDA-approved immunotoxin, Lumoxiti™ from Astra-Zeneca, which achieved 75% objective response rateand 30% durable complete responses in hairy cell leukemia, also targets CD22.
In the phase 3 INO-VATE ALL trial, treatment with INO outperformed standard intensive chemotherapy of ALL patients with regards to complete remission rate (80.7% versus 29.4%, respectively) and OS (median 7.7 versus 6.7 months) as endpoints [57
]. Veno-occlusive liver disease occurred as a serious adverse effect of INO in 11% of treated patients. Although clinical responses with INO therapy have not been durable, there are otherwise limited treatment option for this aggressive disease, and the time gained may be enough to bridge patients to stem cell transplants [7
]. So far, resistance to INO has been primarily reported to occur by the up-regulation of drug transporters that efflux the released payload from the cytosol before it can translocate to the nucleus and damage DNA. In cell lines and primary patient samples, the cytotoxic potency of INO was inversely correlated to the expression levels of P-glycoprotein [58