5-Azacytidine prevents relapse and produces long-term complete remissions in leukemia xenografts treated with Moxetumomab pasudotox

Significance

Moxetumomab pasudotox is a fusion protein of an anti-CD22 Fv and Pseudomonas exotoxin. It is highly active against leukemia in vitro but acute lymphoblastic leukemia (ALL) patients often are resistant. Studies with cultured cells showed resistance is caused by reduced diphthamide, the intracellular target of Pseudomonas exotoxin, but diphthamide is not reduced in most cells from most ALL patients. To study how resistance develops in animals, we injected ALL cells into mice and found that resistant cells occur in discrete bone marrow niches and contain major chromosomal and transcriptional changes. Mice pretreated with 5-azacytidine show greatly improved responses, supporting a trial of the combination in leukemia patients.

Abstract

Moxetumomab pasudotox (Moxe) is a chimeric protein composed of an anti-CD22 Fv fused to a portion of Pseudomonas exotoxin A and kills CD22-expressing leukemia cells. It is very active in hairy-cell leukemia, but many children with relapsed/refractory acute lymphoblastic leukemia (ALL) either respond transiently or are initially resistant. Resistance to Moxe in cultured cells is due to low expression of diphthamide genes (DPH), but only two of six ALL blast samples from resistant patients had low DPH expression. To develop a more clinically relevant model of resistance, we treated NSG mice bearing KOPN-8 or Reh cells with Moxe. More than 99.9% of the cancer cells were killed by Moxe, but relapse occurred from discrete bone marrow sites. The resistant cells would no longer grow in cell culture and showed major chromosomal changes and changes in phenotype with greatly decreased CD22. RNA deep sequencing of resistant KOPN-8 blasts revealed global changes in gene expression, indicating dedifferentiation toward less-mature B cell precursors, and showed an up-regulation of myeloid genes. When Moxe was combined with 5-azacytidine, resistance was prevented and survival increased to over 5 months in the KOPN-8 model and greatly improved in the Reh model. We conclude that Moxe resistance in mice is due to a new mechanism that could not be observed using cultured cells and is prevented by treatment with 5-azacytidine.

Treatment of childhood acute lymphoblastic leukemia (ALL) results in 80% long-term remissions (1). Treating relapsed ALL is a challenge and the likelihood of achieving a long-term remission decreases dramatically with each relapse (2). Consequently, novel therapies are needed to treat relapsed ALL more successfully (3). Emerging new treatment concepts include engineered chimeric antigen receptor (CAR) T cells (4–6), CD19-targeting BiTEs (7–9), B cell receptor (BCR)-signaling inhibitors (10), or antibody drug conjugates (3, 11).

The recombinant immunotoxin Moxetumomab pasudotox (Moxe) consists of a CD22-targeting antibody fragment and the Pseudomonas exotoxin A (12). After binding to CD22, Moxe is internalized, and is transported to the cytosol where it ADP ribosylates elongation factor 2 at a unique diphthamide residue (13). The ADP ribosylation leads to protein synthesis arrest and cell death. The diphthamide moiety (DPH) is generated by seven enzymes (DPH1 through DPH7). The homozygous deletion of the genes for DPH2, -4, or -5 (14) and DPH7/WDR85 (15) or a substantial decrease of DPH1 (16) or DPH4 (17) protein renders cells resistant to immunotoxins in vitro.

Moxe is highly active in 85% of patients with relapsed/refractory hairy-cell leukemia (HCL) (12) and in ∼30% of relapsed/refractory childhood ALL (18). Even-though CD22 is uniformly expressed (19), some HCL and two-thirds of ALL patients do not respond to Moxe. To study resistance to Moxe, three leukemia cell lines were treated in vitro until a resistant clone grew out (15–17). We found that resistant CA46 cells had a small chromosomal deletion containing WDR85; the deletion was preexisting in the population of cells. Resistance in KOPN-8 and HAL-01 was due to promoter silencing of DPH1 and DPH4, respectively. The resistance was associated with methylation of CpG islands in the specific gene promoters, which was partially reversed by treating the cells with the demethylating agent 5-azacytidine (5-AZA).

To identify mechanisms of resistance to Moxe in patients, we studied primary ALL blasts from children with CD22⁺ pre–B-ALL treated on the pediatric phase I trial (NCT01891981) and found resistance in most patients is not associated with low diphthamide gene expression. To understand Moxe-resistance in vivo, we then developed two ALL mouse models. In mice, we found resistant leukemia cells develop in discrete bone marrow (BM) niches and contain major chromosomal and transcriptional changes. Because these cells do not grow in culture when removed from mice, this mechanism could not have been discovered using cultured cells.

Results

Reduced DPH-Enzyme Expression Correlates with Resistance in Selected Patients.

We previously found that CD22-expressing leukemia cell lines that were repeatedly treated with Moxe in vitro develop resistance caused by the loss or a substantial reduction of one of the enzymes required for the diphthamide synthesis (15–17). To study whether primarily Moxe-resistant ALL blasts from patients are resistant due to low DPH expression, we analyzed their expression by RNA deep sequencing in eight primary patient samples (Fig. 1). Compared with the average expression in seven Moxe-sensitive cell lines, two of six nonresponders showed a two- to eightfold reduced expression of DPH4 (Fig. 1A). However, the BM-blasts of Moxe-sensitive patients also showed reduced DPH gene expression. To assess whether the decrease is functionally relevant, we compared the average DPH-expression of the sensitive cell lines to the previously established resistant cell clones (Fig. 1B). A 17-fold reduction of DPH1 in KOPN-8-R (16) and a 10-fold reduction of DPH4 in HAL-01-R (17) are sufficient to cause Moxe-resistance in vitro. The data indicate that for two patients, Moxe-resistance may have been due to reduced DPH4 expression. However, unlike as has been suggested by previous cell line studies, reduced DPH enzymes were likely not involved in the Moxe resistance of the other patients. Due to a lack of patient material, these data were not validated by an independent method.

Fig. 1.

Decreased DPH-gene expression correlates with resistance of one-third of the tested ALL samples. The reads for the indicated DPH enzymes were quantified by RNA deep sequencing of a total of seven Moxe-sensitive cell lines, normalized to reads per kilobase and million mapped reads, averaged, and set to 1. (A) The reads of the DPH-enzymes of six Moxe-resistant and two Moxe-sensitive patients were similarly determined by RNA deep sequencing and then normalized to the average of the seven sensitive cell lines. Nonresponder 6 was determined in duplicates all other patient reads were measured only once. (B) To determine the relative expression levels of the previously published resistant cell lines (16, 17), the DPH-expression was determined for their respective Moxe-sensitive cell line and multiplied by the changes in DPH for the resistant subclones, as determined by RT-PCR. The RT-PCR changes were determined in triplicates (16, 17).

KOPN-8 Relapse at Distinct BM-Sites Predates Systemic Relapse.

To investigate resistance in an animal setting, we developed a NSG (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) mouse model to study resistance in vivo, which we hypothesized would more closely resemble resistance in patients. NSG mice injected intravenously with 5 million KOPN-8 cells die from systemic leukemia after a median of 20 d. When treated with three doses of 0.4 mg/kg Moxe every other day from day 8, mice survived much longer but died from ALL after 40 d (Fig. 2A). The regrowing KOPN-8 cells are detected in the murine BM before they are detected in other organs (20). To assess response to Moxe, we monitored BM-disease burden after treatment (Fig. 2B). BM-infiltration on day 8 was 4%. On day 16, 4 d after the last dose of Moxe, no leukemia was detected in six of eight mice; in the two remaining mice, 1 in 100,000 events was KOPN-8. However, 10 d later, all mice relapsed with an average BM-infiltration rate of 0.05%, which rapidly rose to 43% on day 34.

Fig. 2.

Moxe treatment clears the BM from ALL but disease relapses from discrete BM-sites. (A) Kaplan–Meyer survival curve for mice treated with three doses of 0.4 mg/kg Moxe every other day from day 8 or vehicle. The P value was determined by log-rank test. (B) Rate of BM infiltration for KOPN-8 bearing mice treated with the schedule in A and killed at indicated times. Symbols show individual mice, lines indicate means, error as SD, significance determined by unpaired t test, with P value as ns = not significant, ***P < 0.001. (C) Mice were injected with KOPN-8L and treated with three doses of 0.4 mg/kg Moxe every other day or vehicle from day 8. Moxe-treatment was repeated from day 29 (red arrows). Bioluminescence signal was determined at treatment start (day 8) and at indicated days. All measurements were done at identical camera settings. Pictures shown are scaled from a radiance of 100–30,000. (D–F) Mouse 2 was killed, the right leg excised, formalin-fixed, paraffin-embedded, and stained for a human-specific mitochondrial marker (Mito-A/B, peroxidase signal). Images were digitalized using 40-fold optical magnification.

To visualize where the relapsed cells are located, we generated a KOPN-8 subclone (KOPN-8L) expressing GFP and luciferase (Fig. S1). This clone closely resembles the parental KOPN-8 in growth kinetics and sensitivity to Moxe. Eight days after intravenous injection with KOPN-8L cells, mice showed bioluminescence over the spine, hip, and calvarium (Fig. 2C). Vehicle-treated mice showed a rapid increase in bioluminescence correlating with disease progression. In Moxe-treated mice, there was no bioluminescence detected 4 d after the last dose of Moxe. On day 29, relapse occurred as evidenced by recurrence of the bioluminescence in both mice. The mice were retreated with a second cycle of Moxe. The bioluminescence fell, but persisted over the right tibia in mouse 2, indicating local persistence of ALL cells. Mouse 2 was killed, the leg paraffin-embedded, and stained with a human-specific mitochondrial marker to identify human ALL. Dense ALL infiltration was detected in the distal tibia (Fig. 2 D and E), but no ALL signal was present in the proximal femur or in the surrounding soft tissue (Fig. 2 D and F). Immunohistochemistry confirmed local ALL persistence in the BM of the distal tibia. These results show that the majority of KOPN-8L cells die after initial Moxe treatment, but a few cells persist at an isolated BM site.

We then treated a group of KOPN-8L–bearing mice with two (Fig. 3A) or three (Fig. 3B) cycles of Moxe. The major signal on day 8 was found in close association with bone. Three doses of Moxe abrogated the bioluminescence, but all Moxe-treated mice relapsed. The bioluminescence in relapse appeared over discrete sites in bone, which were randomly distributed over the skeleton. When the bioluminescence after the first cycle reached levels similar to day 8, mice received a second cycle of five doses Moxe every other day. The systemic bioluminescence fell below the detection limit, but the signal over the local sites persisted, and then spread to produce a systemic bioluminescence in relapse. Mice not receiving a third cycle died from progressive disease. Of the three mice receiving a third cycle, two mice responded well but the systemic bioluminescence of the third mouse increased, indicating a generalized Moxe-resistance after the second cycle. These resistant cells were termed KOPN-8L-R.

Fig. 3.

Leukemia relapse stays sensitive to retreatment for several cycles. Mice bearing KOPN-8L cells were treated with two (A), or three (B) cycles of Moxe from days 8, 38, and 57, as indicated by the arrows. One cycle consisted of three doses of 0.4 mg/kg Moxe every other day. Bioluminescence was measured at indicated days after leukemia inoculation. All measurements were done at identical camera settings. Pictures shown are scaled from a radiance of 100–30,000. Shown are representative images of four independent experiments with similar results. (C) RehL-bearing mice were treated with three cycles of Moxe from days 6, 27, and 41. Mice were imaged at indicated days. All measurements were done at identical camera settings. Pictures shown are scaled from a radiance of 200–30,000 with the exception of days 48 and 59, which were scaled from a radiance of 1,000–30,000 to reduce background.

To test whether localized resistance occurred in a second animal model, we generated a GFP/luficerase-expressing RehL subclone (Fig. S2). The Reh cell line was chosen, because they are equally sensitive to Moxe as KOPN-8 (20), but distinct from the KMT2A-ENL translocated KOPN-8, the Reh cells harbor a TEL-AML1 translocation. Mice bearing RehL where treated with repeated cycles of Moxe (Fig. 3C). Vehicle-treated mice progressed as expected. The bioluminescence of RehL-bearing micetreated with Moxe fell after the first treatment cycle but the signal persisted at isolated spots over bones. If not treated again, the Reh-bearing mice developed hind-limb paralysis and splenomegaly and died of progressive systemic leukemia (Fig. S3). Mice receiving a second cycle of Moxe showed stabilization of the bioluminescence and bioluminescence increased 2 d after the third Moxe cycle. The RehL cells after the third Moxe cycle were considered Moxe-resistant. The two xenografts show that although Moxe is very active initially, a resistant subclone develops and persists at local BM sites, where it survives treatment until a systemically resistant clone emerges after several cycles of treatment.

Leukemia Cells That Acquired Resistance Showed Profound Changes in Phenotype.

We attempted to grow the resistant KOPN-8L-R cells from mice in cell culture to facilitate their study. But the KOPN-8L-R cells would not grow, indicating that they had undergone major changes in the mice and suggesting that they had become dependent on the BM environment. Because in vitro generated Moxe-resistant cells show decreased expression of DPH enzymes, we assessed changes in mRNA levels by real-time PCR (Fig. 4A). The KOPN-8L-R had no decrease in any of the DPH-enzyme mRNAs; however, the CD22 mRNA was more than fourfold reduced. Flow cytometry confirmed the reduction of cell surface CD22 and of other B cell markers (Fig. 4B). The surface CD22 was reduced 2.5-fold (P < 0.0001), CD19 was reduced 2.1-fold (P = 0.0001), and CD10 was reduced 80-fold (P < 0.0001); surface CD28, however, increased 3.7-fold (P = 0.001). No CD22⁻ population was detected (Fig. S4). The total CD22 protein by Western blot in FACS-sorted cells from five resistant and sensitive mice was reduced sevenfold (Fig. 4C). Because surface CD22 was reduced only twofold, these data suggest a profound depletion of intracellular CD22 needed for replenishing cell surface CD22 (21).

Fig. 4.

Moxe-resistant relapse shows major phenotype and chromosome changes. KOPN-8L and KOPN-8L-R were engrafted in five mice each, their BM sorted for ALL cells, which were split to produce total RNA and total protein. (A) mRNA levels by real-time PCR in KOPN-8L-R from BM or control KOPN-8-R from cell culture (CC) (16) relative to KOPN-8L (dotted line). Bars indicate average fold-changes of CT-values in triplicates, error as SD. P values as ns = not significant, *P < 0.033, **P < 0.002, ***P < 0.001, ****P < 0.0001. (B) The BM-derived ALL blasts from five independent mice bearing KOPN-8L or KOPN-8L-R were measured for the indicated markers by flow cytometry, t tests determined P values as ns = not significant, *P < 0.033, **P < 0.002, ***P < 0.001. (C) Western blot for total CD22 protein of BM-derived ALL blasts from five independent mice, densitometry values are actin normalized. (D) BM from three individual mice bearing RehL or RehL-R (from Fig. 3C) were analyzed as described in B. *P < 0.033, **P < 0.002, ***P < 0.001. (E) KOPN-8 GL (GL) cells were cultured alone or GL cells and GL-R cells were cultured with OP-9 stromal cells, treated with Moxe, viability determined by flow cytometry for the GFP⁺ cells, which was normalized to untreated control (100%) and 0% is a true 0. Each symbol represents the mean percentage of living cells in triplicates, error as SD. (F) Chromosome analysis by spectral karyotyping showing the karyotype of KOPN-8L [44,XX, add(8q), t(11;19), t(13;14), −13, −14] acquired six additional chromosomal aberrations in KOPN-8L-R [+6, +7, +8, +t(8;5) ×2, and +20]; >10 karyotypes were analyzed, shown is the representative. (G) The expression level of the fusion transcript of the KMT2A (chromosome 11) and the MLLT1 (chromosome 19) translocation was measured by RNA deep sequencing, as described in Materials and Methods. Each symbol represents the number of uniquely mapped reads of individual mice, bars represent the average, error as SD, P values were determined by unpaired t test.

We also examined the resistant RehL-R cells that emerged after mice were treated with three cycles of Moxe and found they had acquired similar changes in surface markers as KOPN-8L-R (Fig. 4D). We found the resistant cells had a 1.6-fold decrease of CD22, a 1.6-fold decrease in CD19, a 1.8-fold decrease in CD10, and a 3.6-fold increase in CD33.

KOPN-8L-R Acquired Partial Moxe Resistance.

Because the resistant KOPN-8L-R cells alone cannot grow in culture, we hypothesized that they need BM support to survive. We investigated whether BM stromal cell lines would allow the resistant cells to grow and demonstrated that the murine calvarium-derived mesenchymal stromal cell line OP-9 rescued cell growth of KOPN-8L-R cells in vitro. Comparing KOPN-8L-R and KOPN-8L cells growing with OP-9 cell support, we found KOPN-8L-R had an IC₅₀ of 3.0 ng/mL, which was 10-fold higher than the IC₅₀ of the sensitive KOPN-8L cells with or without OP-9 cells (Fig. 4E). These data show that KOPN-8L-R cells need stromal cell support to survive and that the OP-9 support itself does not change the Moxe activity against sensitive KOPN-8L cells. This suggests that the KOPN-8L-R cells acquired a partial and nonreversible resistance to Moxe.

Moxe-Resistant ALL Showed Major Chromosomal Changes.

Because it is characteristic for ALL, we investigated whether the resistant cells had developed additional chromosomal alterations and found many changes (Fig. 4F). The parental KOPN-8L cells showed the expected karyotype of 44,XX, t(11;19), t(13;14), −13, and −14. The KOPN-8L-R cells showed additional changes including trisomy of chromosomes 6, 7, 8, and 20 and two additional t(5;8)-chromosomes, resulting in a 50,XX karyotype. These chromosomal changes could be responsible for the changes in phenotype of KOPN-8L-R. In addition, we found that the Moxe-resistant RehL-R cells from three individual mice also showed major changes in their karyotype by G-band staining (Fig. S5). Compared with the Moxe-sensitive RehL, the resistant subclone of one mouse had loss of chromosomes 2 and 18, while the subclone of another had gained chromosomes 2 and 16, and a subclone of a third mouse had gain of chromosomes 2 and 3 and loss of chromosomes 18, 19, 21, and 22. The predominant resistant subclone of each mouse presented distinct karyotype changes, which were different from the untreated cells growing in mice.

Because the KMT2A-rearrangement distinguishes the KOPN-8 cell line from other ALL subtypes (22), we determined if expression of the KMT2A-fusion gene changed significantly in resistant KOPN-8 cells. Analyzing the fusion transcript expression from the RNA deep-sequencing data, we found no difference between sensitive and resistant cells (Fig. 4G). On average, there were 13 uniquely aligned reads for KOPN-8L cells and 12.8 reads for KOPN-8L-R cells (P = 0.74). The major changes in karyotype did not alter the expression level of the KMT2A-fusion mRNA.

RNA Deep Sequencing Identifies Down-Regulation of B Cell-Specific Gene Sets.

To study effects of the chromosomal and other changes in the resistant ALL cells, we analyzed our RNA deep-sequencing data on flow-sorted ALL cells from five mice bearing KOPN-8L and KOPN-8L-R. The dot-plot in Fig. 5A shows the overall distribution of the average log₂ (counts per million) for the 16,561 genes expressed in KOPN-8L and KOPN-8L-R. In accord with the flow data, there is a substantial decrease in mRNA for MME (=CD10, 16-fold) and MS4A1 (=CD20, 28-fold) and a less than threefold but significant (P < 0.001) decrease for CD19 (1.7-fold), CD22 (2.2-fold), CD79A (2.2-fold), and CD79B (2.8-fold). In addition, there was an increase in the monocyte-specific genes TLR7 (5.2-fold), TNFRSF21 (=DR6, 14.4-fold), PLEK (8.8-fold), CSF2RB (=GM-CSF-R, 1.9-fold), and FLT3 (1.6-fold), and the activation marker CD28 (>50-fold).

Fig. 5.

RNA-deep sequencing confirms profound changes in KOPN-8L-R. BM-derived cells were FACS-sorted and their RNA isolated for RNA-deep sequencing. (A) The mapped reads were counted, normalized to counts per million (cpm), log₂-transformed, and each individual gene from resistant cells (Res, x-coordinate) plotted against sensitive cells (Sens, y-coordinate). Every dot represents the average log₂ (cpm) from five independent mice. Highlighted are B cell marker genes (blue rhombi) that are down-regulated and non-B cell marker genes (red stars) that are up-regulated in KOPN-8L-Res. (B) All genes were preranked using limma’s t-statistic from most up-regulated to most down-regulated and the list was compared with MSigDb gene lists. The total number of genes and the number of overlapping genes between a pathway and the experimentally determined gene list are indicated. Every gene present in the preranked gene list and in the respective pathway is represented by a black line above the red (up-regulated) to blue bar (down-regulated), indicating the gene’s rank in the preranked list. The enrichment score (ES) (y axis) is generated by an algorithm, which increases a “running-sum statistic” when a gene is present in the preranked list or decreases when the gene is absent [green line (23)]. The enrichment score is then normalized to the gene set size, resulting in the NES (23). The NES and the false-discovery rate “q” are indicated. DN, down-regulation; PID, pathway interaction database; UP, up-regulation. (C) Shown are TFs central to the early development of B cells, common lymphoid progenitor (CLP), lymphoid-primed multipotent progenitor (LMPP), multipotent progenitor (MPP), and prepro-B cells. Indicated by dashed arrows are interactions between the TF driving development toward B cells or DCs. The fold up- (blue) and down-regulation (green) of TFs in KOPN-8L-Res are shown. Graph adapted from ref. 27.

Because smaller but concordant changes in several genes of one pathway can be biologically relevant, the changes were analyzed using gene-set enrichment analyses (GSEA) (23). The changes were ranked from most up-regulated to most down-regulated and compared with the “c2-curated pathway” gene sets and the “c7-immunological” gene sets provided by the Molecular Signatures Database (MSigDb). GSEA generates the q-value describing the chance that differences between genes in the experimental list and a curated list are false. A low q-value indicates a high probability that the GSEA result is not found by chance. In addition, GSEA provides the enrichment score, which is generated by an algorithm that increases when a gene is present in both the experimentally determined list and the MSigDb list, or decreases when it is not. The enrichment score (ES) is then normalized (NES) (23).

Fig. 5B shows four gene sets that closely resemble the differences between KOPN-8L-R and KOPN-8L. Genes that are down-regulated in large prelymphocytes compared with mature B-lymphocytes (24) are down-regulated in KOPN-8L-R (q < 0.0009, NES = −2.33). Genes defined as “BCR pathway” (25) (q = 0.004, NES = −2.07) are also down-regulated. Both sets of data suggest a change of KOPN-8L-R to less mature B-cells. Additionally, genes that are down-regulated in B-cells compared with plasmacytoid dendritic cells (pDC) (26) were up-regulated in KOPN-8L-R (q < 0.0009, NES = 2.15), indicating a change toward pDCs. In addition, genes up-regulated in B cells compared with monocytes (26) were down-regulated in KOPN-8L-R cells, indicating a change toward a monocytic expression pattern (q = 0.003, NES = −2.04). Taken together, GSEA analysis supports a dedifferentiation of KOPN-8L-R cells from pre-B cells toward less-mature B cells and cells of non-B hematological lineages.

The ingenuity pathway analysis (IPA) further supported a change toward less-mature B cells, as indicated by the decreased expression of BCR-pathway genes. IPA identified the BCR-pathway as the second-most altered canonical pathway in KOPN-8L-R compared with KOPN-8L (Fig. S6).

The transcription factors (TFs) determining early and intermediate B cell fates also changed significantly in agreement with the pathway changes (Fig. 5C) (27). We found the B cell-specific TFs TCF3 (−1.5×, P < 0.0001), FOXO1 (−3.2×, P < 0.0001), EBF1 (−1.5×, P < 0.0001), and BCL6 (−6.3×, P < 0.0001) decreased, while DC-determining TFs ID2 (2.4×, P < 0.0001), IRF4 (2.1×, P < 0.0001), and GFI1 (>50-fold, P < 0.0001) increased (Fig. 5 A and C) (28). Together, these data indicate that a significant down-regulation of several key TFs in B cell development lead to a marked down-regulation of B cell-specific genes (27), whereas the up-regulation of ID2, IRF4, and GLI1 likely result in the up-regulation of genes expressed in monocytic cells.

The Combination of 5-AZA and Moxe Prevented Local Resistance.

Moxe-resistance in vitro can be caused by methylation of promoter regions, which can be prevented by 5-AZA (16, 17). Consequently, we hypothesized that the resistance observed in mice might also be due to DNA methylation and we tested 5-AZA in our model. Considering that the methylation changes might occur early after cell implantation, we treated mice with daily doses of 5-AZA from day 3 (29) followed by five doses of 0.3 mg/kg Moxe intravenously from day 8. Untreated mice progressed rapidly (Fig. 6 A and B, Upper). The bioluminescence for mice treated with Moxe alone fell and rose again in relapse as expected. Daily treatment with 5-AZA alone slowed disease progression, as indicated by the lower bioluminescence on day 8 (P = 0.003). The bioluminescence then increased between days 8 and 19 (P = 0.06). The combination of Moxe and 5-AZA resulted in a total abrogation of the bioluminescence and none of the mice showed any luciferase signal thereafter. The changes in bioluminescence correlated with animal survival (Fig. 6B, Lower). Animal survival was marginally prolonged by 5-AZA alone, and significantly increased by Moxe alone. However, mice that received the combination treatment were healthy and disease-free on day 120, when the experiment was terminated. Pathological examination and BM flow at day 120 showed no evidence of leukemia. It appeared that these mice were cured.

Fig. 6.

The combination of 5-AZA and Moxe greatly improves responses of the KOPN-8L xenografts. Mice inoculated with KOPN-8L were treated with vehicle, five doses of 0.3 mg/kg Moxe every other day from day 8, 2 mg/kg 5-AZA from day 3, or the combination. Bioluminescence was determined before treatment start on day 8 and at indicated days. All measurements were done at identical camera settings. Pictures shown are scaled from a radiance of 100–30,000. Mice shown in A and B were inoculated with KOPN-8L and mice shown in C. with KOPN-8L-R. Symbols in B and C, Upper represent the average bioluminescence intensities per mouse in the respective treatment group; P values were determined by t test. The respective Kaplan–Meyer survival graphs are shown in the in B, Lower, for KOPN-8L mice and C for KOPN-8L-R; P values were determined by log-rank test. t test-derived P values as ns = not significant, *P < 0.033, **P < 0.01, ****P < 0.0001.

5-AZA Did Not Reverse Established Moxe-Resistance.

To determine whether 5-AZA could overcome the established resistance of KOPN-8L-R in mice, we harvested ALL cells from mouse 5, reinjected the cells in new mice, and tested the effect of Moxe, 5-AZA, or the combination (Fig. 6C, Upper). Untreated mice showed a rapid increase in bioluminescence and died on day 25. Mice treated with 5-AZA from day 3 showed a slow, but steady increase in bioluminescence (P < 0.001). The combination of 5-AZA and Moxe slowed the increase of bioluminescence, but all mice progressed after the treatment ended. The mice treated with 5-AZA survived on average 4 d longer, Moxe-treated mice 11 d longer, and mice treated with the combination 16.5 d longer than untreated mice (Fig. 6C, Lower). None of the mice showed long-term disease-free survival, as seen in the mice bearing KOPN-8L cells. These results indicate that once resistance is established it cannot be reversed with 5-AZA.

5-AZA Enhanced the Activity of Moxe in RehL Xenografts.

Mice bearing the RehL cells were treated as the mice bearing KOPN-8L. Vehicle-treated mice showed an increase in bioluminescence (Fig. 7A). In mice treated with 5-AZA, bioluminescence increased between days 8 and 20 (P = 0.02). In mice treated with Moxe alone, bioluminescence on day 20 was stable compared with day 8 (P = 0.94). Only the Moxe/5-AZA combination greatly reduced the signal intensity on day 20 in all 10 mice (P = 0.001). The signal fell below the detection limit in four mice and the mice stayed free of bioluminescence up to 50 d. Nevertheless, all mice relapsed from discrete sites closely associated with bones. Untreated mice showed a median survival of 37 d, mice treated with Moxe of 45 d, 5-AZA treated mice of 51 d, and mice receiving the combination of 67 d (Fig. 7B).

Fig. 7.

5-AZA enhances Moxe activity in the RehL xenograft. (A) Mice were inoculated with RehL cells expressing GFP/luciferase and treated with vehicle, five doses of 0.3 mg/kg Moxe every other day from day 8, 2 mg/kg 5-AZA intraperitoneally from day 3, or the combination. All measurements were done at identical camera settings. Pictures shown are scaled from a radiance of 100–30,000. (B) Kaplan–Meyer survival graph for mice in A; P values were determined by log-rank test as ns = not significant, ***P < 0.001, ****P < 0.0001.

Discussion

We have developed a mouse model of resistance of ALL cells to Moxe, to understand the patho-biologic basis for reduced clinical responsiveness. We find that the ALL cells growing in murine BM become resistant to Moxe and that the resistant cells have developed major chromosomal changes and changes in gene expression, indicating dedifferentiation toward less mature B cell precursors. These changes were not observed when Moxe-resistant cells were selected in cell culture, probably because the resistant cells from mice cannot grow in vitro.

An unexpected feature of the resistant cells growing in mice, which is more likely to recapitulate resistance in patients in comparison with in vitro cell line studies, is that resistant cells first appear at discrete locations in various bones and then spread to other locations. Treatment initially clears the disseminated cells, which then regrow, originating from discrete BM-sites and gradually become resistant. This behavior indicates that resistance develops in several steps. It is also consistent with a protective BM-niche providing local supportive signals (10, 30, 31) and inducing a therapy-resistant state (32, 33). BM niche-induced Moxe resistance is in line with findings in pediatric ALL patients treated with anti-CD22 immunotoxin, where clearance of peripheral blood blasts was observed despite persistence of BM disease (34). Because the development of resistance is prevented by treatment with 5-AZA, but 5-AZA is not effective after resistance has developed, we believe that early changes in gene-expression patterns contribute to the Moxe-resistance and possibly to the major chromosomal changes observed in cells isolated after several cycles of treatment.

Relapse of ALL after chemotherapy commonly presents with chromosomal changes resulting in substantial phenotype changes (35, 36). Relapse of ALL after targeted therapy with CD19-targeting BiTEs (37, 38) or CD19-targeting CAR-T cells (39, 40) similarly present with the loss of CD19, which is accompanied by a lineage switch to a myeloid leukemia not expressing B cell markers with additional chromosomal changes (37–40). For aggressive B cell malignancies, such chromosomal changes in relapse have been described as linear evolution out of the predominant clone at first diagnosis (41–43). Our model suggests that these chromosomal changes develop step-wise and by chance, and take place in cells that are protected by discrete BM-niches. Therapeutic pressure by repeated cycles of CD22-targeted immunotoxin gives rise to a linear subclone with substantial chromosomal and phenotypic changes.

The therapy-resistant subclones of both, the KMT2A-ENL translocated ALL cell line KOPN-8 and the TEL-AML1 translocated cell line Reh, have a marked down-regulation of B cell-specific markers, including CD22 the target of Moxe. An only twofold reduction of the surface CD22 is paralleled by a profound reduction of intracellular CD22. The surface CD22 is internalized after receptor binding and then replenished by the intracellular pool (21). Thus, the reduction of the intracellular pool of CD22 may affect the internalization of Moxe over time. In accord with our previous data showing that CD22-targeted immunotoxins are more active the longer the cells are exposed (20, 44), the reduction of intracellular CD22 suggests that the resistant cells must be exposed to Moxe longer for them to die. Because Moxe has a short 20-min half-life in mice, the drug concentration after a bolus dose falls rapidly to low levels not capable of killing the resistant cells (20, 44). Even though suggested by our data, CD22 expression levels and Moxe activity does not necessarily correlate (45) and CD22-targeted immunotoxin can be highly active against ALL that expresses little CD22 (34).

In summary, we showed in two mouse models that Moxe resistance is due to major structural changes in chromosomes and in gene expression. Systemic resistance develops after several cycles of treatment. By combining Moxe with 5-AZA, resistance and subsequent relapse is completely prevented in KOPN-8 xenografts and delayed in Reh xenografts. The marked increase of Moxe efficacy by 5-AZA makes this combination a promising approach for future clinical testing in children and adults with ALL.

Materials and Methods

Cell Lines.

KOPN-8 and Reh cells were described previously (20). Cellular identity was established by short tandem-repeat analysis. Cells were grown in RPMI-1640 with 10% FBS, 100 U penicillin, and 100 mg streptomycin. Generation of luciferase⁺ KOPN-8L and RehL cells is described in Figs. S1 and S2, respectively. The Moxe-resistant derivate KOPN-8L-Res emerged spontaneously in vivo after repeated cycles of treatment, as described in Results. The stromal cell line OP-9 (CRL-2749) was purchased from ATCC.

For coculture experiments, OP-9 cells were seeded at 3,000 cells per well in a 96-well plate on day 0. On day 1, 50,000 GL-Res or GL cells were added and 6 h later treated with Moxe at indicated concentrations. The cells were stained 48 h later with Annexin V/7-AAD and viability determined by flow cytometry.

Reagents.

Moxe was supplied by Medimmune. 5-AZA (Sigma) was dissolved in 0.9% NaCl at 2 mg/mL, sterile-filtered, and stored at −80 °C. Panobinostat (Selleck) was dissolved in 2% DMSO, 2% Tween80, 48% PEG-300, and 48% water at 5 mg/mL.

Secondary Western blot antibodies were purchased from Santa Cruz, polyclonal α-CD22 was produced in rabbits. Antibodies for flow cytometry (α-CD10-Cy5.5, α-CD19-PE, α-CD22-PE, α-CD28-APC, α-CD33-PerCP, and α-CD34-PE) were purchased from Becton Dickinson.

Animal Studies.

Animals were handled according to NIH guidelines; studies were approved by the National Cancer Institute Animal Care and Use Committee.

Five million KOPN-8L (20), 1 million KOPN-8L-Res, or 5 million RehL cells (34) were injected on day 1 via tail vein into 6- to 8-wk-old NSG mice. Moxe was given intravenously, as indicated. 5-AZA was given at 2 mg/kg and panobinostat at 5 mg/kg intraperitoneally daily from day 3 until the last dose of Moxe (29). To assess BM response, mice were killed on days indicated. For survival studies mice were followed until they showed disease progression and were killed. Disease progression for KOPN-8L-bearing mice was defined as >10% loss of highest body weight, for RehL-bearing mice as development of hind-limb paralysis or weight loss >10% of highest body weight.

BM was extracted with mortar and pestle, mesh-filtered, washed, and either viably frozen or sorted with a FACS Aria (for FSC/SSC, FSC-H/FSC-W, and GFP^pos). Six million sorted cells per mouse were split in half and lysed for extraction of total RNA using the RNEasy kit (Qiagen) or lysed in modified RIPA buffer [50 mM Tris⋅HCl pH 7.5, 0.25% Na-Deoxycholate, 1% Nonidet P-40, 150 mM NaCl, HALT protease inhibitor (Pierce)]. Total RNA was quality controlled and samples with a registrant identification number of 10 sent to the National Cancer Institute RNA deep-sequencing facility; 1 μg was reverse-transcribed and amplified with the QuantiFast SYBR Green PCR kit (Qiagen) for qRT-PCR using primers, as indicated in Table S1. The data were analyzed using the comparative C_T method (ddC_T) with SDS manager (Applied Biosystems) and normalized to the average C_T values of three housekeeping genes.

For immune-phenotyping, cells were F_c-receptor blocked, stained in PBS/1% BSA/2 mM EDTA/0.1% Na-Azide, analyzed on a FACS Calibur, and quantified using FlowJo (TreeStar).

For immunohistochemistry, murine tissue was formalin-fixed for 24 h, stored in 70% ethanol for 48 h, and paraffin-embedded; 4-μm sections were stained with anti-human Mito A/B.

RNA Deep Sequencing.

RNA libraries were prepared using the TruSeq stranded RNA protocol and sequenced on an Illumina HiSeq2500 sequencer using TruSeq V4.0. Quality was checked using FastQC, reads were trimmed using Trimmomatic, and mapped to the human genome (hg19) by STAR in “2-pass” mode (46). The reads at gene level were quantified using subread feature counts (47) and normalized to library size as counts per million. Quantile normalization and differential expression was carried out using limma-voom (48). For GSEA, the genes were preranked by t-statistic and compared with curated gene lists in “pre-ranked mode” (23). To generate the reads corresponding to the KMT2A-ENL fusion, the BOWTIE aligner for gene mapping (49) was combined with the BLAT and STAR aligner for mapping the genes to the fusion junction (50).

Patient Studies.

Primary ALL blast samples from patients with CD22⁺ pre–B-ALL treated on a phase I trial of Moxe (NCT01891981) were collected with informed consent under protocol 04-C-0102 and approved by the National Cancer Institute Institutional Review Board. Total RNA was extracted and RNA deep sequencing performed as described above.

Luciferase Imaging.

d-luciferin (75 mg/kg, VivoGlo; Promega) was injected intraperitoneally, mice anesthetized with Isoflurane, and images were taken 5 min after injection using a Xenogen IVIS-100 (Caliper).

Statistics.

Statistical analyses were performed with Graph Pad Prism v7.00 as unpaired t tests (two group comparison), as ANOVA (multiple comparison analyses), or as log-rank test (animal survival).