A Novel l-Asparaginase with low l-Glutaminase Coactivity Is Highly Efficacious against Both T- and B-cell Acute Lymphoblastic Leukemias In Vivo

Bacterial l-asparaginases are enzymes with dual activities. The predominant one, the l-asparaginase activity that gives these enzymes their name, is the ability to hydrolyze the amino acid l-asparagine (Asn) into l-aspartic acid (Asp) and ammonia. The secondary activity present in l-asparaginases is an l-glutaminase activity, which drives hydrolysis of l-glutamine (Gln) to l-glutamic acid (Glu) and ammonia. For the FDA-approved l-asparaginases [Escherichia coli (EcA) and Erwinia chrysanthemi (ErA), approved in 1978 and 2011, respectively], the l-glutaminase activity ranges from 2% to 10% of their primary l-asparaginase activity (1). The dual l-asparaginase and l-glutaminase property of EcA and ErA is expected to manifest itself in the depletion of both Asn and Gln in the patient's blood, a notion that is supported by several studies (2–5).

The anticancer effect of l-asparaginase is believed to be predominantly due to the depletion of Asn from the blood. Indeed, leukemic blasts from acute lymphoblastic leukemia (ALL) patients completely depend on scavenging Asn from the blood, as they lack or display very low levels of the asparagine synthetase (ASNS) enzyme (6–8). In contrast, the clinical importance of the l-glutaminase activity present in all FDA-approved versions of l-asparaginases is still under debate, with conflicting reports in the literature about its putative antileukemic effect (9, 10). On one hand, pharmacodynamic analyses showed that deamination of Gln is critically required for optimal Asn deamination (2) and other more recent studies indicated contribution of l-glutaminase activity to the cytotoxicity of l-asparaginase on leukemic cells (10, 11). In contrast, others have found that the l-glutaminase activity is not required for the drug's in vitro anticancer effect, as long as the ALL cells lack ASNS (9).

Common side effects in patients treated with l-asparaginases, in addition to an immune response against the bacterial enzymes, include hepatotoxicity, hyperglycemia, dyslipidemia, perturbations in blood coagulation factors, and pancreatitis (12–14). Several clinical studies have documented the Gln depletion resulting from the l-glutaminase coactivity of current l-asparaginase preparations (4, 15, 16), and suggested that the aforementioned side effects can, at least in part, be attributed to this property of the drugs. For example, a link between the l-glutaminase activity and the immunosuppressive effects of these drugs have been reported (17, 18), as well as its role in hepatotoxicity (19), which was proposed to be due to deleterious effects on Gln homeostasis (3). Likewise, Gln depletion could likely contribute significantly to the disrupted protein synthesis in the liver and spleen that is a cause of the coagulopathy aspects of drug toxicity (20). Moreover, hydrolysis of both Asn and Gln will produce ammonia as a byproduct of the reaction. However, given that Gln concentrations are much higher in the blood as compared with Asn, Gln hydrolysis will have a more profound effect on the eventual concentration of ammonia in the blood. Indeed, hyperammonemia has been observed in patients undergoing l-asparaginase treatment (16, 21–25), which has been associated with neurotoxicity.

Additional information on the putative interplay between l-glutaminase activity and drug toxicity came from at least four clinical trials of l-asparaginases. First, in the early 1980s, a clinical trial, which examined an l-asparaginase from Acinetobacter with very high l-glutaminase activity, was forced to terminate early due to central nervous system toxicity (26). Second, between 2001 and 2008, the l-asparaginase from Wolinella succinogenes, which was initially thought to be a low l-glutaminase enzyme, was evaluated clinically through a US National Cancer Institute Rapid Access to Intervention Development (NCI RAID) grant. However, the enzyme produced via this program was found to be toxic in patients and we recently showed that it actually does contain significant l-glutaminase activity (27). Third, in 2008, a phase II clinical trial examining the FDA-approved EcA in ovarian cancer patients had to be terminated early due to excessive toxicities (28). Interestingly, while weight loss was reported as one of the main drug-related toxicities in the phase II ovarian cancer study, it is also a significant l-glutaminase–related toxicity indicator in our actual preclinical study. Finally and very recently, a clinical trial of eryaspase (red blood cell encapsulated EcA) showed that, for a yet unclear reason, the encapsulation process reduced the l-glutaminase activity (i.e., increased the selectivity for Asn hydrolysis over Gln hydrolysis), a factor pointed out as an explanation for the decrease in adverse events in the eryaspase clinical trial compared with naked EcA (29). Hence, these trials support the notion that certain side effects observed in patients undergoing l-asparaginase treatment might be associated with the level of l-glutaminase activity. Therefore, reducing the l-glutaminase activity of available l-asparaginases may be advantageous to lessen toxic side effects, but for now, it is unclear whether this would be detrimental for the antileukemic efficacy of these drugs.

Previously, we engineered variants of ErA with decreased l-glutaminase activity while maintaining near wild-type l-asparaginase activity (30). Here we evaluated these novel ErA variants in vitro and in vivo for their ability to kill ALL cells, and compared them with their wild-type counterpart. It is important to appreciate the experimental complexity when comparing different l-asparaginases for their efficacy and toxicity, as in addition to the kinetic properties of the enzyme drugs, pharmacokinetics and immunogenicity (when tested in patients) play a major role in determining the outcome. To simplify the interpretation of the results, here we present the comparison of l-asparaginases that have similar l-asparaginase activities and that only differ by 1–3 residues, suggesting very similar pharmacokinetic properties, but that have vastly different l-glutaminase activity. Together, our results suggest that high l-glutaminase activity, as present in current FDA drugs, is not essential for efficient in vivo elimination of l-asparaginase–sensitive ALL cells. In addition, reduced toxicity was observed in the low l-glutaminase variants compared with the high l-glutaminase enzymes. This sets up the rationale for further evaluation of such low l-glutaminase variants, which are predicted to have fewer side effects, as alternatives to the current FDA-approved bacterial l-asparaginases for the treatment of ALL.

Enzymes used for kinetic, NMR, and cell culture studies were expressed and purified as previously reported in Nguyen and colleagues' study (30, 31) for ErA-WT, ErA-E63Q, ErA-DM, and ErA-TM; and as in Schalk and colleagues (32) for EcA-WT.

l-asparaginase and l-glutaminase activities were determined using a continuous spectroscopic enzyme–coupled assay as described previously (32, 33).

The LOUCY cell line was established from the peripheral blood of a T-cell ALL patient (34). The luciferase-positive LOUCY cell line was generated as described previously (35). The SUP-B15 cell line was established from cells harvested from the bone marrow of a Philadelphia chromosome–positive B-cell ALL patient (36). The luciferase-expressing SUP-B15 cell line was a kind gift from Dr. Michael Jensen (University of Washington School of Medicine, Seattle, WA). All cell lines were analyzed by STR (short tandem repeat) and confirmed to match 100% to corresponding STR profile data from the Global Bioresource Center ATCC. All cell lines were verified to be mycoplasma free. The Alamar Blue assay for cell viability is described in Supplementary Methods. IC₅₀ values were determined by GraphPad Prism 6.0 using sigmoidal interpolation model with 95% confidence intervals.

Nonobese diabetic/severe combined immunodeficient γ (NSG) mice (The Jackson Laboratory) were intravenously injected at 6 weeks of age with 150-μL DPBS containing 5 × 10⁶ luciferase-positive LOUCY or SUP-B15 cells. At regular time points, the bioluminescence was measured using the IVIS Lumina II imaging system (PerkinElmer). After evidence of leukemic cell engraftment, the mice were randomly divided into different groups that were administered via intraperitoneal injection at a dose of 50 IU/mouse daily for 14 days with either ErA-WT, ErA-E63Q, ErA-DM, ErA-TM or the same volume of DPBS. In another experiment, LOUCY-engrafted mice were treated with 25 IU/mouse at days 0, 2, 4, 7, 9, 11, 12, 13, and 14 via intraperitoneal injection. The bioluminescent imaging (BLI) signal was measured every two to three days as indicated in Fig. 2 and Supplementary Figs. S1, S2, and S3. During the experiment, the mice were observed and weighed every day. The ethical committee on animal welfare at University of Illinois at Chicago approved this animal experiment.

The experimental design of this study incorporated a blinded strategy where the toxicologist was provided with samples labeled as #1 and #2, without knowing the identity of the enzymes (ErA-WT or ErA-TM). In this dose escalation study, 6 animals (3 males, 3 females) per dose group were administered the enzymes intravenously at a starting dose of 40 IU/g, increasing to 80 and finally 160 IU/g. Because of a shortage of the enzymes, ErA-TM group 6 was limited to 4 animals (3 females, 1 male), and a few animals did not receive the full intended dose (one animal of ErA-WT group 3 received 136 instead of the intended 160 IU/g, one animal of the ErA-TM group 5 received 60 instead of the intended 80 IU/g). The unexpected shortage of enzymes was due to higher than expected loss during filtration through a 0.22-μm filter and the adjustment needed for bigger body weight of the mice. After enzyme administration, the animals were monitored daily and clinical signs (hunched posture, decreased activity, sunken eyes, and rough coat) were noted if observed. None of the animals died during the 4-day observation period, and all the animals were euthanized at the end of day 4.

C57BL/6 mice of 7–10 weeks old were intraperitoneally injected with two batches of 50 IU of ErA-WT or ErA-TM. Twenty-four hours after the injection, peripheral blood was collected (5 animals per group) via cardiac puncture under anesthesia (5% isoflurane in oxygen). Shortly after collection, blood was centrifuged in heparin-coated tubes (2,000 × g, 10 minutes, 4°C) for plasma preparation. Plasma l-asparaginase activity was quantified by incubating the samples with an excess amount of l-aspartic acid β-hydroxamate (AHA; Sigma-Aldrich A6508) at 37.0°C. l-Asparaginase hydrolyses AHA to l-aspartic acid and hydroxylamine, which was detected at 690 nm with a SpectraMax M3 (Molecular Devices) spectrophotometer, after condensation with 8-hydroxyquinoline (Merck 8.20261) and oxidation to indo-oxine. Detailed procedure can be found in the Supplementary Data. The ethical committee on animal welfare at Ghent University Hospital approved the experiment.

C57BL/6 mice of 7–10 weeks old were intraperitoneally injected with 50 IU of ErA-WT or ErA-TM. For the pharamacodynamic study, peripheral blood was collected at days 1, 3, 7, and 14 (5 animals per group) via cardiac puncture under anesthesia (5% isoflurane in oxygen). In addition, blood of seven untreated mice was collected to determine the baseline value (day 0). Shortly after collection, blood was centrifuged in heparin-coated tubes (2,000 × g, 10 minutes, 4°C) for plasma preparation. Plasma was diluted with equal volume of a 10% 5′-sulfosalicylic acid dihydrate solution in water and stored at −80°C for the determination of amino acid levels.

For the amino acid analysis, the plasma samples (50 μL) were deproteinized by adding 100 μL of a 10% sulfosalicylic acid solution containing 50 μmol/L internal standards mix. After vortexing, 50 μL of UPLC-grade water was added. Centrifugation occurred for 10 minutes at 9,960 × g. Derivatization of 10 μL of supernatant was performed according to the manufacturer's instructions of the AccQ-Tag kit of Waters. The 4 amino acids (asparagine, aspartic acid, glutamine, and glutamic acid) were measured on an Acquity UPLC with QDA detector of Waters and quantified based on a 5-point calibration curve. The ethical committee on animal welfare at Ghent University Hospital approved these experiments.

A xenograft of a pediatric primary human T-ALL sample was established in NSG mice. Upon establishment of disease, human leukemic cells were isolated from the spleen via Ficoll-Paque (GE Healthcare) density gradient centrifugation. Next, these cells were injected in the tail vein of 15 female NSG mice at 7 weeks of age. Each mouse received 150-μL PBS containing 1.2 × 10⁶ cells. Engraftment of the cells was followed by measuring the percentage of human CD45-positive (%huCD45⁺) cells in the blood. Upon evidence of leukemic cell engraftment, mice were randomly divided into 3 groups (day 0), and treated daily via intraperitoneal injection with 50 IU/mouse for 13 days of either ErA-WT, ErA-TM or the same volume of PBS. At day 0, 7, and 13, blood was collected via the tail vein. At day 13, all mice were sacrificed and the spleen and bone marrow were collected. The %huCD45⁺ cells in the blood, bone marrow, and spleen was analyzed by staining with a phycoerythrin-labeled antibody for human CD45 (130-098-141; Miltenyi Biotec), performing red blood cell lysis and measuring the percentage on a LSRII flow cytometer using FACSDiva software (BD Biosciences). During the experiment, mice were observed and weighed every day. The experiment was approved by the ethical committee on animal welfare at Ghent University Hospital.

Total RNA was isolated using the miRNeasy Mini Kit (Qiagen) and the RNAse-Free DNAse set (Qiagen). cDNA was synthesized with the iScript Advanced cDNA synthesis kit (Bio-Rad). The SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) was used and the PCR reactions were run on the LightCycler 480 (Roche, model LC480). Every sample was analyzed in duplicate. qBasePLUS software (Biogazelle) was used for analysis. Gene expression was normalized against 3 reference genes (GAPDH, TBP, YWHAZ). ASNS primers: (F) 5′-CCCTGCACGCCCTCTATG-3′, (R) 5′-GGATCCTGAGGTTGTTCTTCACA-3′; GAPDH primers: (F) 5′-TGCACCACCAACTGCTTAGC-3′, (R) 5′- GGCATGGACTGTGGTCATGAG-3′; TBP primers: (F) 5′- CACGAACCACGGCACTGATT -3′, (R) 5′- TTTTCTTGCTGCCAGTCTGGAC -3′ and YWHAZ primers: (F) 5′-ACTTTTGGTACATTGTGGCTTCAA-3′, (R) 5′- CCGCCAGGACAAACCAGTAT-3′.

See Supplementary Data file for details on biostatistics.

To determine whether l-asparaginase variants with low l-glutaminase activity may hold clinical potential, we investigated several l-glutaminase–deficient ErA variants [denoted as ErA-E63Q, ErA-DM (double mutant) and ErA-TM (triple mutant)] that retain most of their wild-type l-asparaginase activity (30). Comparisons of kinetic parameters between these ErA variants and the FDA-approved wild-type versions of ErA and EcA (denoted as ErA-WT and EcA-WT) are summarized in Table 1. We selected ErA-WT over EcA-WT as the backbone to develop low l-glutaminase variants because of its superior l-asparaginase activity (∼2.5-fold higher rate than EcA-WT in hydrolyzing Asn at the physiologic concentration of 50 μmol/L). However, ErA-WT also has >70-fold higher l-glutaminase activity compared with EcA-WT at the physiologic concentration of 500 μmol/L Gln. As both k_cat (i.e., the rate at saturating substrate concentration) and K_m influence the turnover of enzyme-catalyzed reactions, the k_cat/K_m ratio is often taken as a measure of enzyme efficiency, and this ratio is shown in Table 1 for the l-asparaginase and l-glutaminase activities of the examined enzymes. However, to assess the specificities of the enzymes, we also calculated the ratio between the k_cat/K_m of the l-asparaginase reaction to the k_cat/K_m of the l-glutaminase reaction (the larger the number, the higher is the specificity for the l-asparaginase reaction). From this calculation, it is clear that ErA-TM and ErA-DM are much more l-asparaginase specific (ratio = 68,750 and 4,842, respectively) compared with the original ErA-WT enzyme (ratio = 58.6), but that only ErA-TM is significantly more specific than EcA-WT (ratio = 4,625). However, this analysis based on the ratios of the l-asparaginase and l-glutaminase k_cat/K_m values may not best reflect the physiologic conditions. Therefore, we also calculated specificity ratios based on the observed rates (k_obs) at physiologic substrate concentrations (50 μmol/L for Asn, 500 μmol/L for Gln). These calculations show that our engineered ErA variants have significantly superior Asn:Gln specificity as compared with ErA-WT (Table 1). Using this calculation, even ErA-DM is about 2-fold more l-asparaginase–specific compared with EcA-WT, with ErA-TM being 47-fold more specific. The reduced rate of l-glutaminase activity for each ErA mutant was also demonstrated through measuring changes to Gln concentrations over time by NMR spectroscopy (Fig. 1A and B). Notably, while EcA-WT completely hydrolyzed Gln in approximately 45 minutes, solutions with ErA-DM and ErA-TM contained >80% of the starting Gln after 1 hour, demonstrating their exceptionally low l-glutaminase activities (Fig. 1B).

The development of ErA mutants with comparable l-asparaginase but variable l-glutaminase activity (ErA-E63Q > ErA-DM > ErA-TM) allowed us to test whether the high intrinsic l-glutaminase activity of ErA-WT is truly required for its clinical efficacy. Notwithstanding the limitations of evaluating l-asparaginase in cell culture, we first validated the antiproliferative effect of wild-type and mutant l-asparaginases in vitro on the human leukemic cell lines, LOUCY (T-ALL) and SUP-B15 (B-ALL). Results indicated that both ALL cell lines were similarly sensitive to ErA-WT, EcA-WT, and to each of the l-glutaminase–deficient ErA mutants (Table 2; Supplementary Fig. S4). As most cell lines depend on high Gln levels in culture, the slightly lower IC₅₀ values for ErA-WT and EcA-WT compared with the l-glutaminase-deficient ErA mutants is not surprising.

Given that in vitro studies cannot unambiguously clarify whether l-glutaminase activity is required for in vivo effectiveness of l-asparaginases, we subsequently used xenograft models of luciferase-positive LOUCY and SUP-B15 cells to perform in vivo drug treatment experiments. Engraftment of human leukemic cells in mice is often considered successful when the percentage of peripheral blood (PB) cells positive for the human CD45 antigen (%huCD45⁺) is ≥1%–2% (37, 38). Four weeks after NOD-scid IL2Rgamma^null mice (NSG) received cell line injections, bioluminescence imaging (BLI) flux signals corresponding to a PB %huCD45⁺ greater than 8% confirmed successful engraftment and showcased the high level of disease burden in the examined animals (Supplementary Figs. S1, S2, S3, and S5 report the calibration between BLI flux and PB %huCD45⁺). With this level of engraftment, daily drug treatment with ErA-WT (intraperitoneal injection of 50 IU/day for 14 days) was initiated. Surprisingly, this FDA-approved l-asparaginase failed to reduce tumor cell growth in vivo. Of note, ErA-WT has a half-life of only 0.65 days in humans, compared with 1.24 days for EcA-WT (39). Furthermore, half-lives of these drugs are dramatically shortened in mice (40). Thus, we hypothesized that the short half-life of ErA-WT prevented therapeutic efficacy. To evaluate whether drug instability indeed hindered the anticancer effect, we retained the N-terminal SUMO tag, which was originally incorporated to increase stability and facilitate the heterologous expression of the enzymes in E. coli. The SUMO tag has only moderate impact on the enzymatic activity of the variants (see Table 1), and dosing of the drugs according to their activity (i.e., adjusted to deliver the same IU) largely accounts for the effect of the SUMO tag on activity. Notably, treatment of the LOUCY cell–xenografted mice with this adapted SUMO-ErA-WT enzyme (50 IU/day i.p. for 12 days) resulted in a marked decrease in tumor burden (Supplementary Fig. S1), suggesting that maintaining the SUMO tag increased the stability of the enzyme with no or minimal effects toward the therapeutic enzymatic properties of ErA-WT. Therefore, this stability tag was incorporated in the wild-type and mutant ErA enzymes used in subsequent in vivo drug treatment experiments.

We next evaluated the in vivo efficacy of the low L-glutaminase variants listed in Table 1. ErA-E63Q and ErA-DM were as efficient as ErA-WT in reducing the BLI signal in mice engrafted with LOUCY (Supplementary Fig. S1) or SUP-B15 cells (Supplementary Fig. S2). For a more stringent evaluation of the requirement for the l-glutaminase coactivity for l-asparaginase efficacy against ALL in vivo, we compared ErA-TM, the variant with the lowest l-glutaminase activity, with ErA-WT, which has inherently high l-glutaminase activity. Results indicated that both enzymes are indistinguishable in their ability to rapidly reduce leukemic burden (Fig. 2A; Supplementary Figs. S3 and S6). After 14 days of consecutive treatment, both ErA-WT- and ErA-TM-treated mice displayed BLI signals diminished to background levels (Fig. 2B). In addition, analysis of huCD45⁺ cells in PB and bone marrow (BM) of ErA-WT- and ErA-TM-treated mice showed undetectable levels of leukemic cells at day 14 of treatment (Fig. 2C and D). Smaller spleen size and reduced lymphoblastic invasion of liver tissue further confirmed the comparable antileukemic efficacy of ErA-WT and ErA-TM (Fig. 2E and F).

Since the dosing regimen used for the above studies (50 IU/mouse daily) induced an abrupt decrease in the BLI signal for both ErA-WT and ErA-TM–treated groups (BLI signal below 7% at day 3 and less than 1% relative to treatment start already by day 7, Fig. 2A; Supplementary Figs. S3 and S6), we asked whether a less aggressive dosing regimen would reveal differences in efficacy between the two enzyme variants. Therefore, we conducted an additional efficacy study where the animals were treated with 25 IU/mouse (half the dose of the previous experiments). Since after the initial 6 drug treatments, which were administered on a Mon–Wed–Fri schedule, the BLI was not below background level for both groups (1%–5% relative to day 0), we added three daily drug injections at the same dose (for a total of 9 drug administrations per group). The ALL burden, quantitated by the BLI signal, was monitored during the treatment period followed by an additional 15 days after the last treatment (Fig. 2G). Even at this reduced dose, both enzymes rapidly decreased the BLI signal. Multiple t tests were calculated for every single imaging data point revealing no statistically significant difference in BLI between the two treated groups (P >0.01). The BLI signal was below 0.5% at the last day of treatment and a week later (day 22), and remained below 1% even at day 29 (15 days post last treatment).

However, the residual BLI signal indicated that the cancer was not eradicated in both treated cohorts, suggesting that further drug dosing would be needed. Whereas the ErA-TM treated group could tolerate additional doses, this was not the case for the ErA-WT–treated group, a point clearly illustrated by the condition of the shredding toys (untouched by the ErA-WT-treated group, shredded by the ErA-TM–treated group), shown both at the last day of drug treatment and 10 days later (Supplementary Fig. S7). Collectively, these results convincingly show that the ultra-low l-glutaminase ErA-TM enzyme maintains a very similar ability to combat ALL cells in vivo as compared with the high l-glutaminase ErA-WT enzyme but with an improved tolerability profile.

In addition to noting the reduced activity of the high l-glutaminase ErA-WT–treated mice, but not of the low l-glutaminase ErA-TM–treated mice, we also compared the impact of the drugs on the animals' body weight. At the end of the experiment, we observed 27% mean weight loss in ErA-WT–treated mice (Fig. 2H), which is consistent with results from previous studies (41, 42). This effect was largely mitigated in the animals treated with ErA-TM, which only experienced 10% mean weight loss. Hence, by treatment day 14, the ErA-WT group lost on average an additional 4 grams of body weight (7 versus 3) compared with ErA-TM (P <0.0001). In addition, the correlation between percentage weight loss and level of enzyme l-glutaminase activity level was reproducible across several independent experiments (Supplementary Fig. S8A and S8B). Finally, we examined the acute toxicity of ErA-WT and ErA-TM in a blinded single dose-escalation study in CD-1 mice, using a concentration range of 40 to 160 IU/g. For comparison, the dose used to treat the ALL-bearing mice was approximately 2.5 IU/g. No fatalities occurred in either group, but clear physical and behavioral signs of toxicity were observed in the ErA-WT–treated group (Table 3). These signs were virtually absent in ErA-TM–challenged mice (Table 3). Together, these data show a clear correlation between reduced l-glutaminase activity and reduced drug toxicity. As previously noted, a common side effect of l-asparaginase treatment is hepatotoxicity, which presents histopathologically as macrovesicular hepatic steatosis (43), coupled with abnormally elevated serum levels of the liver enzymes alanine aminotransferase (ALT; ref. 44) and aspartate aminotransferase (AST; ref. 45). However, blood chemistry analysis did not display a clear difference in ALT or AST levels between ErA-WT- and ErA-TM–treated mice. The absence of a difference in serum liver enzyme levels in the studied animals may not be surprising since l-asparaginase induced hepatotoxicity in humans has been linked to fatty liver disease (46), elevated BMI (12, 47), and increased age (48), whereas the mice studied did not present histologically with hepatic fat accumulation (Fig. 2F), were lean and young.

The similar anti-ALL power of ErA-WT and ErA-TM but dissimilar toxicity profile could be due to their different l-glutaminase activity, but could also be due to different pharmacokinetic properties. To investigate this possibility, mice were injected intraperitoneally with 50 IU of ErA-WT or ErA-TM from two different batches—we chose this drug dose to be consistent with the majority of our efficacy studies. The l-asparaginase activity in blood plasma samples was measured 24 hours after the initial injection (for both batches) and in addition at days 3, 7, and 14 for the second batch. We measured comparable l-asparaginase activity for both enzymes at all time points, indicating a similar in vivo stability (Fig. 3A and B). We also determined the levels of the amino acids asparagine, aspartate, glutamine, and glutamate prior to the treatment and at days 1, 3, 7, and 14 postadministration of ErA-WT or ErA-TM. Both enzymes depleted the blood asparagine, which was below detection at day 1 and 3 and then slowly recovered (Fig. 3C). Consistent with this, the levels of aspartate increased at day 1 and 3, and then decreased to the pretreatment level by day 7 (Fig. 3D).

Notably, when we examined the effect of the enzymes on glutamine, we noted that ErA-WT reduced the glutamine level from approximately 600 μmol/L pretreatment to approximately 250 μmol/L at day 1, which then promptly recovered by day 3. In contrast, ErA-TM did not reduce the glutamine levels (Fig. 3E). As expected from this, the glutamate levels increased for the ErA-WT–treated mice, but not for the ErA-TM–treated animals (Fig. 3F). These results are consistent with the predictions from the kinetic properties of these enzymes (Table 1) and the NMR experiments (Fig. 1), showing that the in vitro observed similarity in the l-asparaginase activity but differences in the l-glutaminase activity translates into the in vivo setting. Moreover, whereas the glutamine levels recovered by day 3, we would expect that the daily dosing schedule, as used in our efficacy studies, would continuously impact the glutamine levels.

The previous experiments demonstrated the anti-ALL equivalence between ErA-WT and ErA-TM in both T- and B-ALL cell lines. To further probe the potential clinical relevance of ErA-TM, we compared these enzymes using a patient-derived xenograft (PDX) model from a primary human T-ALL. After verifying leukemia engraftment (%huCD45⁺ range 1.8%–7.5%), animals (n = 5/group) were randomized into control, ErA-WT-, and ErA-TM–treated groups (50 IU/dose daily for 13 days). PB %huCD45⁺ increased dramatically in the control group (average 4% increasing to 78% at day 13; Fig. 4A). In contrast, both of the l-asparaginase–treated groups displayed very low %huCD45⁺ in peripheral blood at day 13. These data were further recapitulated when examining %huCD45⁺ in bone marrow after the animals were sacrificed at day 13 (Fig. 4B), and in spleens at that time point (Fig. 4C). Likewise, analysis of the spleen weights showed enlarged spleens for the controls but normal sized spleens for the treated animals (Fig. 4D). These data demonstrate that both ErA-WT and ErA-TM carry strong preclinical in vivo activity, while bigger cohorts might be essential to demonstrate whether or not differences in efficacy between the enzymes are present. The seemingly moderate increase in potency of ErA-WT comes with increased toxicity, as indicated by the amount of weight lost by the animals. Indeed, recapitulating the weight loss trend we observed before (Fig. 2H), mice treated with ErA-WT lost nearly twice as much weight compared with ErA-TM (30% versus 15% of starting body weight, P <0.05; Fig. 4E). As in the previous experiment with the LOUCY cell line, we could not detect differences in ALT or AST levels between ErA-WT- and ErA-TM–treated mice.

Whereas treatment of mice engrafted with the T-ALL cell line LOUCY with ErA-WT or ErA-TM resulted in complete loss of the BLI signal and in undetectable %huCD45⁺ in the peripheral blood and bone marrow (Fig. 2A–D), treatment of mice engrafted with primary T-ALL cells resulted in lowered but still detectable leukemia burden (Fig. 4A–D). One possible explanation for the difference in efficacy between the human cell lines and the PDX could be the ASNS expression level. To probe this point, we used quantitative PCR to measure the ASNS mRNA levels in the LOUCY cell line, in the patient-derived cells used in the PDX experiment, and in 4 additional T-ALL patient samples. Indeed, the ASNS mRNA in the LOUCY cells is 20 to 30-fold lower than that measured in the patients' sample (Supplementary Fig. S9). However, we note that a growing body of evidence conclusively revealed no correlation between ASNS expression and sensitivity to l-asparaginase in patients (49–52). Hence, the genetic reason(s) that make ALL susceptible to l-asparaginase therapy is/are not fully understood. On the basis of the data presented here, it appears that patients with measurable ASNS levels will have similar responsiveness but better drug tolerance to the low l-glutaminase ErA-TM variant compared with the FDA-approved ErA-WT drug.

Identifying the l-asparaginase with the best clinical properties is a challenge. Scoring l-asparaginases based solely on cell culture data, where l-glutamine is indispensable, may be misleading, due to the significant l-glutaminase activity of these enzymes. Scoring l-asparaginases based on in vivo data is much more clinically relevant but also presents several challenges. The enzyme's kinetic properties (k_cat and K_m values) for both Asn and Gln are important parameters to consider. In addition, the drug's persistence in circulation (i.e., half-life) will determine the duration of enzymatic Asn and Gln depletion, a parameter that can be fine-tuned by appropriate dose and frequency of drug administration. Ideally, one would like to combine a prolonged half-life as seen with the introduction of pegylated EcA, high l-asparaginase activity, and only low or absence of l-glutaminase activity, the latter predicted to be in part responsible for acute toxicities of the drug. Currently, a pegylated version of ErA-WT is being evaluated (53). While such a version is predicted to solve the problem of short in vivo persistence of native ErA, we caution that with the concomitant longer time for Asn depletion, such an enzyme will also have a longer duration of Gln depletion. Considering the very high l-glutaminase activity of ErA-WT, this would predict increased side effects.

Here we present ErA variants with significantly lower l-glutaminase but comparable l-asparaginase activities relative to ErA-WT and extended circulation time achieved by maintaining the SUMO tag. We demonstrated that our engineered low l-glutaminase Erwinia l-asparaginase variants have preserved cell killing properties, similar to ErA-WT. Comparable pharmacokinetic properties of the SUMO-tagged ErA-WT versus ErA-TM enzymes accompanied by similar Asn but remarkably different Gln depletion profiles convincingly discount the possibility that the observed anti-leukemic effect in ErA-TM was due to glutamine depletion. Moreover, the comparable serum persistence and antileukemic properties of the enzymes indicate that the significant difference in the toxicity profile is linked to the difference in impact on Gln levels, which suggests a superior tolerability for the low l-glutaminase variant ErA-TM.

Others have also recognized the potential advantages of l-asparaginases with low l-glutaminase activity. One in particular is the low l-glutaminase S121 variant of the Wolinella succinogenesl-asparaginase (WoA-S121). For additional details about this enzyme, see Supplementary Material. The ratio of the l-asparaginase to l-glutaminase rates (measured at the physiologic substrate concentrations of 50 μmol/L for Asn and 500 μmol/L for Gln) best conveys the clinically relevant substrate specificity of these drugs (a high ratio describes a more specific l-asparaginase with low l-glutaminase). We recently investigated the properties of WoA-S121 and discovered that this ratio is 62 (27), which is actually inferior to that for EcA-WT (ratio = 188), but superior to ErA-WT (ratio = 6.6; Table 1). However, this ratio for ErA-TM is approximately 9,000, showcasing the extremely high l-asparaginase preference of this variant.

In conclusion, this study convincingly shows that high l-glutaminase activity, as present in current FDA-approved l-asparaginase drugs, is not essential for efficient in vivo elimination of l-asparaginase–sensitive ALL cells. Furthermore, our results suggest a decline in in vivo toxicity when the drug's l-glutaminase activity is reduced. Since the debilitating side effects of current l-asparaginases often result in treatment cessation, an event associated with inferior event-free survival (48, 54), l-asparaginases with diminished toxicity are highly pertinent to improving ALL treatment outcome. In line with this notion, a recent pilot study (55) that evaluated an intensified l-asparaginase treatment in a population of high-risk pediatric ALL patients using pegylated EcA (the current standard of care in the USA with >20-fold higher l-glutaminase activity compared with ErA-TM) had to be aborted due to an unacceptable frequency of adverse effects. However, since the patients were receiving additional chemotherapeutic drugs, the causative agent behind the increased toxicity cannot be directly linked to the more frequent dosing of the l-asparaginase. Notably, the ultra-low l-glutaminase ErA-TM variant, as presented in this study, now provides an alternative l-asparaginase that can directly probe this point.

H.A. Nguyen is a chief scientific officer at Enzyme by Design Inc. and has ownership interest (including patents) in University of Illinois at Chicago. Y. Su has ownership interest (including patents) in Enzyme by Design, Inc. A.M. Schalk is a chief operating officer and has ownership interest (including patents) in Enzyme by Design Inc. T. Lammens provided expert testimony for Jazz Pharmaceuticals (support from Jazz Pharmaceuticals to attend the Annual ASH Conference 2015 and 2016). B. De Moerloose provided expert testimony for Jazz Pharmaceuticals (travel grant from Jazz Pharmaceuticals for ASH Annual Meeting in 2015 and in 2016). Y. Saunthararajah is a consultant/advisory board member for EpiDestiny. A. Lavie is a CEO and has ownership interest (including patents) at Enzyme by Design, Inc. No potential conflicts of interest were disclosed by the other authors.

The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

Conception and design: H.A. Nguyen, M. Caffrey, T. Lammens, V. Mondelaers, B. De Moerloose, A.V. Lyubimov, B.J. Merrill, A. Lavie

Development of methodology: H.A. Nguyen, Y. Su, J.Y. Zhang, A. Antanasijevic, D. Rondelli, S. Goossens, A. Lavie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.A. Nguyen, Y. Su, J.Y. Zhang, A. Antanasijevic, M. Caffrey, A.M. Schalk, D. Rondelli, A. Oh, D.L. Mahmud, A. Kajdacsy-Balla, S. Peirs, T. Lammens, V. Mondelaers, B. De Moerloose, S. Goossens, M.J. Schlicht, K.K. Kabirov, A.V. Lyubimov, P.V. Vlierberghe, A. Lavie

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.A. Nguyen, Y. Su, A. Antanasijevic, M. Caffrey, L. Liu, D. Rondelli, A. Oh, M.C. Bosland, S. Peirs, S. Goossens, K.K. Kabirov, A.V. Lyubimov, P.V. Vlierberghe, A. Lavie

Writing, review, and/or revision of the manuscript: H.A. Nguyen, J.Y. Zhang, A. Antanasijevic, M. Caffrey, A.M. Schalk, L. Liu, D. Rondelli, A. Oh, M.C. Bosland, A. Kajdacsy-Balla, S. Peirs, T. Lammens, V. Mondelaers, B. De Moerloose, S. Goossens, A.V. Lyubimov, Y. Saunthararajah, P.V. Vlierberghe, A. Lavie

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.A. Nguyen, Y. Su

Study supervision: H.A. Nguyen, A.V. Lyubimov, A. Lavie

A. Lavie was supported in part by NIH grant RO1 EB013685 and by Merit Review Award I01BX001919 from the United States Department of Veterans Affairs Biomedical Laboratory Research and Development Service, and by the UIC Cancer Center. P. Van Vlierberghe was supported by the Fund for Scientific Research Flanders (FWO) grant 3G0C4713, by the Belgian Foundation Against Cancer grant 365W3415W, and by Kom op tegen Kanker (Stand Up To Cancer) grant 116000000251, and by the Flemish Cancer Society research grant 365Y9115W. S. Peirs was supported by doctoral and postdoc FWO grant FWO17/PDO/111. We thank Dr. Michael Jensen (University of Washington School of Medicine) for the generous gift of the SUP-B15-luciferase cell line and Béatrice Lintermans for excellent technical assistance. The histology work was carried out by the UIC Research Resources Center's Research Histology and Tissue Imaging Core.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.