Copper depletion inhibits electron transport chain activity and nucleotide synthesis to boost leukemia therapy

Abstract

The nutrient-sparse cerebrospinal fluid (CSF) poses a major challenge to spreading cancer cells¹. Despite this challenge, leukemia cells can spread to the CSF, creating a life-threatening complication that requires aggressive treatment^2,3. Unfortunately, this treatment strategy can lead to dose-limiting neurotoxicity^4,5, and many patients ultimately relapse with central nervous system (CNS) involvement^6,7. Here we used a targeted in vivo CRISPR screen to identify targetable nutritional dependencies of systemic and CNS acute lymphoblastic leukemia (ALL). Our screen revealed copper metabolism as a targetable nutritional dependency in ALL. Copper depletion by either genetic deletion of the transporter SLC31A1 or dietary intervention significantly slowed the growth of both systemic and CNS leukemia in a murine xenograft model. Mechanistically, copper depletion compromised the proliferation of leukemia cells by inhibiting complex IV activity and nucleotide synthesis. Furthermore, dietary depletion of copper combined with the standard-of-care therapy methotrexate (MTX) to inhibit leukemia progression in cell-line and patient-derived xenograft models (PDXs). Our findings identify copper as an actionable micronutrient to disrupt nucleotide synthesis in ALL and proposes copper depletion as a way to boost leukemia therapy in the hard-to-treat CNS.

Spread of cancer into the cerebrospinal fluid (CSF) is a deadly complication in both solid and liquid tumors^1,2. In the case of Acute Lymphoblastic Leukemia (ALL), the most common childhood cancer, prophylactic therapy delivered to the CSF is necessary to prevent spread of the cancer and to achieve long-term remission⁸. However, CSF-directed treatments can become ineffective⁶ leading to relapse. In addition, CNS directed treatment contributes to the risk of long-term cognitive deficits in survivors⁵. The anti-folate drug methotrexate (MTX) is a cornerstone of both systemic and CNS ALL therapy^9,10 and restrains leukemic growth by inhibiting purine and pyrimidine synthesis^11,12; however, MTX treatment is linked to acute neurotoxicity, leukoencephalopathy, and the potential for long-term neurological complications^13–15. Even with excellent 5 and 10-year survival rates in pediatric ALL, there is still a critical need to improve survivorship. A better mechanistic understanding of MTX-induced neurotoxicity^16,17 can inform strategies to reduce MTX-induced neurotoxicity. However novel therapies that target cancer-specific vulnerabilities while sparing healthy brain tissue can further improve patient outcomes and quality of life of survivors.

The cerebrospinal fluid (CSF) is nutrient sparse compared to the blood^18,19, and yet cancer cells are able to spread into and grow in this metabolically-unfavorable environment^1,3. Cancer cells often rewire their metabolism to adapt to nutritional constraints, and in some cases these adaptations create therapeutic vulnerabilities^20,21. Motivated by the need to identify ALL-specific vulnerabilities in the CNS milieu, we asked whether nutritional adaptation of leukemia cells to the CSF creates metabolic dependencies that can be targeted therapeutically. Therefore, we undertook a targeted in vivo CRISPR screen in a xenograft mouse model of ALL. This screen identified genes involved in copper metabolism and oxidative phosphorylation (SLC31A1, SDHA, and GSR) as therapeutic vulnerabilities in ALL cells. We found that in leukemia cells, copper ions primarily support electron transport chain (ETC) activity. Using mechanistic rescue experiments and metabolite profiling, copper depletion restrains nucleotide synthesis to compromise leukemia cell proliferation. Finally, we validated the therapeutic potential of combining copper depletion with methotrexate, a core ALL therapy, to reduce systemic and CNS disease in a xenograft model of ALL.

In vivo CRISPR screen identifies copper metabolism as a dependency in ALL

For our CRISPR screen we used the SEM cell line, a human ALL cell line driven by the high-risk t(4;11) KMT2A-AFF1 translocation^22,23. This cell line readily spreads to the CNS when engrafted in immunodeficient mice²¹. Since the engraftment efficiency of cells significantly limits the single-guide RNA (sgRNA) coverage in in vivo screens²⁴, a genome-wide screen becomes practically challenging. To determine how many cells survive the engraftment process and could be represented in an in vivo CRISPR screen in our model, we first utilized a genetic barcoding approach (Extended Data Fig. 1a). This approach revealed variable recovery of barcodes across organs (Extended Data Fig. 1b-c, Supplementary Data S1), with millions of cells recovered from the leptomeninges at the experimental endpoint (Extended Data Fig. 1b) emerging from only ~196 clones (Extended Data Fig. 1c). Importantly, a significant number of barcodes remained unique to the leptomeninges (Extended Data Fig. 1d). Given the low engraftment efficiency in our model, we determined that an in vivo screen using 4 gRNAs per gene would only allow for roughly 40 genes to be screened per mouse while retaining sufficient coverage in the CNS. Therefore, to increase the total number of genes screened, we conducted four independent CRISPR screens in four independent cohorts of mice. Each CRISPR screen targeted 43 unique genes each, for a total of 172 genes. Each CRISPR library contained 172 unique sgRNAs targeting 43 metabolic genes, along with 20 identical control, intergenic sgRNAs, and 2 sgRNAs targeting essential genes (DHFR and DHODH). These shared sgRNAs enabled us to compare the performance of each library (assessing for depletion of the sgRNAs targeting essential genes) and unify our analysis by normalizing our results to the intergenic sgRNAs (Extended Data Fig. 1e). Since the number of genes to be screened was so limited, our gene list was not selected in an unbiased manner; instead, it was hand-curated and enriched for genes in metabolic pathways that were of interest to us, and that we predicted to be important for cell survival in the metabolic environment of the CSF (Supplementary Data S2).

To conduct our CRISPR screen, SEM cells stably expressing Cas9-RFP were transduced with CRISPR library lentivirus encoding for sgRNAs under a vexGFP reporter. Transduced cells were FACS-sorted and engrafted into irradiated NOD-SCID mice (see Methods). At the humane endpoint, cells were harvested by rough dissection from the leptomeninges, spleen, bone marrow and blood, and submitted for next generation sequencing (NGS) (Fig. 1a, Supplementary Data S3).

Figure 1: Targeted in vivo CRISPR-knockout screen identifies copper metabolism as an in vivo dependency in ALL.

(a) Overview of the in vivo CRISPR screen. (b) CRISPR screen results. Grey dots represent genes in screen, black dots represent intergenic guideRNAs with linear regression and 99% confidence band overlaid. Red dots highlight notable genes from the screen. Axes represent mean log(2) fold change of the MAGeCK-normalized read counts of guides for each gene, relative to the mean read counts of the intergenic guides. (c) A scheme of the linked functions of SLC31A1, GSR and SDHA. (d) Experimental overview of in vivo competitive mixing experiments to validate single genes from the screen. (e-f) Change in %GFP+ cells compared to input, from mice at clinical endpoint that received intergenic or SLC31A1-targeting sgRNA cells mixed with control mCherry-expressing cells. Maximal depletion is indicated by a dashed line at −50%. Data are normalized to the mean of the intergenic group, and are from either spleens of mice receiving cells by intravenous injection (e), or from the leptomeninges of mice receiving cells by intracisternal injection (f). Plotted data are mean ± SD, p-values from Šídák's multiple comparisons test. Each dot represents an individual mouse. (g) Kaplan-Meier survival curve of the top and bottom quartiles of patients from the MP2PRT-ALL cohort (n = 1,290 patients in total) or TARGET-ALL-P2 cohort (n = 374 patients in total) with available bone marrow mRNA-sequencing and overall survival (OS) data. Vertical dashes represent censored patients. Patients were sorted by MT1X expression in transcripts per million (TPM), and the survival data of the top (”high”) and bottom quartiles (”low”) were plotted. p-value from log-rank (Mantel-Cox) tests. HR = hazard ratio, FC = fold change, i.v. = intravenous, i.c. = intracisternal, sgRNA = single-guide RNA.

sgRNAs targeting CXCR4 were the most significantly depleted from all sites (Extended Data Fig. 1f), consistent with published data highlighting CXCR4’s role in regulating leukemia stem cell engraftment and regulating cell survival in ALL^25,26. Additionally, sgRNAs targeting the de novo purine synthesis enzymes GART, PAICS, and ATIC were also significantly depleted (Extended Data Fig. 1f), consistent with broad sensitivity of ALL cells to the nucleotide synthesis inhibitor and core ALL therapy: methotrexate (MTX). Pairwise comparisons between the spleen, bone marrow, and leptomeninges also revealed several organ-specific dependencies (Extended Data Fig. 1g).

Given the small number of genes in our screen, we looked for genes that scored significantly and are functionally linked. sgRNAs targeting the mammalian cell-surface copper transporter SLC31A1 were depleted in all sites, along with sgRNAs targeting glutathione disulfide reductase (GSR) and succinate dehydrogenase complex flavoprotein subunit A (SDHA) (Fig. 1b). Together, these genes connect the import and intracellular chaperoning of copper ions²⁷ with a critical downstream function of copper: supporting ETC activity^28–30 (Fig. 1c). SDHA may be the only ETC gene that scored in our screen because other ETC genes are essential and and likely to drop out at during in vitro transduction without supplementation of uridine and pyruvate in the media. Although loss of SDHA is functionally distinct from ETC inhibition due to copper deficiency, recent reports have implicated a role for SDHA in the regulation of purine synthesis³¹, strengthening the connection between a dependency on ETC and nucleotide synthesis in ALL. We found CSF levels of copper in NOD-SCID mice to be significantly lower than levels in the plasma (Extended Data Fig. 2a), matching what is seen in humans³², raising the possibility that copper depletion might limit availability of an already scarce micronutrient. Notably, copper levels are increased in the serum of leukemia patients^33,34 and in ALL blasts compared to normal lymphocytes³⁵. Furthermore, relapse-fated ALL cells and CNS-derived ALL cells exhibit transcriptional and functional signatures of increased oxidative phosphorylation, suggesting an opportunity to target metabolic shifts associated with treatment resistance^36,37. Our screen therefore extends on these findings to suggest that inhibition of copper metabolism and oxidative phosphorylation can inhibit the growth of ALL in vivo.

To validate the importance of copper metabolism in SEM cells, we used a competitive mixing experiment in vivo²¹ (Fig. 1d). FACS-purified SLC31A1-knockout GFP (SLC31A1-KO) (Extended Data Fig. 2a-b) or control GFP SEM cells were mixed with mCherry-expressing SEM cells, validated by flow cytometry (Supplementary Fig. 1a), and then engrafted into mice using either intravenous injection into the peripheral blood or intracisternal injection into the CSF. Intravenous injection primarily models a systemic disease in which the leukemia cells engraft in hematologic organs such as the spleen and the bone marrow and migrate to the CNS, while intracisternal injection bypasses CNS-homing and results in primary CNS leukemia (Extended Data Fig. 2c-d). At the humane endpoint, mice were euthanized and cells harvested from either the spleen or the leptomeninges for flow cytometry to determine the percentage of GFP positive cells (Supplementary Fig. 1b-c).

When introduced intravenously, SEM cells with knockout of SLC31A1 exhibited significantly reduced the growth in the spleen (Fig. 1e, Supplementary Fig. 1c) and in the leptomeninges (Extended Data Fig. 2e) compared to control cells. To assess whether this effect is due to reduced engraftment, we repeated this mixing experiment using an sgRNA driven by a doxycycline-inducible promoter and administered doxycycline in the drinking water starting at day 5 after engraftment (Extended Data Fig. 2f-g). In this system, SLC31A1-KO cells still exhibited a growth defect compared to control cells in the spleen (Extended Data Fig. 2h), suggesting that negative selection against SLC31A1-KO cells occurs beyond initial engraftment. There was a mild but not statistically significant growth defect in the CNS, which could be explained by low penetrance of doxycycline into the CNS³⁸ (Extended Data Fig. 2i). Lastly, engraftment of mixed KO and control cells directly into the CSF by intracisternal injection also demonstrated reduced growth of SLC31A1-KO cells compared to control cells, albeit to a lesser degree (Fig. 1f), suggesting that inhibition of cellular copper import impairs leukemic growth in the CSF and not just the cells’ ability to migrate to the CNS. Altogether, these findings suggest that inhibiting cellular copper import impairs the growth of ALL both in peripheral organs as well as in the CNS.

Finally, to assess whether the expression of copper metabolism genes may have prognostic value in ALL patients, we probed RNA-sequencing data from diagnostic bone marrow samples in the MP2PRT-ALL and TARGET-ALL-P2 trials on the NIH Genomic Data Commons³⁹. Neither SLC31A1 nor ATP7A (the primary copper exporter in lymphocytes)⁴⁰ mRNA levels were responsive to copper availability in cultured SEM cells (Extended Data Fig. 3a) , and only ATP7A had prognostic value (Extended Data Fig. 3b-g). Instead, we focused metallothionein 1X (MT1X), a copper-binding protein whose transcription correlates with copper availability^46–48 (Extended Data Fig. 3h). High MT1X expression is associated with worse overall survival in both the MP2PRT-ALL and TARGET-ALL-P2 datasets (Fig. 1g). Since MT1X expression could be induced by exposure to other heavy metals, we also looked at COX17, the mitochondrial copper chaperone, and found that high expression of COX17 is associated with worse survival in the MP2PRT-ALL but not the TARGET-ALL-P2 dataset (Extended Data Fig. 3i-j). Altogether, our in vivo screen, genetic perturbations, and human survival data analyses suggest that copper is a metabolic dependency of ALL in vivo.

Copper depletion inhibits leukemic proliferation through inhibition of the electron transport chain

Next, we asked what intracellular function of copper is critical to support leukemia cell proliferation. To address this question, we assessed the cellular consequences of copper depletion by either knockout of SLC31A1 or copper chelation using bathocuproinedisulfonic acid (BCS)⁴⁹ in four B-ALL (SEM, NALM6, REH) and T-ALL (MOLT16) cell lines in culture. We assessed three major intracellular functions of copper relevant to oncogenesis: antioxidant defense⁵⁰, kinase signaling (ERK1/2)²⁸, and electron transport chain (ETC) activity through complex IV assembly^28,29 (Extended Data Fig. 4a). Copper depletion in culture had minimal effect on the intracellular GSH/GSSG levels/ratio and NADP/NADPH levels/ratio in all cell lines tested (Extended Data Fig. 4b-c), arguing against decreased capacity for buffering oxidative stress. Given copper’s role in facilitating MEK1-mediated phosphorylation of ERK1/2^51,52, we expected that copper depletion should reduce phosphor-ERK1/2 levels. However, copper depletion did not yield consist effects on phospho-ERK1/2 levels across cell lines, with SEM and MOLT16 cells exhibiting increased phospho-ERK1/2 following BCS treatment and NALM6 and REH exhibiting decreased levels (Extended Data Fig. 4d-e). We cannot rule out a role for copper in antioxidant defense or MAPK signaling to support leukemia cell proliferation in vivo, especially in the nutrient-deprived CNS. However, findings from both published work and the present work suggest that this is not the case. Notably, copper’s role MEK1/2 and MAPK signaling was shown to only affect cancer growth in models that are dependent on oncogenic MAPK signaling⁵², which as far as we are aware does not apply to the four ALL cell lines we used. Moreover, we consistently found suppressed oxygen consumption following copper depletion in vitro across all four cell lines (Fig. 2a, Extended Data Fig. 4g) as well as a proliferative defect (Extended Data Fig. 4f), pointing towards inhibited respiration as the culprit behind slowed proliferation at least in culture.

Figure 2: Copper depletion inhibits leukemic proliferation through inhibition of the electron transport chain.

(a-b) Seahorse MitoStress Test on (a) SEM cells or (b) NALM6 cells treated with BCS with or without CuCl₂. (b) Blue-native (BN-PAGE) followed by western blot detection of COX1 in native complex IV in SEM cells or NALM6 cells treated with vehicle or BCS. Representative data of two biological replicates shown; an additional replicate can be found in Source Data. (c) Seahorse assay to measure OCR in permeabilized SEM or NALM6 cells following treatment with complex IV substrates. (d) Complex IV enzymatic activity in BCS-treated SEM or NALM6 measured colorimetrically by monitoring the oxidation of reduced cytochrome C over time from 5 μg of mitochondrial lysate immunocaptured on microplate wells coated with anti-Complex IV antibodies. (e) Schematic of AOX rescue of copper depletion by SLC31A1-KO or BCS treatment. (f) Relative cell number after four days of control or SLC31A1-KO SEM cells overexpressing RFP or AOX. n = 4 experimental repeats, each dot represents the mean of three technical replicates. (g) Doublings per day from days 5–11 (SEM and NALM6) or days 2–8 (MOLT16) of BCS treatment in the indicated cell lines overexpressing RFP or AOX. n = 3 biological replicates, doubling rate calculated as linear regression of growth over 2 passages. SEM cells were treated with BCS for 8 days and NALM6 cells were treated for 6 days prior to the indicated mitochondrial or Seahorse assays. For Seahorse assays, plotted data are mean±SD with 4–6 technical replicates (wells). BCS and CuCl₂ were used at 50 μM. All p-values from Šídák's multiple comparisons test. Other plotted values are mean±SD. BCS = bathocuproinedisulfonic acid, oligo = oligomycin, AA = antimycin A, rot = rotenone, FCCP = carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, perm = XF Plasma Membrane Permeabilizer, ADP = adenosine diphosphate, TMPD = N,N,N,N-tetramethyl-p-phenylenediamine, asc = ascorbate, OD = optical density.

Therefore, we turned our attention to complex IV assembly and copper’s role in the ETC. Given the respiration defect of copper depleted cells (Fig. 2a, Extended Data Fig. 4g), we further interrogated the status of the various respiratory complexes in the SEM and NALM6 cell lines. Copper depletion led to specific reductions in complex IV levels, including the monomeric, dimeric and supercomplex assemblies, as determined by Blue Native PAGE (BN-PAGE) (Fig. 2b). Western blotting for specific complex IV subunits in denaturing and reducing conditions revealed significant depletion of COX2, mild depletion of COX1 and minimal change in COX4 (Extended Data Fig. 5a, Supplementary Fig. 2). Functional assessment of complex IV activity using permeabilized-cell Seahorse and a plate-based immunosorbent biochemical assay confirmed significant loss of complex IV activity (Fig. 2c-d). Although levels of complex I, II and III remained either unchanged or increased upon copper depletion (Extended Data Fig. 5b), metabolite profiling demonstrated that copper depletion decreases NAD/NADH ratio and increases the succinate/fumarate ratio, suggesting that forward electron flow through complex I and II are ultimately inhibited by copper depletion (Extended Data Fig. 5c-e). Furthermore, permeabilized Seahorse assays showed no change in oxygen consumption following treatment with either complex I or II substrates (Extended Data Fig. 5f-g), corroborating complex I and II dysfunction resulting from copper depletion. ETC deficiency resulting from copper depletion was rescued by genetic overexpression of SLC31A1, or treatment with exogenous copper chloride (Fig. 2a, Extended Data Fig. 5h). Thus, we conclude that a major role of copper in ALL cells is to support ETC activity through complex IV assembly.

We next asked whether the observed proliferative defect is a result of copper depletion-induced ETC disruption. To address this, we rescued the activity of the ETC in a copper/complex IV-independent manner by overexpression of alternative oxidase (AOX) from Ciona intestinalis^53,54 (Fig. 2e). AOX over-expression partially rescued the proliferative defect induced by SLC31A1-KO in SEM cells (Fig. 2f) and BCS-treatment in SEM, NALM6 and MOLT16 cells without affecting the proliferative rate of control cells(Fig. 2g, Extended Data Fig. 6a-c). Furthermore, AOX over-expression rescued whole cell NAD/NADH levels as well as the NAD/NADH ratio (Extended Data Fig. 6d-e). These data indicate that a key function of copper is to support ETC activity to maintain cell proliferation in ALL cells. The incomplete rescue of growth by AOX may be due to the loss of proton motive force from complexes III and IV (leading ultimately to reduced ATP production), or due to loss of other copper-dependent intracellular processes (such as lipoylation) that are not rescued by AOX. BCS-treatment does not alter mitochondrial mass nor mitochondrial membrane potential, arguing against loss of mitochondria or mitochondrial permeabilization as a cause for cell death (Extended Data Fig. 6f). Furthermore, treatment of AOX-overexpressing cells two different porphyrin superoxide dismutase (SOD) mimetics (MnTE-2-PyP and MnTBAP)⁵⁵ did not further rescue proliferative rate (Extended Data Fig. 6g), suggesting that loss of SOD1 activity due to copper depletion is not a major contributor to the observed growth defect. Finally, overexpression of yeast NADH dehydrogenase NDI1 to bypass complex I did not rescue the proliferative rate of BCS treated cells (Extended Data Fig. 6h-i), likely because NDI1 depends on ubiquinone pools that cannot be replenished without functional complex IV. Ultimately, we conclude that copper depletion can restrain leukemia growth through inhibition of ETC activity.

Copper depletion slows leukemia proliferation through inhibition of nucleotide synthesis

It seemed unlikely that ATP depletion as a result of ETC inhibition was the sole cause for slowed growth of copper-depleted leukemia cells, because cancer cells can rely on glycolysis to produce a significant portion of ATP and biosynthetic precursors^56,57. To better understand the impact of copper depletion on leukemia cell metabolism, we performed targeted metabolite profiling of SLC31A1-KO (SEM) and BCS-treated (SEM, NALM6, MOLT16) cells. Copper depletion significantly increased levels of pyrimidine synthesis intermediates: carbamoyl aspartic acid (CAA) and dihydroorotate (DHO), purine monophosphates (IMP, AMP and GMP), and decreased levels of aspartate and pyrimidine triphosphates (CTP, UTP) (Fig. 3a-f, Extended Data Fig. 7a-f). CAA and DHO were unable to be detected in MOLT16 cells. Levels of other amino acids were not significantly changed by copper depletion (Extended Data Fig. 7g-h). Finally, low-dose treatment of SEM cells with the complex IV inhibitor ADDA-5 also decreased aspartate levels with similar effects on pyrimidine synthesis inhibition as BCS treatment (Supplementary Fig. 3), suggesting that perturbed nucleotide metabolism following copper depletion is a direct result of loss of complex IV activity.

Figure 3: Copper depletion slows leukemia cell proliferation through inhibition of nucleotide synthesis.

(a) Volcano plot of changing metabolites between in day 14 BCS-treated SEM cells. n = 3 biological replicates. (b) Aspartate, carbamoyl aspartic acid (CAA) and dihydroorotate (DHO) levels in BCS or CuCl₂ treated SEM cells. (c) Nucleotide levels in BCS or CuCl₂ treated SEM cells. (d) Volcano plot depicting day 6 metabolite changes between BCS-treated and vehicle-treated NALM6 cells. n = 3 biological replicates. (e) Aspartate, carbamoyl aspartic acid (CAA) and dihydroorotate (DHO) levels in BCS or CuCl2 treated NALM6 cells. (f) Nucleotide levels in BCS or CuCl₂ treated NALM6 cells. (g) Isotopic labelling scheme of tracing ¹⁵N-amide glutamine into pyrimidine synthesis and showing steps that can be inhibited by copper depletion. (h-i) Fractional labelling of metabolites in BCS and/or CuCl₂ treated SEM, NALM6 (h), and SEM, NALM6, and MOLT16 (i) cells pulsed for 4 hours with ¹⁵N-amide glutamine. Labelling is expressed as a fraction of total unlabeled abundances measured in a parallel experiment for (h) DHO, and (i) CTP. n = 3 biological replicates. Cells were treated with BCS for: 14 days (SEM), 6 days (NALM6) or 8 days (MOLT16). (j) Scheme showing contribution of pyruvate and uridine to metabolic rescue following copper depletion. (k) Relative cell number after four days of control or SLC31A1-KO cells treated with vehicle or combination of pyruvate and uridine. n = 5–6 experimental replicates. (l) Doublings per day from days 6–10 (SEM cells) or 5–9 (NALM6 and MOLT16 cells) of BCS treatment and either vehicle or pyruvate and uridine. n = 3 biological replicates. Plotted data are mean ± SD, p-values from: (a, d) FDR-adjusted t-tests from MetaboAnalyst, (b, e, k, l) Šídák's multiple comparisons test, or (h, i) Tukey’s multiple comparison test. Cells were treated with uridine (400 μM), pyruvate (1 mM), BCS (50 μM) and CuCl₂ (50 μM). BCS = bathocuproinedisulfonic acid, CAA = carbamoyl aspartic acid, DHO = dihydroorotate, FC = fold change, CTP = cytidine triphosphate, OAA = oxaloacetate.

We reasoned that the observed disruptions of nucleotide metabolism could be a result of three separate effects: first, inhibition of the ETC leads to defective aspartate synthesis and decreased aspartate abundance, which is critical for nucleotide synthesis (Fig. 3g)^58,59. Second, inhibition of the ETC reduces ATP production, resulting in the accumulation of purine monophosphates⁶⁰. Lastly, ETC inhibition could lead to inability of complex III to reduce ubiquinone back to ubiquinol (Fig. 3g), therefore inhibiting the activity of pyrimidine synthesis enzyme DHODH⁵⁴. This results in the observed accumulation of DHO despite no change in DHODH protein levels (Extended Data Fig. 7i). Together, these three mechanisms could explain the decreased aspartate levels, increased CAA/DHO levels, and broad perturbations in nucleotide levels.

To address whether perturbed nucleotide metabolism following copper depletion is due to ETC dysfunction, we performed metabolite profiling on copper depleted cells overexpressing AOX. AOX overexpression in the copper-depleted SEM cells was sufficient to restore DHO, aspartate, and most nucleotide levels back to normal (except for IMP, discussed below) (Extended Data Fig. 7j-l). Additionally, isotope tracing with ¹⁵N-amide labelled glutamine in BCS-treated cells revealed increased de novo synthesis and total abundance of CAA and DHO (Fig. 3h, Extended Data Fig. 8a), and significantly reduced de novo synthesis of UTP and CTP (Fig. 3i, Extended Data Fig. 8b), highlighting a substantial pyrimidine synthesis defect. We were unable to detect purine synthesis intermediates in these experiments (PRPP, AICAR nor GAR); however, we observed proportionally increased levels of m+2 IMP and GMP but not AMP. This suggests that loss of aspartate may inhibit AMP synthesis and lead to accumulation of IMP (Extended Data Fig. 8c-d). In line with this, AOX overexpression did not fully restore aspartate nor IMP levels in BCS-treated cells despite fully rescuing the NAD/NADH ratio (Extended Data Fig. 6e, 7j-l). Analysis of the mono- and di-phosphate pyrimidines (UMP, UDP and CDP) and purine triphosphates (ATP and GTP) also revealed reduced labelling from ¹⁵N-amide glutamine, suggesting a combination of synthesis and phosphorylation defects (Extended Data Fig. 8e, Supplementary Fig. 4a-c). Notably, we observed increased DHO synthesis concurrently with decreased aspartate abundance, suggesting that ETC inhibition both increases aspartate consumption through conversion into DHO in addition to reducing aspartate production. Finally, across the three cell lines we tested, we observed increased m+1 GTP (a product of purine salvage) and decreased m+3 GTP (a product of de novo purine synthesis) following copper depletion, consistent with reports that purine salvage increases upon ETC inhibition^60,61. Therefore, we conclude that copper depletion disrupts pyrimidine and purine synthesis and results in perturbations in nucleotide levels.

We also investigated the effects of another copper chelator: ammonium tetrathiomolybdate (TTM) on cellular metabolism. TTM-treated SEM cells exhibit suppressed respiration (Supplementary Fig. 5a-b), increased levels of IMP and GMP and decreased levels of aspartate and pyrimidine triphosphates (Supplementary Fig. 5c). However, this dose of TTM was rapidly toxic to cells within 3 days (Supplementary Fig. 5d), consistent with TTM’s distinct, intracellular mechanism of cytotoxicity as compared to BCS, which chelates copper extracellularly^62,63.

We next asked whether perturbed nucleotide synthesis is the culprit behind slowed proliferation of copper depleted ALL cells. To address this, we supplemented copper depleted leukemia cells with pyruvate and uridine to bolster aspartate and pyrimidine synthesis respectively^54,58,59 (Fig. 3j). Individual supplementation with either pyruvate or uridine partially rescued the growth rate of SLC31A1-KO SEM cells, while dual supplementation further rescued the growth rate of BCS-treated SEM, NALM6 and MOLT16 cells (Fig. 3k-l, Extended Data Fig. 9a). Notably, pyruvate and uridine supplementation rescued the proliferation rate of copper-depleted cells to a similar degree as AOX over-expression, suggesting that the key function of the ETC in leukemia cells is to support nucleotide synthesis. Metabolite profiling revealed that dual supplementation with pyruvate and uridine rescued the NAD/NADH ratio and partially restored aspartate and CAA and DHO levels in copper-depleted SEM cells (Extended Data Fig. 9b-d). Additionally, supplementation of BCS-treated cells with inosine and uridine (Ino/Urid) also partly rescued cell proliferation and rescued DHO and aspartate levels without changing the NAD/NADH ratio (Extended Data Fig. 9e-f). Notably, inosine and uridine supplementation did not rescue to the same extent as AOX (except in MOLT16 cells, in which the trend was reversed), suggesting that at least part of the AOX-driven rescue is dependent on NAD/NADH ratio or ATP production. Overall, these data indicate that nucleotide synthesis is a limiting factor for cell proliferation upon copper depletion.

Leukemia progression and therapy alter levels of nucleotide precursor levels in mouse and human cerebrospinal fluid

Since both nucleotide salvage and synthesis play key roles in promoting cancer growth⁶¹, we asked what are the intracellular levels of de novo synthesis intermediates and salvage precursors in CNS versus splenic leukemia cells, and how the progression of and use of systemic therapy affects the availability of salvage and synthesis precursors in the CNS microenvironment. First, we developed a method using magnet-activated cell sorting (MACS) to purify SEM cells from the CNS and spleens of mice for metabolite profiling (Fig. 4a). Comparing paired samples from the same mice, we found that SEM cells in the CNS have increased levels of purine synthesis intermediates and decreased levels of salvage precursors (Fig. 4b-c) which may suggest that cells in the CNS exhaust the substrates for nucleotide salvage and increase rates of de novo synthesis. Metabolite profiling of CSF from mice bearing CNS leukemia and control mice revealed that guanine, cytidine, cytosine and xanthine, as well as glutamine levels are significantly decreased (Fig. 4d-f). Surprisingly, thymine and thymidine levels were increased suggesting that leukemia cells might selectively scavenge nucleosides within the CSF microenvironment. Lastly, we performed targeted and untargeted metabolite profiling of serial CSF samples collected from 57 ALL patients at 5 standardized lumbar puncture timepoints within the Dana Farber-Cancer Institute (DFCI) 16–001 protocol (Fig. 4g). Asparagine levels were decreased at the F1 and F3 timepoints, consistent with treatment of the patients with systemic asparaginase (an asparagine depleting therapy) (Fig. 4h). Levels of 5-methyl tetrahydrofolate (the predominant circulating form of folate) were particularly reduced at F3, consistent with administration of methotrexate (an antifolate) during the CNS phase of treatment (Fig. 4h). Furthermore, levels of the purine salvage precursors: hypoxanthine and xanthine, as well as modified nucleosides such as N4-acetylcytidine, 1-methylguanine and 1-methyladenosine, significantly decreased from diagnosis (F0) to day 18 of induction (F1), suggesting that levels of nucleotide salvage precursors are reduced during a potential period of emergence of drug-resistant clones within the CSF (Fig. 4i, Supplementary Data S5-9).

Figure 4: Leukemia progression and therapy alter nucleotide precursor levels in mouse and human cerebrospinal fluid.

(a) Experimental scheme of SEM leukemia cell isolation from the central nervous system (CNS) and spleens of mice using magnet-activated cell sorting (MACS). (b) Volcano plot depicting metabolite changes between CNS and splenic SEM leukemia cells from mice bearing terminal disease. (c) Abundance of significantly changing nucleotide synthesis and salvage intermediates or precursors in CNS versus splenic leukemia cells. Plotted values are relative to the mean of the splenic cells. n = 6 pairs of CNS and splenic samples from 6 independent mice, with lines indicating paired samples. (d) Experimental scheme of cerebrospinal fluid (CSF) collection from mice with CNS leukemia and control mice. (e) Volcano plot depicting metabolite changes between CSF from the terminal stage of mice with CNS leukemia compared to control mice. (f) Abundance of significantly changing nucleotide salvage intermediates or precursors in CSF. Plotted values are relative to control CSF. (g) Overview of human cerebrospinal fluid (CSF) sample collection from patients in the DFCI 16–001 protocol. (h) Asparagine and 5-methyl tetrahydrofolate (5-methyl THF) levels as measured by targeted LC-MS metabolite profiling, normalized to each patient’s F0 sample. (I) Hypoxanthine and inosine levels as measured by untargeted metabolite profiling using CompoundDiscoverer. Each dot represents an individual sample. Each line indicates samples from a different patient. Error bars indicate mean ± SD. p-values from: (b, e) FDR-adjusted unpaired t-tests from Metaboanalyst, two-stage step-up FDR-adjusted (c) paired or (f) unpaired t-tests, or (h-i) Šídák's multiple comparisons test. CSF = cerebrospinal fluid, LP = lumbar puncture, THF = tetrahydrofolate, FC = fold change, DFCI = Dana Farber Cancer Institute.

Overall, these data could suggest that CNS leukemia cells exhaust salvage precursors and increase de novo synthesis. Furthermore, nucleotide precursor levels in the CSF are reduced by induction therapy or CNS leukemia progression, which might create increased reliance on de novo synthesis that could be targeted therapeutically by copper depletion.

Copper depletion combines with methotrexate to reduce leukemic growth in vivo

We then asked whether copper depletion can slow ALL progression and combine with existing ALL therapy. Several copper chelators are FDA-approved and used in clinic for treatment of Wilson’s disease and other syndromes of copper overload^64–66. Importantly, mild copper depletion is being investigated for prevention of metastasis in high-risk breast cancer cases, and is well tolerated by patients⁶⁷. To achieve significant copper depletion in our mouse model, we combined zinc acetate (ZnAc) (Galzin^™), an FDA-approved copper depleting therapy, with a copper deficient diet (CuD). After 2 weeks, mice on CuD and ZnAc in the drinking water had significantly lower plasma total copper levels compared to mice on a copper replete diet (CuR) receiving vehicle (Fig. 5a). To track leukemic progression in mice, we used luciferase-expressing SEM cells and serial bioluminescent imaging (BLI) (Fig. 5b). We found that copper depletion as a monotherapy slows the progression of SEM cell-derived, systemic leukemia (Extended Data Fig. 10a-b). To assess for potential combinatorial effects of CuD with existing anti-metabolite therapies used in ALL, we performed targeted metabolite profiling on SEM cells treated with BCS and either 6-mercaptopurine (6-MP) or methotrexate (MTX) in culture (Extended Data Fig. 10c). Dual treatment with MTX and BCS further increased AICAR levels, a marker for purine synthesis inhibition (Extended Data Fig. 10D), suggesting an additive effect of copper depletion and MTX. However, copper depletion does not change the IC50 of MTX in SEM cells in culture, arguing against a synergistic effect of the two treatments (Extended Data Fig. 10e). These data suggested an additive effect of combining copper depletion with MTX treatment to inhibit ALL growth.

Figure 5: Dietary copper depletion combines with methotrexate to reduce leukemic growth in vivo.

(a) Plasma total copper levels of mice on copper replete diet and vehicle (CuR + Vehicle) or copper depleted diet and zinc acetate (CuD + ZnAc) treatment for 2 weeks (equivalent to day 0 in scheme in (b), see Methods) as measured by inductively-coupled plasma mass spectrometry (ICP-MS). Plotted values represent mean +/− S.D. (b) Experimental setup for copper depletion and methotrexate treatment of leukemic mice. Intraperitoneal (i.p.) dosing at days 15 and 18. BLI = bioluminescent imaging. (c) Total photon flux from BLI at indicated timepoints expressed as fold change compared to day 14. Plotted values represent mean +/− SEM. n = 9, 8, 9, 6 mice per group. (d) Representative BLI images from the indicated days of mice shown in Main Figure 5c. Each row contains images of the same mouse. (e) Representative images of femur or spleen sections from CuR + MTX or CuD + MTX treated mice stained for human COX4 (green) and DAPI (blue). Scale bar is 20 micrometers. Full quantification of images is shown in Extended Data Fig 10g-h. (f) Violin plots of mean intensity values for COX4 staining per cell for mouse femurs (top) and spleens (bottom). 5 cells were quantified per image. For femurs, n = 205 and 140 total cells and for spleens, n = 205 and 155 cells respectively. Lines indicate quartiles and median. (g) Relative photon flux from BLI of mouse heads at indicated days expressed as fold change compared to day 14. Plotted values represent mean +/− SEM. n = 10 mice per group. (h) Representative BLI images from the indicated days in panel (g). Each row of BLI images is from the same mouse. (i) Experimental set up for in vivo 3-¹³-C serine tracing followed by CNS and splenic SEM leukemia cell isolation by MACS. (j) Unlabelled abundances of UTP and CTP in CNS SEM leukemia cells. Data are relative to the mean of CuR + vehicle treated mice. (k) Sum of m+1 and m+2 ¹³-C labelled fractions of ATP and GTP in CNS SEM leukemia cells and splenic SEM leukemia cells. (l) Scheme of experiment to assess effect of CuD or CuR diets with MTX treatment in patient derived xenograft (PDX) models of ALL. Mice were placed on experimental diets one day after intravenous (i.v.) engraftment of PDX cells, given intraperitoneal (i.p.) doses of MTX or vehicle at indicated timepoints, and monitored for percentage of human CD45+ cells in peripheral blood (hCD45+%) by cheek bleeding. (m) Percentage of hCD45+ cells in peripheral blood at indicated timepoints for the CBAB-62871 PDX model. Plotted values represent mean+/−SEM. n = 6, 5, 16 (10 at week 7), and 17 (11 at week 7) mice per group. Combined data from two experimental replicates is shown. (n) Percentage of hCD45+ cells in peripheral blood at week 7 for CBAB-62871. (o) Percentage of hCD45+ cells in peripheral blood at indicated timepoints in the DFAT-25991 PDX model. Plotted values represent mean +/− SEM. n = 8 mice from week 4–6 and 5 mice per group at week 7. (p) Percentage of hCD45+ cells in peripheral blood at week 7 for DFAT-25991. p-values are from (a, f, j-k, n, p) Welch’s unpaired T-tests, (c) Tukey’s multiple comparison test, (g) Šídák's multiple comparisons test, or (m, o) testing for differences in the best-fit values for the growth constant “k” in multiple logistic regressions. PDX cells were obtained from the Public Repository of Xenografts (ProXE) as a gift from Jennifer Perry in Kimberly Stegmaier’s lab.

We next assessed whether the combination of MTX and CuD could be used to treat ALL. Copper depletion increased the effect of intraperitoneal (i.p.) MTX on SEM-derived, systemic leukemia in mice (Fig. 5b-d, Extended Data Fig. 10f). Immunofluorescence staining using a human-specific antibody against COX4 demonstrated reduced COX4 levels in the femurs and spleens of CuD+MTX compared to CuR+MTX mice (Fig. 5e-f, Extended Data Fig. 10g-i), confirming a direct effect of copper depletion on ETC integrity in leukemia cells. Additionally, we asked whether copper depletion can be combined with systemic methotrexate therapy to treat CNS disease. We found that copper depletion enhanced the effect of i.p. MTX in our model of isolated CNS leukemia as measured by serial BLI (Fig. 5g-h). To assess whether nucleotide synthesis is impacted in leukemia cells from copper-depleted mice, we pulsed mice harboring late-stage systemic leukemia with i.p. 3-¹³C serine 16 hours prior to purifying the CNS and splenic leukemia cells by MACS (Fig. 5i). Analysis of the unlabeled metabolites revealed decreased levels of UTP and CTP in CNS leukemia cells but not in splenic leukemia cells when comparing CuR and CuD mice (Fig. 5j, Extended Data Fig. 10j), which could be a result of increased reliance on de novo synthesis in the CNS. Critically, levels of m+1 and m+2 ATP and GTP were significantly decreased in the CNS and splenic leukemia cells of CuD mice (Fig. 5k). Levels of m+1 labelled serine were slightly decreased in CNS and splenic cells of CuD mice, which could be a result of increased consumption or reduced uptake of the tracer (Extended Data Fig. 10k). Reduced GSH levels did not change with CuD treatment, arguing against increased oxidative stress as a result of copper depletion (Extended Data Fig. 10l). Finally, we tested the combined effects of a copper-depleted diet with MTX on two different patient derived xenograft (PDX) models (CBAB-62871 (B-ALL) and DFAT-25991 (T-ALL)) and found that the combined treatments outperformed MTX alone in inhibiting the progression of disease (Fig. 5l-p, Extended Data Fig. 10m, Supplementary Fig. 7). Notably, mice on CuD treatments did not exhibit significant weight loss compared to control mice in any of our experiments (Supplementary Fig. 8a-d). Overall, these data provide a proof-of-concept that copper depletion represents an actionable, nutritional handle to perturb ETC activity and potentiate leukemia therapy for both systemic disease as well as in the hard-to-treat CNS.

Taken together, our data suggest that copper metabolism is an actionable nutrient dependency for ALL. Using in vivo functional genomics, we elucidated copper metabolism as a dependency of ALL cells both systemically and in the hard-to-treat CNS. We have identified a mechanism by which depletion of the micronutrient copper inhibits ETC activity and disrupts nucleotide synthesis to slow leukemia cell proliferation. Systemic depletion of copper enhances the effect of the standard of care ALL drug methotrexate to inhibit ALL growth in both cell-line derived and PDX models in vivo. This may suggest an additive benefit of copper depletion concurrent with existing therapy in the clinical setting. Lastly, our data highlight a combinatorial effect of copper depletion and methotrexate on slowing the growth of leukemia cells within the CSF, suggesting that a nutritional perturbation could potentiate therapy for CNS leukemia.

Materials and Methods

Cell lines

All cell lines were tested monthly for mycoplasma by PCR. The sources of the cell lines are as follows: SEM, NALM-6: D. M. Sabatini, Massachusetts Institute of Technology (MIT); MOLT4, MOLT16, PF382, REH: A. Gutierrez, Dana-Farber Cancer Institute. Cell lines were validated using STR profiling at the DFCI Molecular Diagnostics Core prior to use. All cells were cultured at 37°C with 5% CO₂.

Cell culture experiments

Cell lines were maintained in RPMI-1640 (Genesee Scientific, 25–506) supplemented with 10% fetal bovine serum (FBS) (Sigma 12306C-500mL or GenClone 25–550) and penicillin-streptomycin (GenClone 25–512). The reagents used in cell culture experiments are as follows: sodium pyruvate (Corning 25–000-Cl or Genesee Scientific 25–775), uridine (Sigma-Aldrich U3750–25G), inosine (Sigma-Aldrich I4125–10G), bathocuproinedisulfonic acid (BCS, prepared in sterile H2O and used at a final concentration of 50 uM unless otherwise specified) (Sigma-Aldrich B1125–500MG), copper(ii) chloride (Fisher Scientific 222011–50G, used at a final concentration of 50 uM), ammonium tetrathiomolybdate (TTM, Sigma-Aldrich 323446–1G), doxycycline (Sigma-Aldrich D5207–1G), trametinib (Selleck S2673–5MG), MnTBAP chloride (VWR 89151–000), MnTE-2-PyP (chloride) (MedChem Express HY-130574–5MG), ADDA 5 (Sigma-Aldrich SML1940–5MG). MnTBAP stocks were prepared in 0.1 M NaOH; other stock solutions were prepared in PBS or DMSO (ADDA 5, tetrathiomolybdate, trametinib, drugs for Seahorse assays).

Cell doubling experiments were seeded with a starting cell density of 250,000 cells/ml and cultured for 2–3 days in the specified culture medium. After 2 days, cells were counted using hemocytometers (Millipore Sigma Z359629–1EA) and reseeded at either 250,000 cells/mL (if to be passaged again in 2 days) or 125,000 cells/mL (if to be passaged again in 3 days) in the specified condition with fresh medium. Cell doubling calculations were calculated using the equation

$$\mathrm{Cell}{\mathrm{Doublings}}_{\mathrm{Day}2}=log2\left({\mathrm{Density}}_{\mathrm{Day}2}/{\mathrm{Density}}_{\mathrm{Day}0}\right)$$

Cumulative doublings were calculated by summing the cell doublings between day 0 and the indicated day. Doubling rate for BCS-treated cells was calculated as a linear regression of at least 3 cell counts across 5–6 days.

For cell growth experiments across 4 days, cells were seeded with a starting cell density of 100,000 cells/mL and cultured for 4 days in the specified culture medium, then counted.

IC-50 dose response curves

Cells were seeded at a cell density of 100,000 cells per ml in 96-well plates with the indicated concentrations of compounds in pentaplicate. Cells were incubated for 4 days, after which an ATP-based, cell viability assay was performed (CellTiter Glo, Promega G7572). Survival curves and IC50 were analyzed by best-fit analyses of the dose-response curve using Prism.

In vitro treatments

Unless otherwise indicated, BCS and copper (ii) chloride were used at 50 uM. For pyruvate and uridine treatments: uridine was supplemented at a final concentration of 400 uM, and pyruvate at a final concentration of 1 mM. For inosine and uridine treatments, both nucleosides were supplemented at 100 uM. Doxycycline (Thermo Scientific AC446060250) was added at a final concentration of 2 ug/mL in culture to induce sgRNA expression. Trametinib dosing was 30 nM for 24 hours. MnTE-2-PyP and MnTBAP were used at final concentrations of 20 uM and 50 uM respectively.

Immunoblot and immunoprecipitation

Cell lysis was performed with RIPA buffer (Santa Cruz Biotechnology, sc-24948) following the manufacturer’s protocol. Protein concentration was measured using bicinchoninic acid (BCA) assay (Thermo Fisher Scientific, 23227). Note that samples were not boiled prior to SDS-PAGE when analyzing mitochondrial encoded proteins. Protein samples were prepared in 5X sample buffer (Boster Bio, AR1112) so that 20 μg of protein lysate per sample could be used for SDS–polyacrylamide gel electrophoresis. Samples were separated on pre-cast 4–12% or 10–20% gels (Invitrogen XP10205BOX or XP04125BOX) and transferred onto 0.45 uM PVDF (Genesee 83–636R). Blots were blocked using 5% milk in TBST (tris-buffer saline + 1:1000 Tween-20) for 1 hour at room temperature, washed 3x with TBST, then incubated in primary antibody at 4°C overnight. Immunoblotting was performed using the following primary antibodies:

For reducing and denaturing western blot:

vinculin (CST 13901; 1:2500)

Beta actin (CST 8457; 1:2000)

GAPDH (CST 2118L; 1:2000)

MT-CO1/COX1 (abcam ab14705; 1:2000)

COX4I1/COX4 (CST 4850; 1:1000)

COX2 (Proteintech 55070–1-AP; 1:5000)

DHODH (Santa Cruz E-8 sc-166348; 1:500)

SDHA (CST 5839S; 1:1000)

SLC31A1 (CST 13086S; 1:500)

Beta tubulin (CST 2128P; 1:1000)

Human Total OxPhos Cocktail (abcam ab110411; 1:1000)

VDAC1/porin (abcam ab154856; 1:1000)

Total ERK1/2 (CST 4696S; 1:1000)

Phospho-ERK1/2 (CST 4370L; 1:1000)

Primary antibodies were diluted in Signal Enhancer HIKARI Solution 1 (Nacalai USA NU00102).

For blue-native PAGE, the following antibodies were used:

MT-CO1/COX1 (ABCAM, ab14705, 1:2000)

SDHA (Santa Cruz, sc-166947,1:20,000)

NDUFS1 (Abcam, ab169540,1:1000)

UQCR4/CYC1 (Millipore Sigma, HPA001247,1:2000)

Secondary antibodies used were:

Peroxidase AffiniPure Goat Anti-Mouse IgG H+L (Jackson ImmunoResearch Laboratories, #115–035-166) and Peroxidase AffiniPure Goat Anti-Rabbit IgG H+L (Jackson ImmunoResearch Laboratories, #111–035-144) were both diluted 1:10,000 in 5% milk in TBST and incubated with immunoblots for 1 hour at room temperature.

For SLC31A1 immunoblotting from cell lysates, immunoprecipitation was performed using Protein A/G magnetic beads (Thermo Scientific 88802) without crosslinking. Buffers were prepared as follows: 20X coupling buffer (0.2 M sodium phosphate, 3.0 M sodium chloride, pH = 7.2), elution buffer (0.1 M glycine, pH = 2), neutralization buffer (1.0 M Tris, pH = 9), wash buffer (1 X Tris-buffered saline + 1:2000 Tween-20), MCLB (50 mM Tris (pH = 7.5), 300 mM NaCl, 0.5% NP40, filtered using 0.40 uM PES filters and stored at 40° C). 25 μL of beads were used per sample. Beads were washed three times with 500 μL coupling buffer, using a magnetic stand (Promega Z5342) in between washes to remove the supernatant without disturbing the beads. Then, 100 μL of coupling buffer and 10 μL of primary antibody were added per sample. Beads were incubated on an inversion mixer for 1hr at room temperature to couple the primary antibody to the beads. Then, the beads were washed twice with 500 μL coupling buffer, then twice with 500 μL MCLB and stored on ice for up to 24 hours. 5 million cells were spun down, washed once with PBS, then resuspended in 250 μL MCLB (with 1 mM DTT and Roche cOmplete^™, Mini Protease Inhibitor Cocktail added, product #11836153001). Lysed cells were incubated on an inverting mixer for 20 minutes at 4° C, then spun at 16.1 rcf for 10 minutes. The supernatant was transferred to another tube, and 50 μL of the supernatant was reserved for protein quantitation by BCA assay and for use as “input” samples for immunoblotting.

200 μL of cleared cell lysate was mixed with antibody-bound beads and incubated on an inverting mixer overnight at 4°C. Beads were then washed twice with 500 μL wash buffer, once with water, then eluted with 100 μL elution buffer for 10 minutes at room temperature on an inverting mixer. The supernatant was transferred to a new tube and 15 μL of neutralization buffer was added. The neutralized sample was then analyzed by immunoblotting.

Crude Mitochondrial Isolation

Crude mitochondrial isolates were used for blue-native PAGE and complex IV activity plate assays. To obtain mitochondrial fractions, 40 million cells were pelleted and resuspended in 800 μL hypotonic buffer (10 mM Tris, pH 7.5; 10 mM NaCl; 1.5 mM MgCl2) and left to swell on ice for 7.5 min. Cells were dounced in a 1 mL glass homogenizer with 2×25 strokes on ice, with 2 minutes rest in between. 2 M sucrose T10E20 buffer (10 mM Tris, pH 7.6; 1 mM EDTA, pH 8.0; 2 M Sucrose) was added to the lysate to bring the sucrose concentration to 250 mM and homogenized cells were centrifuged at 600g for 10 minutes to remove nuclei and unbroken cells. The supernatant was collected and spun at 10,000g to pellet mitochondria. Pellets were then washed twice with 250 mM sucrose T10E20 buffer (10 mM Tris, pH 7.6; 1 mM EDTA, pH 8.0; 250 mM Sucrose) before use. A small fraction of mitochondria were lysed in RIPA and the protein content measured by BCA assay as above to determine the total mitochondrial yield.

Complex IV enzymatic activity assays

Mitochondria were isolated from cells as described above using dounces and total quantity estimated using BCA assay. 20ug of mitochondria were lysed, and then 5 ug of lysed mitochondria were loaded per well into antibody-coated microplates according to the manufacturer’s instructions. Complex IV enzyme activity was measured colorimetrically in antibody-coated microplates by measuring the oxidation of cytochrome-c according to the manufacturer's instructions (Abcam cat# ab109909), and complex IV activity was inferred by calculating the slope of the first 70 minutes of data collection.

Blue Native PAGE

50 μg of isolated mitochondria, as described above, were resuspended in 20 μL sample buffer cocktail (5 μL 4X NativePAGE sample buffer, 8 μL of 5% digitonin, and 7 μL of water). The suspension was left on ice to solubilize for 20 min, after which the lysate was cleared with centrifugation at 20,000g for 10 min at 4°C. 2 μL of Coomassie G-250 was added to the supernatant and proteins were separated using NativePAGE 3%−12% gradient gels (Thermo Fisher Scientific). During electrophoresis, inner chambers were filled with dark blue cathode buffer (Bis-tris 50 mM, Tricine 50 mM and 0.02% Coomassie blue G250) and outside chambers were filled with running buffer (Bis-tris 50 mM). Gels were run for 30 min at 150 V, upon which the dark blue running buffer was replaced with light blue running buffer (Bistris 50 mM, Tricine 50 mM and 0.001% Coomassie blue G250) and run for an additional 90 min at 250 V. Proteins were transferred onto PVDF membranes using wet transfer methods at 200 mA for 90 min, fixed with 8% acetic acid, rinsed twice with water, and then air dried. Dried membranes were then activated in 100% methanol for 5 min, followed by methanol and water rinses to remove Coomassie staining, and then blocked using 5% skim milk in TBST. OXPHOS complexes were analyzed by immunoblotting with indicated antibodies towards each complex.

Mitochondrial membrane potential and mitochondrial mass measurement

For determination of mitochondrial membrane potential and mitochondrial mass, tetramethylrhodamine methyl ester perchlorate (TMRM, Cayman 21437) and MitoSpy Green FM (BioLegend 424805) were used. For staining, cells were resuspended in pre-warmed media at a density of 1 million cells/mL. TMRM (final concentration 500nM) and MitoSpy (final concentration 2.5uM) were added and cells were incubated at 37C in a 5% CO2 incubator for 30 minutes. Afterwards, cells were washed once in PBE and then resuspended in PBE for analysis on a flow cytometer.

Flow cytometry

Flow cytometry was performed on a BD Fortessa and flow sorting was performed on a BD FACSMelody. Briefly, for samples from cell culture or from mice, cells were washed with fluorescence-activated cell sorting (FACS) buffer (PBS + 2% FBS + 2 uM EDTA) and incubated for 30 min in the presence of the indicated antibodies and live/dead stains on ice. After staining, cells were washed once with FACS buffer and then analyzed on ice. For staining SLC31A1 by indirect staining, cells were washed with FACS buffer and then incubated for 30 min with the primary antibody (Abcam AB129067). The cells were then spun down, washed once with FACS buffer, then incubated for 30 min with PE or Alexa 647 secondary antibody (BioLegend 406414 or 406421). For cell sorting, all staining steps and the sorting procedure itself was performed at room temperature to preserve cell viability. For in vivo mixing experiments, SEM cells were distinguished from murine cells by GFP/mCherry expression as well as human CD19+ staining. The following flow cytometry antibodies were used:

For flow sorting of SLC31A1-stained cells:

Anti-SLC31A1/CTR1 antibody (Abcam AB129067))

Donkey anti-Rabbit secondary antibody (PE – BioLegend 406414, or Alexa 647 – Biolegend 406421)

For mouse in vivo mixing experiments and PDX experiments:

Anti-human CD19 – Pacific Blue (Biolegend 302223)

Anti-human CD45 – PE (BioLegend 304008)

Anti-human CD2 – FITC (BioLegend 984902)

Fixable Near IR Live-Dead Stain (Thermo L10119)

Gene expression analysis (RNA, cDNA synthesis, and quantitative polymerase chain reaction)

RNA extraction from cell pellets was performed using RNAzol (Thermo Fisher Scientific, NC0477546). The reverse transcriptase reaction was performed using 1 μg of RNA per sample (New England Biolabs, E3010). Quantitative polymerase chain reaction (qPCR) reactions were performed with SYBR green (CWBiotech, CW2621) on a QuantStudio 7 Pro (Applied Biosystems) qPCR instrument. The reference genes for all qPCR reactions were β-actin and UBC.

Primers used for qPCR are the following:

SLC31A1-forward: 5’-GTGCTAGTGGCTGGACTTGA-3’

SLC31A1-reverse: 5’-CCAAAGTAGAAGGTCATAGGCATC-3’

ATP7A-forward: 5’-CTGGAACATATAGCAAAGGGCA-3’

ATP7A-reverse: 5’- CGTTGTACAAGTTCCACATCCAC-3’

MT1X-forward: 5’- CACGCTTTTCATCTGTCCCG-3’

MT1X-reverse: 3’- CAGGAGCCAACAGGCGAG-3’

β-actin-forward: 5’-CAACCGCGAGAAGATGACCC-3’

β-actin-reverse: 5’-AGGCGTACAGGGATAGCACA-3’

UBC-forward: 5’-att tgg gtc gcg gtt ctt-3’

UBC-reverse: 5’- tgc ctt gac att ctc gat ggt-3’

Seahorse Analysis

The Seahorse XFe96 Analyzer was used to assess oxygen consumption rate (OCR). Prior to Seahorse assay, the culture plate was treated with 1X Poly-L Lysine solution (NewComerSupply 1339A) diluted in water for 15 minutes, and each well was then rinsed 3x with 1X PBS. Cells were then washed once in pre-warmed Seahorse RPMI (Agilent 103576), and 100k cells were plated per well in pentaplicates. The cell plate was then centrifuged at 350g for 1 minute with no braking to adhere the cells to the surface, and then incubated at 37°C for 1 hour in an incubator without CO₂. The standard mitochondrial stress test was performed with injection of carbonyl cyanide p-trifluoromethoxyphenylhydrazone (final concentration: 1 μM; Sigma-Aldrich, C2920), oligomycin (final: 2 μM; Sigma-Aldrich, 75351), rotenone (final: 500 nM; Sigma-Aldrich, R8875), and antimycin A (final: 500 nM; Sigma-Aldrich, A8674). Analysis was performed using Wave software (Agilent Technologies).

Permeabilized cell Seahorse assays were conducted following a published protocol⁶⁸. Briefly, 100,000 cells were seeded on poly-L-lysine coated 96 well plate and just prior to measurements, media was replaced by MAS-BSA buffer (70 mM sucrose, 220 mM Mannitol, 10 mM KH₂PO₄, 5 mM MgCl₂, 2 mM Hepes (pH 7.2), 1 mM EGTA, and 4μg/ml BSA). Three basal measurements were taken, upon which the cells were permeabilized using XF Plasma Membrane Permeabilizer (1.5 nM final concentration; Agilent) together with 1 mM ADP and the following specific respiratory complex substrates: Complex I = pyruvate/malate (5 mM/2.5 mM); Complex II = succinate (10mM); Complex IV = N,N,N,N-tetramethyl-p-phenylenediamine/ascorbate (0.5 mM/2 mM final concentrations). These were followed by injections with oligomycin (1 μg/ml final) and respective Complex inhibitors (Complex I and II = 1 μM rotenone and 20μM Antimycin A; Complex IV = 20 mM potassium azide).

Metabolite profiling by mass spectrometry

Polar metabolite detection

One and a half million cells from culture were collected via centrifugation, washed with 0.9% NaCl, and resuspended in 400 μL of ice-cold 100% LC-MS methanol (supplemented with isotopically labeled amino acid standards [Cambridge Isotope Laboratories, MSK-A2–1.2], aminopterin, and isotopically labeled reduced glutathione standard [Cambridge Isotope Laboratories, CNLM-6245–10]), that was then transferred to a microcentrifuge tube and then vortexed briefly. Next, 100 μL of a 25 mM Ammonium Acetate, 2.5 mM Na-Ascorbate and 20 mM Ellman’s reagent (5,5′-Dithiobis (2-nitrobenzoic acid), D8130, Sigma-Aldrich) solution prepared in LC-MS water was added, for a final composition of 80% methanol and 20% water. The sample was then vortexed for 10 seconds. The Ellman’s reagent solution was prepared fresh on the day of the experiment.

For CSF metabolomics, either 10uL of mouse CSF or 20 μL of human CSF was extracted with a total of 200 μL of extraction buffer (following the same steps as above, 160uL methanol with standards + 40 μL water with additives).

After extraction, samples were centrifuged for 10 minutes at 23,000 g at 4°C to pellet cell debris. The supernatant was transferred to a new microcentrifuge tube and dried on ice using a liquid nitrogen dryer. Dried samples were stored dried at −80°C until analysis on LC-MS instruments.

Dried samples were resuspended in 30 μL LC-MS water (supplemented with QReSS, Cambridge Isotope Laboratories, MSK-QRESS-KIT) by brief vortexing. Resuspended samples were spun again for 10 minutes at 23,000 g at 4°C, and the cleared supernatant was transferred to LC-MS micro vials (ThermoScientific 6ESV9–04PP) with caps (ThermoScientific 6ASC9ST1). A small amount from each sample was taken to create a pooled sample, and this pooled sample was serially diluted 3 and 10-fold for quality control injections throughout the sequence. 2 μL of each sample was injected into a ZIC-pHILIC 150 × 2.1 mm (5 μm particle size) column (EMD Millipore) operated on a Vanquish^™ Flex UHPLC system (Thermo Fisher Scientific).

Chromatographic separation was achieved using one of the three following conditions, all of which result in the same gradient during the period of sample elution off the column:

1. buffer A was acetonitrile; buffer B was 20 mM ammonium carbonate, 0.1% ammonium hydroxide in water; resulting pH is around 9 without pH adjustment. Gradient conditions used were: 0–20 min: linear gradient from 20% to 80% B; 20–24 min: hold at 80% B; 24–24.1 min: from 80% to 20% B; 24.1–32 min: hold at 20% B at 0.150 mL/min flow rate. The column oven and autosampler tray were held at 25 °C and 4 °C, respectively.

2. buffer A was 95% acetonitrile+5% buffer B from above; buffer B was 95% buffer B from above+5% acetonitrile; Gradient conditions used were: 0–20 min: linear gradient from 16.6% to 83.4% B; 20–24 min: hold at 83.4% B; 24–24.1 min: from 83.4% to 16.6% B; 24.1–32 min: hold at 16.6% B at 0.150 mL/min flow rate. The column oven and autosampler tray were held at 25 °C and 4 °C, respectively.

3. buffer A was 95% acetonitrile+5% buffer B from above; buffer B was 95% buffer B from above+5% acetonitrile; Gradient conditions used were: 0–17.5 min: linear gradient from 16.6% to 75.0% B; 17.5–18.0 min: linear gradient from 75.0% to 85.0% B; 18–20.0 min: hold at 85.0% B; 20.0–20.5 min: from 85.0% to 16.6% B; 20.5–24 min: hold at 16.6% B at 0.150 mL/min flow rate. The column oven and autosampler tray were held at 25 °C and 4 °C, respectively.

MS data acquisition was performed using a QExactive benchtop Orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe (Thermo Fisher Scientific) and polarity switching was used. For polar metabolomics, four scans were used: full scans in both positive and negative ionization mode in a range of m/z = 70–1000, with the resolution set at 70,000, the AGC target at 1 × 10⁶, and the maximum injection time (Max IT) at 20 milliseconds (ms) from 0–20 minutes. A third scan in the negative mode was used with range of m/z = 220–700 from 0–20 minutes and the same resolution, AGC settings with 30 ms Max IT. Lastly, a targeted-SIM scan was added with a resolution of 35k, AGC target 1e5, and max IT 20 ms, isolation window = 1.0 m/z, with an inclusion m/z of 503.0552 (corresponding to Ellman-derivatized glutathione). Tune file parameters were: spray voltage = 3.5 kV, capillary temperature = 320°C, S-lens RF = 50, auxiliary gas temperature = 350°C.

For analysis of folate species including 5-methyl tetrahydrofolate, the mass spectrometer was operated in full-scan, positive ionization mode using three narrow-range scans: 438 to 450 mass/charge ratio (m/z), 452 to 462 m/z, and 470 to 478 m/z, with the resolution set at 70,000, the AGC target at 10⁶, and the maximum injection time of 150 ms. Heated electrospray ionization (HESI) settings were as follows: sheath gas flow rate, 40; Auxiliary (Aux) gas flow rate, 10; sweep gas, 0; spray voltage, 2.8 (negative) and 3.5 (positive); capillary temperature, 300; S-lens radio frequency (RF) level, 50; Aux gas heater temp, 350. Levels of folates were normalized to aminopterin as an internal standard and to polar metabolites.

Metabolomics data analysis

Polar metabolites were relatively quantified while referencing an in-house library of chemical standards and using TraceFinder 4.1 (Thermo Fisher Scientific, Waltham, MA, USA), with a 5–part per million mass tolerance. Samples and fractional dilutions were prepared as quality controls and injected at the beginning and end of each run. Pooled samples were interspersed throughout the run to control for technical drift in signal quality and for coefficient of variability (CV) determination for each metabolite. Data normalizations were performed in two steps: (i) Integrated peak area signal from internal standards added to extraction buffers were mean-centered (for every standard, peak area was divided by the mean peak area of the set) and averaged across samples; samples were divided by the resulting factor, thus normalizing for any technical variability due to MS signal fluctuation or pipetting and sample injection errors (usually within 10% variability). (ii) Normalization for biological material was based on detected polar metabolites as follows: CV values (based on pooled sample reinjections) and coefficient of determination (R-squared) (based on linear dilutions of pooled sample) were calculated per metabolite. Metabolites with <30% CV and >0.95 RSQ were mean-centered and averaged across samples. Metabolite peak areas were then divided by the resulting factor (biological normalizer), thus accounting for any global shift in metabolite amounts due to differences in biological material. Normalization was performed in R using scripts found at www.github.com/FrozenGas/KanarekLabTraceFinderRScripts. The TraceFinder library used for analysis can be found in Supplementary Data S11.

Nucleotide heatmaps were generated using Morpheus, https://software.broadinstitute.org/morpheus.

Gene targeting using single sgRNAs

sgRNAs were cloned into the doxycycline-inducible sgRNA vector FgH1tUTG (Addgene #70183). sgRNAs were picked using the Broad Institute CRISPick tool^69,70. The sgRNA sequences used for validation studies were:

SLC31A1 #1: TTGGTGATCAATACAGCTGG

SLC31A1 #2: TCAGCCTCACACTCCCATGG

SLC31A1 #3: TCCTCCACCATGGGAGTGTG

SLC31A1 #4: AGACAGCAGCATGATGATGA

Intergenic #1: AAAGACGCGTAGGTTGTACC.

Guides were ordered as DNA oligos with 5’-TCCCG overhang on the forward sequence, and 5’-AAAC-3’ overhangs on the reverse sequence. After annealing and phosphorylation using T4 Polynucleotide kinase (NEB M0201S), digestion and ligation into the vector was accomplished using BsmBI-v2 (NEB R0739S) and T4 ligase (M0202M), and the subsequent vector was transformed into chemically competent DH5a cells for plasmid propagation.

Bacterial work and plasmid DNA propagation

Chemically competent DH5a cells were prepared using the Inoue method. Transformation was performed by thawing competent cells on ice, adding 3uL of plasmid solution (<200ng),, incubating on ice for 30min, and heat shocking at 42C for 30 seconds. After a brief rest on ice for 5 minutes, 950uL of SOC outgrowth media (RPI S26300–50.0) was added and bacteria were allowed to recover at 37C for 1 hour with shaking at 200rpm. For colony selection, transformed DH5a cells were plated on 2% Agar-LB Broth 10cm plates (Apex Bio Sciences 20–273 and Apex Bio Sciences 11–118) supplemented with 100ug/mL ampicillin (Sigma A0166–5G), wrapped in Parafilm and cultured in a 37C incubator overnight.

For suspension culture, bacterial were grown in LB broth supplemented with ampicillin (100ug/mL). and plasmid DNA extracted using Mini, Midi or MaxiPrep kits (Zymo D4211, D4200-A, D4203-A). Glycerol stocks were prepared by mixing equal parts of a confluent culture with a 50/50 solution of glycerol/H2O in a cryovial and stored in the −80.

Lentivirus production

To generate lentivirus, HEK-293T cells were seeded in 10-cm plates in DMEM (Genesee 25–500) supplemented with 10% FBS and without penicillin-streptomycin. After 24 h (or once the cells achieved 60–70% confluency), the cells were transfected with 5 ug of the above sgRNA encoding plasmids alongside 4 ug of ΔVPR envelope (Addgene #8455) and 1 ug of CMV VSV-G (Addgene #8454) packaging plasmids diluted in either the jetPRIME transfection reagent (Genesee 55–132) or TransIT-VirusGEN transfection reagent (MIR 6703). Twelve to sixteen hours after transfection, the medium was aspirated and replaced with 8 ml of fresh DMEM supplemented with 30% FBS. Virus-containing supernatants were collected 48 hours after the transfection and passed through a 0.45-μm filter to eliminate cells.

Lentiviral transduction of cell lines

Cells were pelleted, then seeded at a density of 2,500,000 cells per ml in 6-well plates in 2 ml of RPMI+10% FBS+penicillin/streptomycin containing 1:1000 polybrene (Sigma-Aldrich TR-1003-G), and then transduced with lentivirus by centrifugation at 931g for 90 min at 37 °C. After spinning, 2mL of RPMI with 10% FBS and penicillin/streptomycin was added dropwise to each well. After 12–16hr incubation, cells were pelleted to remove virus and then re-seeded into fresh culture medium. After 48–72 hours to allow for transgene expression, selection for transduced cells was performed using either puromycin treatment (1ug/mL final concentration, Invivogen ANT-PR-1) or cell sorting for fluorescent protein expression.

For SLC31A1 knockout cells, doxycycline was added at 2ug/mL final concentration into the media 48 hours after lentiviral transduction. Then, after 72 hours of doxycycline treatment, at least 500,000 SLC31A1-negative cells were sorted out using indirect staining (see above). For intergenic sgRNAs, SLC31A1-positive cells were sorted.

15N-amide Glutamine tracing

Glutamine tracing was conducted in glutamine–free RPMI (Genesee, 25–506N) supplemented with 10% dialyzed FBS and unlabeled glutamine (Sigma SLCJ2605) or ¹⁵N-amide labelled glutamine (Cambridge Isotope Laboratories NLM5570.5) and 1X Pen-Strep. A 25X stock of unlabeled and labeled glutamine was prepared in MilliQ water prior to addition to the media. Before tracing, cells were cultured for at least 48 hours in glutamine-free RPMI supplemented with 10% dialyzed FBS and unlabeled glutamine.

Tracing was performed in SEM cells treated with 50 uM BCS and/or CuCl₂ for 14 days. On day 14, cells were seeded in glutamine-free RPMI supplemented with 10% dialyzed FBS, 1X Pen-Strep, labeled or unlabeled glutamine, and respective treatments. Total glutamine concentration under all medium conditions was 300mg/L. Following 4 hours of culture under labeled or unlabeled conditions, cells were quickly washed in 0.9% NaCl and extracted in 500uL of 80% /20% LC-MS grade methanol/water with 25 mM ammonium acetate, 2.5 mM Na-ascorbate, supplemented with isotopic standards as noted above. Chromatographic separation and MS data acquisition were identical as above. Relative quantification of polar metabolites was performed with TraceFinder 4.1 as described above. To identify metabolites with expected ¹⁵N labeling, the mass of the extra neutron (m = 0.997) was added to the expected m/z for each potential nitrogen that could be labeled. Relative quantification of this list of compounds allowed the calculation of percent labeling for each carbon (m + 1, m + 2, m + 3, etc.). Subtraction of the natural abundance of ¹⁵N was performed using the R package, IsoCorrectoR⁷¹. Corrected abundances were presented as the change in percent labeling between experimental conditions, or fraction of the total abundance measured in parallel unlabeled samples from the same experiment.

Human CSF collection and LC-MS analysis

Cerebrospinal fluid (CSF) samples were collected from patients were enrolled on the Dana Farber Cancer Institute (DFCI) ALL Consortium Protocol 16–001, “Treatment of Newly Diagnosed Acute Lymphoblastic Leukemia in Children and Adolescents” (clinicaltrials.gov ID NCT03020030). Protocol 16–001 enrolled patients from 2017–2022 across 8 sites in North America: Dana-Farber/Boston Children’s Hospital, Boston, MA; Columbia University Medical Center, New York, NY; Hasbro Children’s Hospital, Providence, RI; Montefiore Medical Center, Bronx, NY; Roswell Park Cancer Institute, Buffalo, NY; Rutgers Cancer Institute of New Jersey, New Brunswick, NJ; CHU de Québec, Centre Hospitalier de l'Université Laval, Québec City, Quebéc, Canada; CHU Sainte-Justine, Montreal, Québec, Canada. Institutional review board approval was obtained at each site for the treatment protocol and ancillary studies. Informed consent was provided by the subjects’ guardians or by patients over age 18. Written assent was also obtained for the patient based on institutional guidelines.

All participants were asked to consent for the optional retention of CSF for research. CSF samples were collected prior to administration of intrathecal chemotherapy at the four timepoints (F1-F4) indicated. For those patients with a diagnosis of ALL established prior to the first diagnostic lumbar puncture, consent was requested for collection of leftover CSF prior to any chemotherapy (F0). As per standard of care, CSF was removed in a volume approximately equal to the volume of chemotherapy to be administered (typically 5–7 mL). Between 0.5–1 mL was sent for routine analyses of cell count and Cytospin. The remaining, “leftover” CSF was placed on ice immediately, and centrifuged within one hour to remove cellular elements. The supernatant was frozen at −80°C and shipped on dry ice to Rutgers Cancer Institute for analysis of biomarkers. Upon receipt and verification, CSF was thawed on wet ice, divided into 100 microliter aliquots, labeled and frozen at −80°C until analysis.

Secondary use of these specimens was reviewed by the Boston Children’s Hospital IRB and given exempt status under IRB-#P00034501. Of all the CSF samples collected, 57 patients had samples at all five timepoints, and thus we selected these patients for CSF metabolite profiling. Samples were thawed again, aliquoted into 20uL aliquots, refrozen, then sent to Boston Children’s Hospital on dry ice and stored at −80°C until analysis. Frozen aliquots of the 285 human CSF samples (57 patients * 5 timepoints) were divided into five batches, and the batches were analyzed by LC-MS over the course of 6 months using the same chromatography and same LC-MS instrumentation. Samples from the same patient were run within the same batch. 20uL of human CSF was extracted for metabolomic analysis as described in the Supplementary Methods. After polar normalization for targeted analysis, peak areas were normalized to each patient’s diagnostic F0 CSF sample.

Clinical and demographic data for patients whose CSF samples were profiled are available in Supplemental Data S7.

Untargeted metabolomics data analysis using CompoundDiscoverer

Relative quantification for untargeted polar metabolomics was performed with Compound Discoverer (CD) 3.3. A general workflow was built to best suit our polar metabolomics LC-MS method (details on each parameter are given in Supplementary Data S6. Positive and negative modes were analyzed separately. MS1 in-house retention time and chemical formula databases were used (HILIC_all and MSMLS_HILIC, in Supplementary Data S8 and 9 respectively). Filtering steps were performed based on peak noise levels, ppm error, formula annotation, and the relative abundance of the integrated peak area in true sample compared to blank injections, where features with >3-fold higher in samples were retained. Within our HILIC chromatography, we rarely observe chromatographic shifts larger than 1 minute and retention times drifts larger than 40 seconds, thus these parameters were set as limits for retention time correction and matching to internal databases.

To compare peak areas across batches, a normalization factor was calculated per metabolite and per batch. The pooled samples from each batch were assumed to be of roughly similar composition; thus, the average signal per metabolite from the pools of each batch were compared to the pools from batch 1. A normalization factor was calculated per metabolite representing the fold change difference between the pools of each batch and the pools of batch 1. This metabolite-batch correction factor was then applied to the peak areas for each metabolite for each sample by batch. The details of these calculations can be found in Supplementary Data S5. The reference databases used for our CompoundDiscoverer analysis can be found in Supplementary Data S8-9). Data for individual metabolites was then plotted in Prism.

CRISPR-resistant SLC31A1 cloning

A CRISPR-resistant SLC31A1 coding sequence with a 3’ HA tag connected by 3x G4S (4 glycine and 1 serine) linker and 5’ BamHI and 3’ MluI restriction enzyme cut sites was ordered as a gene block from Integrated DNA Technologies. Following PCR amplification using DreamTaq (Thermo Scientific K1082), the PCR product was digested using BamHI and MluI and purified using the Monarch PCR & DNA Cleanup kit (T1130S). Similarly, the pLEX_TRC202 and pLEX_TRC203 vectors were also digested using BamHI and MluI, and linearized vectors gel purified (Zymo Research D4001), and the SLC31A1 ORF ligated into the linearized plasmid using T4 ligase.

Plasmids used in this study:

FUW-RLuc-T2A-PuroR (Addgene 102320), FgH1tUTG (Addgene #70183), ΔVPR envelope (Addgene #8455), CMV VSV-G (Addgene #8454), Cas9-RFP lentiviral particles (Sigma-Aldrich Cas9-RFPV-200uL), Cas9-Blast (Addgene #52962), pAW13.lentiguide.mCherry (Addgene #104375).

pLV-RFP, pLV-AOX-IRES-RFP, and pLV-NDI1-IRES-RFP were obtained as gifts from Navdeep Chandel’s lab.

ICP-MS analysis of copper

Copper (Cu) levels were quantified using inductively coupled plasma-mass spectrometry (ICP-MS). Whole mouse blood was collected by submandibular bleeding into a heparinized tube (VWR #103093–113), then spun at 2000g for 10 minutes at 4C. 10uL of the supernatant plasma was transferred to a nitric acid-washed 1.5mL tube and frozen at −80C until digestion and analysis. Sample digestion and ICP-MS analysis were conducted by the Biomolecular Mass Spectrometry Core (BMSC) within the Center of Excellence in Environmental Toxicology at the University of Pennsylvania. All samples were spiked with internal standards: yttrium-89 (¹⁸⁹Y) at 1 ppb and terbium-159 (¹⁵⁹Tb) at 0.2 ppb final concentrations. Digestion was performed by adding 250 μL of concentrated HNO₃ to each sample, followed by incubation at 60 °C for 48 hours. Digested samples were then diluted to a final volume of 5 mL using metal-free Milli-Q water.

A calibration curve was prepared concurrently using a premixed metal ion standard, starting at 250 ppb and serially diluted two-fold to 0.12 ppb. Copper levels were measured using an iCAP RQ ICP-MS instrument (Thermo Fisher Scientific).

Patient survival curves

Patient phenotypic and transcripts per million (TPM) RNA-seq data for MT1X, SLC31A1 and ATP7A from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative and Molecular Profiling to Predict Response to Treatment in ALL (MP2PRT-ALL) cohorts managed by the NCI were downloaded from the Genomic Data Commons (https://portal.gdc.cancer.gov/projects/ ) and the UCSC Xena Genomics Browser⁷² respectively on October 1, 2025. For patients with multiple represented tumour samples, only the first diagnostic bone marrow sample was used for downstream analyses. Both B and T-cell ALL cases from the TARGET-ALL-P2 cohort were included in our analyses. Overall survival was plotted for the upper and lower quartiles. Survival curves were compared by the log-rank test and P values indicated.

Animal work

The Committee for Animal Care at Boston Children’s Hospital approved all animal procedures carried out in this study. Mice were euthanized if they exhibited symptoms of lethargy, a drop in weight (10% over a day or 20% since the beginning of the experiment), reduced appetite, a lack of movement, a hunched back or hair loss. We have complied with all relevant ethical regulations while conducting animal experiments. All mice used were mixed sex NOD.CB17-Prkdcscid/NCrCrl (originally purchased from Charles River and maintained at the BCH ARCH animal facility by self-breeding), and 5–12 weeks of age at the start of each experiment.

Intravenous (i.v.) and intracisternal (i.c.) injections

To prepare cells for injection, cells were washed once with PBS and then resuspended in 0.9% NaCl at the indicated concentrations below.

For intravenous injections, cells were prepared at a concentration of 3 million cells per 200uL in sterile 0.9% NaCl in individual aliquots. 27G insulin syringes (BD/Embecta 329412) were first flushed using an extra aliquot of cells, then loaded with the 200uL of cells. Mice were briefly restrained using a mouse tail illuminator and restrainer (Braintree Scientific) and cells injected into the tail vein.

For intracisternal injections, mice were anesthetized using isoflurane (3–5% in pure oxygen) in an induction box and then transferred onto a surgical chuck with a low-profile nose cone and facemask (Colonial Medical Supply XRM-S-KIT). Mice were draped with the neck across a 15mL centrifuge tube such that the neck was in flexion, approximately 90 degrees. A 2 cm area between the ears was shaved (Wahl BravMini+) and prepped using cotton swabs dipped in either Betadine or 70% ethanol (3x each). Meanwhile, a 30G insulin syringe (EXEL 26015) was marked at 3.5mM using lab tape, flushed with a cell mixture (50,000 cells per 10uL, with a total of 500uL in a 1.5mL microcentrifuge tube) and 10uL of the cell mixture drawn up. To inject the cells, the space between the posterior skull and C1 vertebrae was palpated, and the needle inserted midline approximately 3.5mm deep. After slowly injecting 10uL of cells with a 5 second pause at the end, the needle was withdrawn, and the injection site wiped with a cotton swab dipped in 70% ethanol. Mice were monitored and euthanized if paralysis, abnormal gait or head tilt was observed following recovery.

Mouse CSF collection

Mouse CSF collection was performed as described previously¹⁷ with mice anesthetized using ketamine/xylazine (100–120mg/kg, and 10mg/kg by intraperitoneal injection).

In vivo barcoding experiment

SEM cells were transduced with a lentiviral library (derived from the pRDA_206 backbone from the Broad Institute) encoding ~25,700 genetic barcodes (gift from Marc Schwartz) at a multiplicity of infection (MOI) of 0.3–0.4. ~15 million vexGFP+ cells were sorted to maintain >500x coverage of all barcodes. The sorted cells were then engrafted by tail vein (i.v.) injection into irradiated NOD-SCID mice (250 rads within 24 hours of injection). After engraftment, mice were allowed to develop terminal disease, and cells were collected from the skull by making a sagittal cut along the superior surface of the skull, opening the skull, pulling the brain out and flushing the exposed skull cavity with PBS. To collect cells from the spleen, the intact spleen was placed on a 70 uM cell strainer and mashed through using a 3 mL syringe plunger. The strainer was flushed with PBS to collect remaining cells. To collect cells from the bone marrow, the femurs and tibiae from mice were placed in a petri dish with PBS, and each bone was flushed using a 25 G needle attached to a 3 mL syringe. The cell mixture was then strained through a 70 uM strainer. To collect cells from the peripheral blood, an EDTA-flushed 1 mL syringe was used to collect between 500–600 μL of whole blood from each mouse by cardiac puncture of the left ventricle. After collection of all samples, the cells were spun down and subjected to RBC lysis (BioLegend 420301 or Thermo Fisher A1049201) for 10 minutes. Cell pellets were then washed with PBS and frozen at −80°C until further processing.

Genomic DNA was extracted using the QIAamp DNA mini kit (Qiagen 51304), and next-generation sequencing (NGS) was used to determine the number of barcodes recovered from each site. NGS was carried out at the Broad Institute by the Genetic Perturbations Platform using Hiseq2500 50 cycles. As a proxy for the number of barcodes retained within each site, we ranked the barcodes by most to least number of reads, and calculated the number of barcodes whose cumulative sum added up to 90% of total reads from each site.

In vivo CRISPR screens

Each library contained 172 sgRNAs targeting 43 library-unique genes, with 4 sgRNAs per gene, and 20 shared, control sgRNAs targeting non-coding intergenic loci (Extended Data Fig. 1e). Plasmid pools were obtained from the Broad Institute (clone pools 1726–1729). For each screen, lentiviral production was performed in-house in triplicate in 15cm plates with 293T cells at 60–70% confluency using TransIT VirusGEN Transfection reagent (MIR6703). 8 ug of ΔVPR, 2 ug of VSV-G, and 10 ug of pool plasmid was used per transfection. SEM cells were transduced at 0.3–0.4 MOI (determined by vexGFP positivity following transduction) using the resulting lentiviral supernatant. Forty-eight hours after lentiviral transduction, at least 20 million vexGFP+ cells were sorted in batches over the course of a day. The next day, 8–12 week old mice were irradiated (250 rads) and engrafted with 3 million CRISPR screen cells each by tail vein injection. Aliquots of cells at injection were frozen down as a reference sample to ensure representation of all guides. 5–6 mice were used per library, and the screens were staggered but all mice for one library were engrafted at the same time. After engraftment, mice were allowed to develop terminal disease, and cells were collected from the skull, spleen, bone marrow, and peripheral blood as described above. Genomic DNA was extracted using the QIAamp DNA mini kit (Qiagen 51304), and next-generation sequencing (NGS) was used to determine the number of sgRNA reads from each site. NGS was carried out at the Broad Institute by the Genetic Perturbations Platform using Hiseq2500 50 cycles.

sgRNA reads were processed using MAGeCK¹⁹, using paired analysis comparing the brain to the sum of reads from the spleen and bone marrow, and normalizing to the median of control intergenic sgRNAs. Each mouse was considered its own replicate for each library. For scatterplots, the mean sgRNA reads for each gene were taken from the MAGeCK output files.

Mouse xenograft SEM-cell line derived model of ALL

Cells were engrafted into mice as described above. For methotrexate treatments, a 1 mg/mL stock of methotrexate (SigmaAldrich M8407–100MG) was prepared in sterile PBS. Mice were given 5 mg/kg doses of methotrexate intraperitoneally in a total volume of 200 μL , or sterile PBS as vehicle. Mice were weighed either the day prior or the same day to calculate the appropriate dose.

Bioluminescent imaging was performed using the IVIS Spectrum in vivo imaging system (PerkinElmer) operating the LivingImage (v4.8.2) software. Mice were given intraperitoneal injection of 50 μg dissolved in 100uL 0.9% NaCl of the water-soluble substrate Coelenterazine (Nanolight Technologies #3031). After ten minutes, mice were anesthetized with isoflurane (3–4% in oxygen) and then imaged prone then supine using the automated exposure setting.

For in vivo doxycycline induction, doxycycline (Thermo Scientific AC446060250) was supplemented in the drinking water at a final concentration of 0.4 mg/mL with 2% sucrose.

Copper depletion in vivo:

For SEM-cell line derived xenografts, mice 5 weeks or older were switched to either a copper replete diet (6 ppm, Teklad TD.240097) with 2% sucrose in the drinking water (SigmaAldrich S0389–1kg) or copper deficient diet (0.3–0.7 ppm, Teklad TD.240098) with 2% sucrose and zinc acetate (SigmaAldrich 383317–25g) in the drinking water. The zinc acetate dose was 2 g/L for 4 weeks (2 weeks prior to injection and 2 weeks after); after this treatment period, all mice were switched to autoclaved water. Mice were allowed to consume food and water ad libitum, and the food and water was replaced every week.

Patient Derived Xenograft (PDX) Models of ALL

PDX cells (CBAB-62871 and DFAT-25991) were obtained from the Public Repository of Xenografts (ProXE)⁷³ as a gift from Jennifer Perry in Kimberly Stegmaier’s lab. All PDX cells used were under passage number 5.

All PDX cells were passaged and expanded exclusively in mice. For expansion, PDX cells were thawed into PBS and counted using a hemocytometer. After spinning 5 mins at 350g at room temperature, cells were resuspended to a concentration of 1 million cells per 200uL in 0.9% sterile saline. 1 million cells were then injected intravenously per mouse into irradiated (250 rads within the past 4 hours) mice. At the human endpoint, cells were harvested by blunt dissection from the spleen and femurs. To collect cells from the spleen, the intact spleen was placed on a 70 uM cell strainer and mashed through using a 3 mL syringe plunger. The strainer was flushed with PBS to collect remaining cells. To collect cells from the bone marrow, the femurs and tibiae from mice were placed in a petri dish with PBS, and each bone was flushed using a 25 G needle attached to a 3 mL syringe. The cell mixture was then strained through a 70 uM strainer. The cells from both sites were then mixed, counted, and frozen in 20–50million cell aliquots in 90% FBS + 10% DMSO. After each passage, the recorded passage number was increased by 1.

For experiments to assess the efficacy of treatments on PDX models, mice were irradiated (250 rads) within 4 hours of being engrafted with 1 million PDX cells as above. The next day, mice were randomized to either copper replete diet (CuR) or copper depleted diet (CuD) for the rest of the experiment, with both groups receiving normal autoclaved drinking water. Methotrexate treatments were carried out as above. To monitor the progression of the disease, 2–3 drops of mouse blood was collected by submandibular bleeding into a heparinized tube (VWR #103093–113), then lysed for 5 minutes at room temperature using 1mL of either ACK lysis buffer (Thermo Fisher Scientific A1049201) or a 1x prepared solution of 10x RBC lysis buffer (BioLegend 420301). Samples were then spun for 5 minutes at 350G at 4C, resuspended in 250uL PBE and transferred into a 96-well U-bottom plate (GenClone 25–224) for staining. Cells were pelleted, and resuspended in 250 μL of staining buffer (hCD19 1:200, hCD2 1:200, hCD45 1:200, Near IR Live Dead stain 1:1000; catalog numbers listed above) and incubated for 30 minutes at 4C in the dark. Cells were pelleted once more, washed 1x in PBE, pelleted, then resuspended in 250 μL PBE and strained through 35 μm cell strainer caps (Genesee Scientific 28–155) into FACS tubes for analysis.

A summary of available clinical data and mutational data for the PDXs used in this study is available in Supplementary Data S10.

Magnet Activated Cell Sorting (MACS) of Human Leukemia Cells from Mice

Mice were anesthetized with ketamine/xylazine (100–120 mg/kg; and 10 mg/kg by intraperitoneal injection), and leukemia cells were harvested from the skulls and spleens of mice by rough dissection as described above. Cells were resuspended in 160 μL of PBE, and 50 μL of Mouse Cell Depletion Cocktail (Miltenyi Biotec 130–104-694) was added per sample and samples incubated on ice for 5 minutes. While incubating, one LS column (Miltenyi Biotec 130–042-401) per sample was affixed to a QuadroMACS magnet (Miltenyi Biotec 130–091-051) and primed using 3 mL of ice-cold PBE. The cell suspension was then applied to each column and the flow-through containing unlabelled, human leukemia cell-enriched fraction was collected. The column was washed with an additional 2×1 mL of PBE and the flow-through collected as well. The columns were then discarded. The flow-through was spun at 4000 g for 1 minute at 4° C, and the cell pellet resuspended in 1 mL of 0.9% NaCl. After an additional 20 second spin at 4° C and 21,000rcf, cells were resuspended in for 400uL of 100% LC-MS grade methanol with internal standards for polar metabolite detection as described above.

For in vivo isotope tracing using 3-¹³C serine (Cambridge Isotope Laboratories CLM-1572-PK), an 80mg/mL solution of labelled serine was prepared in sterile 0.9% NaCl and kept at −20 until use. Mice were intraperitoneally with 800mg/kg of labelled serine 16-hours prior to cell harvest by MACS. Mice were staggered in groups such that 4–6 mice were sacrificed per hour. To identify metabolites with expected ¹³C labeling, the mass of the extra neutron (m = 1.00335) was added to the expected m/z for each potential carbon that could be labeled. Relative quantification of this list of compounds allowed the calculation of percent labeling for each carbon (m + 1, m + 2, m + 3, etc.). Natural abundance was corrected as described above for glutamine tracing. Corrected abundances are presented as the change in percent labeling between experimental conditions.

Histology and quantification

After harvest, mouse tissues were fixed for 24 hours in 10% neutral buffered formalin (Thermo Scientific #5700), then washed briefly in deionized water and stored in 70% ethanol/30% water until embedding. For histology of mouse skulls, intact heads were fixed for 24 hours, then skinned and fixed for another 24 hours. All tissues were paraffin-embedded and sectioned by the Boston Children’s Hospital Pathology Core. H&E staining was performed by the same core.

For human COX IV staining, slides were rehydrated with serial washes in xylene, then 100%, 95%, 85% and 75% ethanol, each for 5 minutes. Slides then were subjected to heat-induced epitope retrieval by immersion in citrate buffer (10mM sodium citrate, 0.05% Tween-20, pH = 6.0) and heating in a steamer for 30 minutes. Slides were washed 3x with TBST (TBS + 0.1% Tween 20), permeabilized for 5 minutes with 0.25% Triton-X 100 in 1X Tris-buffered saline (TBS), blocked for 1 hour at room temperature with 10% goat serum (Colorado Serum Company #30920) in TBST, and incubated overnight in primary antibody (CST #4850) diluted 1:500 in blocking buffer at 4°C. After three washes the next morning with TBST, slides were then incubated with Alexa-488 goat anti-rabbit secondary antibody (Abcam AB150077) diluted 1:500 in blocking buffer for 1 hour at room temperature. Slides were then washed 3x for 10 minutes in TBST. In the first TBST wash, DAPI was added at a final concentration at 1 ug/mL. Slides were mounted with 200 μL of ProLong Gold Antifade Mountant (Invitrogen P36930) and allowed to harden for 1 hour at room temperature before imaging.

Both H&E and fluorescently-stained slides were imaged on a Zeiss Axio Imager.Z2 Microscope (BCH Cellular Imaging Core) using the ZEN 3.11 software for Windows. For COX IV quantification, 4–6 high-powered (63x objective) fields were acquired per tissue and the mean AF488 intensity was calculated on a per tissue, per mouse basis. All slides were stained simultaneously, imaged using the same acquisition settings (DAPI – 200 ms at 2%, AF488 – 120 ms at 6%, AF594 – 1050 ms at 16%), and all immunofluorescence images were acquired in the same imaging session. One slide of tissue was analyzed per mouse. Slide labels were blinded using tape and randomized using a numeric code prior to imaging and quantitation, and labels were re-linked following quantitation. Mean intensity on the AF488 channel per image was calculated using ZEN lite 3.11 for Windows. For per-cell AF488 signal, 5 cells were selected randomly per image and encircled using the spline contour function in ZEN and the resulting mean intensity value was taken. Images selected for display in figures were exported from ZEN using identical display settings by tissue: for spleen, the settings were: AF488 range 30–255, gamma = 1.0, DAPI range 29–107, gamma = 1.0. For femurs/bone marrow, the settings were: AF488 range 25–249, gamma = 1.0, DAPI range 25–110, gamma = 1.0.

Statistical analysis and software

Analysis and visualization of metabolomics data (PCA plots, volcano plots) were generated with MetaboAnalyst 5.0 (www.metaboanalyst.ca). Schemes were created in BioRender and edited in Adobe Illustrator. Statistical analysis was performed using GraphPad Prism 9. All error bars represent SD except for mouse IVIS data, in which the error bars report SEM.

Extended Data

Extended Data Figure 1: In vivo CRISPR screen identifies organ-specific metabolic vulnerabilities.

(a) Overview of barcoding experiment. i.v. = intravenous. (b) Total cell recovery by organ in millions as determined by hemocytometer counting with trypan blue. Each dot represents an individual mouse, n = 4 mice. lepto = leptomeninges. (c) Total number of barcodes recovered from each organ per mouse as determined by next-generation sequencing. The number of barcodes recovered from an organ was determined by the number of barcodes summing up to 90% of total reads from each sample. (d) Venn diagrams (1 per mouse) illustrating unique and shared barcodes across organs within each mouse. (e) An overview of CRISPR library design. (f) CRISPR screen results. Grey dots represent genes in screen, black dots represent intergenic guideRNAs with linear regression and 99% confidence band overlaid. Blue dots represent known ALL dependencies. Axes represent mean log(2) fold change of the MAGeCK-normalized read counts of guides for each gene, relative to the mean read counts of the intergenic guides. (g) Genes from the CRISPR screen are ranked by relative depletion or enrichment in one organ versus another. The 5 most depleted (blue) or enriched (red) genes are highlighted in each plot.

Extended Data Figure 2: Knockout of SLC31A1 inhibits growth of ALL in vivo.

(a) Total copper levels in the plasma and cerebrospinal fluid (CSF) of untreated NOD-SCID mice as measured by inductively-coupled plasma mass spectrometry (ICP-MS). (b) flow cytometric assessment of SLC31A1 expression using indirect antibody staining in live SEM cells with knockout of SLC31A1 with GFP or CRISPR-resistant SLC31A1 rescue. (c) Western blot validation of SLC31A1 knockout. (d) Representative bioluminescent imaging (BLI) of mice 3 weeks after receiving luciferase-expressing SEM cells by either intravenous or intracisternal injection. (e) Hematoxylin and eosin staining of sections from skulls, spleens and femurs of terminal mice receiving luciferase-expressing SEM cells by intravenous or intracisternal injection. Scale bar for whole-skull images are 1cm, all other scale bars (including skull inset images) are 100 micrometers. Orange rectangles denote inset of skull images. (f) Change in %GFP+ cells compared to input, from CNS of mice at clinical endpoint that received intergenic or SLC31A1-KO sgRNA cells mixed with control mCherry-expressing cells by intravenous injection. Data are normalized to the mean of the intergenic group. (g) Quantitative real-time polymerase chain reaction (qRT-PCR) measurement of SLC31A1 mRNA knockdown following doxycycline-inducible sgRNA expression. mRNA was collected from cells after 9 days of doxycycline induction. Each group is normalized to the -doxycycline condition. (h) Experimental setup of in vivo mixing experiment in which doxycycline was administered via the drinking water starting at day 5 post engraftment. (i-j) Change in %GFP+ cells compared to input, from mice at clinical endpoint that received intergenic or SLC31A1-KO sgRNA cells mixed with control mCherry-expressing cells. Data are normalized to the mean of the intergenic group and are from the (i) spleen or (j) CNS of mice receiving cells by intravenous injection. Plotted data are mean ± SD. p-values from Šídák's multiple comparisons test. Each dot in bar graphs represents a biological replicate or mouse.

Extended Data Figure 3. Analysis of TARGET-ALL-P2 survival and SLC31A1/ATP7A gene expression data.

(a) qRT-PCR measurement of SLC31A1 or ATP7A mRNA levels in 8-day BCS or CuCl₂ treated SEM cells relative to vehicle treated cells. n=3 biological replicates. (b-g) Kaplan-Meier survival curves of patients from MPT2PRT-ALL or TARGET-ALL-P2 cohorts with available bone marrow mRNA-sequencing and overall survival (OS) data, vertical dashes represent censored patients. Patients were sorted by (b, e) ATP7A, (c ,f) SLC31A1, or (d, g) log(SLC31A1/ATP7A) expression in transcripts per million (TPM), and the survival data of the top (”high”) and bottom quartiles (”low”) were plotted. (h) qRT-PCR measurement of MT1X mRNA levels in 8-day BCS or CuCl₂ treated SEM cells relative to vehicle treated cells. n=3 biological replicates. (i-j) Kaplan-Meier survival curves of patients from (i) MPT2PRT-ALL or (j) TARGET-ALL-P2 cohorts sorted by COX17 expression with the top and bottom quartiles plotted. p-values from log-rank (Mantel-Cox) tests or (h) Dunnett’s multiple comparisons test.

Extended Data Figure 4: Copper depletion does not significantly affect baseline NADP/NADPH levels, glutathione ratios nor hinder MAPK signaling in ALL cells in vitro.

(a) A scheme illustrating primary intracellular functions of copper ions. (b-c) NADP, NADPH levels and NADP/NADPH ratio, as well as oxidized and reduced glutathione levels as determined by LC-MS targetted metabolite profiling in (b) SLC31A1-KO SEM cells with or without genetic rescue, or (c) BCS-treated SEM cells (day 14), NALM6 (day 6), REH (day 8) or MOLT16 (day 8) with or without CuCl₂ rescue. Levels of reduced glutathione were determined by derivatization with Ellman’s reagent (see Supplementary Methods). (d-e) Western blotting for total and phospho (p-)ERK1/2 levels in (d) SLC31A1-KO or (e) BCS treated SEM, NALM6, REH or MOLT16 cells. (f) Doublings per day normalized to vehicle-treated cells after either 4 days (NALM6) or 6 days of BCS treatment (SEM, REH, MOLT16, MOLT4, PF382). Doubling rate was calculated by taking the linear regression of cell number across at least 2 passages over 4 days. (g) MitoStress Seahorse Assay for indicated cell lines treated with BCS and/or CuCl₂ after 10 days of treatment. For Seahorse assays, plotted data are mean ± SD with 4–6 technical replicates (wells). All other plotted data are mean ± SD, p-values from Šídák's multiple comparisons test. Each dot represents a biological replicate. OE = overexpression, sgRNA = single guide RNA, oligo = oligomycin, AA = antimycin A, Rot = rotenone, FCCP = carbonyl cyanide-p-trifluoromethoxyphenylhydrazone.

Extended Data Figure 5: Copper depletion disrupts complex IV and perturbs ETC activity.

(a) Western blot for COX2, COX1 and COX4 in BCS-treated SEM (day 8), NALM6 (day 6) and MOLT16 (day 8) cells. (b) Blue-native (BN-PAGE) followed by western blot detection of indicated subunits in native complex I-III in BCS treated SEM cells (day 8) or NALM6 cells (day 6). Representative data of two biological replicates shown. Additional replicate shown in Source Data. (c-d) Levels of NAD, NADH and the NAD/NADH ratio were determined by LC-MS targeted metabolite profiling in (c) SLC31A1-KO SEM cells or (d) BCS treated SEM (day 8), NALM6 (day 6) and MOLT16 (day 8) cells. (e) Levels of succinate, fumarate, and succinate/fumarate ratio in BCS treated SEM and NALM6 cells as determined by targeted LC-MS metabolite profiling. (f-g) Seahorse assay to measure OCR in permeabilized BCS-treated (f) SEM (day 8) or (g) NALM6 (day 6) cells following treatment with complex I or II substrates. (h-i) MitoStress Seahorse assay performed on SLC31A1-KO SEM cells with (h) eGFP or SLC31A1 rescue, or (i) 100 μM CuCl₂ pre-treatment for 24 hours. For Seahorse assays, plotted data are mean ± SD with 4–6 technical replicates (wells). Other plotted data are mean±SD, p-values from Šídák's multiple comparisons test. Each dot represents a biological replicate. BCS and CuCl₂ were used at 50 μM. OE = overexpression, BCS = bathocuproinedisulfonic acid, oligo = oligomycin, AA = antimycin A, rot = rotenone, FCCP = carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, perm = XF Plasma Membrane Permeabilizer, pyr/mal = pyruvate and malate, succ = succinate.

Extended Data Figure 6: Rescue of ETC activity with AOX over expression partially rescues cell proliferation in the context of copper depletion.

(a) Relative cell number of SEM, NALM6 or MOLT16 cells expressing RFP or AOX and treated with vehicle or Antimycin A (5 nM) for 3 days. (c-c) Seahorse MitoStress Assay on (b) 6-day BCS-treated SEM cells overexpressing RFP or AOX, or (c) SLC31A1-KO SEM cells overexpressing RFP or AOX. Plotted data are mean ± SD with 5 technical replicates (wells) for Seahorse Assays. (d-e) Levels of NAD, NADH and NAD/NADH ratio were determined by LC-MS targeted metabolite profiling in (d) SLC31A1-KO SEM cells over expressing RFP or AOX, or (e) day 8 BCS-treated SEM cells over expressing RFP or AOX. (f) Flow cytometry histograms to measure mitochondrial mass (MitoSpy Green FM) and mitochondrial membrane potential (TMRM) in BCS-treated SEM (day 8), NALM6 (day 6), and MOLT16 (day 8) cells. Bar graphs depict the median TMRM/MitoSpy ratio for all recorded events. Cells were also gated for live cells using the Near IR Fixable LIVE/DEAD stain measured on the APC-Cy7 channel (APC-Cy7 negative). (g) Doublings per day from day 5–11 of BCS-treated AOX-overexpressing SEM or NALM6 cells also treated with MnTE-2-PyP (20 μM) or MnTBAP (50 μM). (h) Relative cell number of SEM, NALM6, or MOLT16 cells expressing RFP or NDI1 and treated with vehicle or Antimycin A (5 nM) for 3 days. (i) Doublings per day from day 5–11 of BCS treatment of SEM, NALM6, or MOLT16 cells overexpressing RFP or NDI1. Plotted data are mean±SD, p-values from Šídák's multiple comparisons test. Each dot represents a biological replicate. oligo = oligomycin, AA = antimycin A, Rot = rotenone, FCCP = = carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, MnTE-2-PyP = manganese (III) meso-tetrakis (N-ethylpyridinium-2-yl) porphyrin, MnTBAP = manganese (III) tetrakis(4-benzoic acid) porphyrin chloride, AOX = alternative oxidase from Ciona intestinalis, NDI1 = yeast NADH-ubiquinone reductase.

Extended Data Figure 7: Copper depletion disrupts nucleotide synthesis in an ETC-dependent manner.

(a) Volcano plot of changing metabolites between SLC31A1-KO SEM cells and control cells. n = 3 biological replicates. (b) Aspartate and dihydroorotate (DHO) levels at baseline in SLC31A1-KO or control SEM cells with eGFP or SLC31A1 rescue. (c) Nucleotide levels in SLC31A1-KO SEM cells with eGFP or SLC31A1 rescue. (d) Volcano plot of changing metabolites between day 8 BCS-treated MOLT16 cells and control cells. n = 3 biological replicates. (e-f) Aspartate and (f) nucleotide levels in day 8 BCS-treated MOLT16 cells. (g-h) Volcano plot of significantly changing metabolites between day 14 BCS-treated (g) or SLC31A1-KO (h) SEM cells, with amino acids highlighted in pink. n = 3 biological replicates. (i) Western blot for DHODH levels in BCS-treated SEM (day 8), NALM6 (day 6) and MOLT16 cells (day 8); each lane represents an independent biological replicate. Actin is a loading control. . (j) Nucleotide levels in AOX-overexpressing SEM cells treated with BCS for 8 days. (k-l) Levels of aspartate, DHO, and IMP in (k) 14-day BCS-treated SEM cells overexpressing RFP or AOX, (l) or SLC31A1-KO SEM cells overexpressing RFP or AOX. DHO = dihydroorotate, CAA = carbamoyl aspartic acid, AOX = alternative oxidase, IMP = inosine monophosphate. Plotted data are mean ± SD. p-values from Šídák's multiple comparisons test.

Extended Data Figure 8: Both purine and pyrimidine synthesis are impacted by copper depletion.

(a-b) Fraction abundance of ¹⁵-N labelled metabolites relative to levels in vehicle-treated cells in BCS-treated SEM cells (14 days), NALM6 (6 days) or MOLT16 (8 days). (c) scheme of substrates consumed to synthesize adenosine monophosphate (AMP) and guanosine monophosphate (GMP) from inosine monophosphate (IMP). (d) Fraction abundance of ¹⁵-N labelled purine monophosphates in 14-day BCS treated SEM cells relative to vehicle-treated cells. (e) Relative abundance (normalized to vehicle-treated group) (left bar graphs) or fractional ¹⁵N label incorporation (right bar graphs) of indicated metabolites in BCS-treated SEM (14 days), NALM6 (6 days) or MOLT16 (8 days) cells. Plotted data are mean ± SD. p-values from Tukey’s multiple comparisons tests.

Extended Data Figure 9: Supplementation with nucleoside or nucleoside precursors bolsters nucleotide synthesis in copper-depleted cells.

(a) Cell number after 4 days of growth for SLC31A1-KO cells treated with pyruvate and/or uridine relative to vehicle-treated intergenic sgRNA-containing cells, all seeded at 0.1 million/mL. (b) Nucleotide levels in SEM cells treated with BCS and/or pyruvate and uridine for 8 days. (c-d) Whole cell NAD, NADH and NADH ratios in (c) SLC31A1-KO SEM cells with or without pyruvate and uridine treatment, or (d) 14-day BCS-treated SEM cells with or without concurrent pyruvate and uridine treatment. (e) Doublings per day from days 5–11 (SEM and NALM6) or days 2–8 (MOLT16) of cells treated with BCS and/or uridine and inosine (100 μM each). Doubling rate was calculated from the linear regression of cell number across at least 2 passages. (f) Levels of dihydroorotate (DHO), aspartate, or the NAD/NADH ratio in SEM and NALM6 cells after 8 days of treatment with BCS and/or inosine and uridine (Ino/Urid) relative to vehicle treated cells. CAA = carbamoyl aspartic acid. Plotted data are mean ± SD, p-values from Šídák's multiple comparisons test. Each dot represents a biological replicate.

Extended Data Figure 10: Dietary copper depletion combines with methotrexate to reduce leukemic growth in vivo.

(a) Experimental setup for copper depletion in leukemic mice. BLI = bioluminescent imaging. i.v. = intravenous, (b) Total photon flux from BLI at indicated timepoints expressed as fold change compared to day 14. n = 5 (CuR) or 6 (CuD). Plotted values represent mean +/− SEM. Representative BLI images shown on the right. (c) Principal component analysis (PCA) plot of targeted LC-MS metabolomics data of SEM cells. Cells were treated with 50 μM BCS for 8 days, 6-mercaptopurine (6-MP, 500 nM) for 24 hours, and/or methotrexate (MTX, 5 nM) for 24 hours. (d) Relative abundances of AICAR in MTX and BCS treated SEM and NALM6 cells, related to panel (c). (e) Dose-responsive curves of either SLC31A1-KO SEM cells or 8-day BCS-treated SEM cells treated with MTX for 3 days. Cells were plated in 96-well plates at a density of 0.25 million/mL. Relative luminescence was measured as a proxy for cell number using CellTiter-Glo. (f) Spleen weights at day 30 of mice from Main Figure 4c as a percentage of total body weight. (g-h) Quantification of mean intensity value for COX IV staining per whole image using ZEN for mouse (g) femurs and (h) spleens. 4–6 images were quantified per tissue per mouse. For femurs, n = 41 and 28 total images and for spleens, n = 41 and 31 images were quantified respectively. Lines indicate quartiles and median. (i) Example of per-cell quantification of COX IV staining intensity. Numbers in red indicate mean intensity value as reported from ZEN. All scale bars are 20 micrometers. (j) Relative unlabelled UTP and CTP levels in SEM leukemia cells harvested by MACS from the spleens of CuR+Vehicle or CuD+ZnAc treated mice. (k) m+1 ¹³C-labelled fraction of serine as measured from SEM leukemia cells harvested from the CNS or spleens of mice by MACS. (l) Ellman’s-derivatized glutathione (GSH) levels in SEM leukemia cells harvested from CNS or spleen. (m) Spleen weights from the endpoint of mice engrafted with PDX CBAB-62871 at week 7, related to Fig 5m-n. CuR = copper replete diet, CuD = copper depleted diet, AICAR = 5-aminoimidazole-4-carboxamide ribonucleotide, MACS = magnet-activated cell sorting. Plotted values are mean+/−SD unless otherwise specified. n.s. = not significant. p-values from Tukey’s multiple comparison tests (b and d), or unpaired t-tests (f-h, j-m).

Supplementary Material

Additional information: Supplementary information is available for this paper. Correspondence and requests for materials should be addressed to Naama Kanarek at naama.kanarek@childrens.harvard.edu. Reprints and permissions information is available at www.nature.com/reprints.

Data and Code Availability:

Raw mass spectrometry data generated as part of this study is deposited at Metabolomics Workbench⁷⁴ (Study numbers #:ST003819, ST003820, ST003821, ST003832, ST003822, ST003823, ST003824, ST003825, ST003826, ST003827, ST003831, and ST003838; additional study IDs are pending). Metabolite mass-to-charge ratios and retention times for targeted analysis were compared to in-house metabolite standards, and reference values can be found in Supplementary Data S4 and S11. R scripts used for formatting and normalization of metabolomics datasets can be found at: https://github.com/FrozenGas/KanarekLabTraceFinderRScripts.