Erratum 22 April 2022. See Erratum.

RES EARCH

CELL DEATH death is primarily dependent on intracellular

copper accumulation.
Copper induces cell death by targeting lipoylated Copper ionophores induce a distinct form of
TCA cycle proteins regulated cell death
We first asked whether copper ionophore–
Peter Tsvetkov *, Shannon Coy
1 2,3,4,5 5,6
, Boryana Petrova , Margaret Dreishpoon , Ana Verma 1 2,3,4,5
, mediated cell death is regulated and specif-
Mai Abdusamad1, Jordan Rossen1, Lena Joesch-Cohen1, Ranad Humeidi1, Ryan D. Spangler1, ically whether short-term exposure leads to
John K. Eaton1, Evgeni Frenkel7, Mustafa Kocak1, Steven M. Corsello1,5,8, Svetlana Lutsenko9, irrevocable, subsequent cell cytotoxicity. Pulse
Naama Kanarek1,5,6, Sandro Santagata2,3,4,5,10, Todd R. Golub1,5,11,12* treatment with the copper ionophore elesclomol
at concentrations as low as 40 nM for only
Copper is an essential cofactor for all organisms, and yet it becomes toxic if concentrations exceed a 2 hours resulted in a 15- to 60-fold increase
threshold maintained by evolutionarily conserved homeostatic mechanisms. How excess copper induces in intracellular copper levels (fig. S2, B and C)
cell death, however, is unknown. Here, we show in human cells that copper-dependent, regulated cell that triggered cell death more than 24 hours
death is distinct from known death mechanisms and is dependent on mitochondrial respiration. We show later (Fig. 1C). This result suggests that copper-
that copper-dependent death occurs by means of direct binding of copper to lipoylated components mediated cell death is indeed regulated.
of the tricarboxylic acid (TCA) cycle. This results in lipoylated protein aggregation and subsequent Cell death involves signaling cascades and
iron-sulfur cluster protein loss, which leads to proteotoxic stress and ultimately cell death. These molecularly defined effector mechanisms (20)
findings may explain the need for ancient copper homeostatic mechanisms. that involve proteins and lipids such as those
characteristic of apoptosis (21), necroptosis

T
(22), pyroptosis (23), and ferroptosis (24), which
he requirement of copper as a cofactor nates the cytotoxicity of the compounds [fig. is a recently discovered iron-dependent cell
for essential enzymes has been recognized S1, B to C, and (5)]. death pathway. Previous reports suggested that

Downloaded from https://www.science.org on June 12, 2024

across the animal kingdom, spanning A clear picture of the mechanisms under- elesclomol induces ROS-dependent apoptotic
bacteria to human cells (1). However, intra- lying copper-induced toxicity has not yet cell death (6, 12), but elesclomol-induced cell
cellular copper concentrations are kept emerged, with contradictory reports suggest- death did not involve either the cleavage or
at extraordinarily low levels by active homeo- ing either the induction of apoptosis (11, 12), activation of caspase 3 activity, the hallmark of
static mechanisms that work across concen- caspase-independent cell death (5, 7, 13), reac- apoptosis (25) (Fig. 1, D and E, and fig. S2D).
tration gradients to prevent the accumulation tive oxygen species (ROS) induction (14–16), or Similarly, elesclomol killing potential was main-
of free intracellular copper that is detrimental inhibition of the ubiquitin-proteasome system tained when the key effectors of apoptosis, BAX
to cells (1–4). Whereas the mechanisms of tox- (17–19). The cross-kingdom efficacy of copper- and BAK1, were knocked out (Fig. 1F and fig.
icity of other essential metals, such as iron, are binding molecules as cell death inducers sug- S2, E to I) or when cells were cotreated with pan-
well established, the mechanisms of copper- gests that they target evolutionarily conserved caspase inhibitors (Z-VAD-FMK and Boc-D-FMK)
induced cytotoxicity remain unclear (5–7). cellular machinery, but such mechanisms have (Fig. 1G), again indicating that the copper-
Copper ionophores are copper-binding small yet to be elucidated. induced cell death is distinct from apoptosis.
molecules that shuttle copper into the cell and To further establish whether copper iono- Furthermore, treatment with inhibitors of other
are thereby useful tools to study copper toxicity phore cytotoxicity is dependent on copper known cell death mechanisms—including fer-
(8, 9). Multiple lines of evidence indicate that itself, we analyzed the killing potential of the roptosis (ferrostatin-1), necroptosis (necrostatin-1),
the mechanism of copper ionophore–induced potent copper ionophore elesclomol. The source and oxidative stress (N-acetyl cysteine)—all
cell death involves intracellular copper accumu- of copper in cell culture medium is serum (fig. failed to abrogate copper ionophore–induced
lation and not the effect of the small molecule S1D), and accordingly, cells grown in the ab- cell death (Fig. 1G), suggesting a mechanism dis-
chaperones themselves. Multiple, structural- sence of serum were resistant to elesclomol. By tinct from known cell death pathways (Fig. 1H).
ly distinct small molecules that bind copper contrast, elesclomol sensitivity was completely
share killing profiles across hundreds of cell restored by the addition of copper in a 1:1 ratio Mitochondrial respiration regulates copper
lines [Fig. 1A, fig. S1A, and (5, 10)]. Structure- (fig. S1, E and F). Copper supplementation ionophore–induced cell death
function relationship experiments show that similarly sensitized cells to treatment with six One hint to the pathways that mediate copper
modifications that abrogate the copper bind- structurally distinct copper ionophores (Fig. ionophore–induced cell death is the observa-
ing capacity of these compounds result in loss 1B and fig. S1, G to P), but supplementation tion that cells that are more reliant on mito-
of cell killing (5), and copper chelation elimi- with other metals, including iron, cobalt, zinc, chondrial respiration are nearly 1000-fold more
and nickel, failed to potentiate cell death (Fig. sensitive to copper ionophores than cells under-
1
Broad Institute of Harvard and MIT, Cambridge, MA, USA. 1B and fig. S1, G to P). Consistent with this going glycolysis (Fig. 2A and fig. S3, A to F).
2
Laboratory of Systems Pharmacology, Department of Systems observation, depletion of the endogenous in- Treatment with mitochondrial antioxidants,
Biology, Boston, MA, USA. 3Ludwig Center at Harvard, Harvard
Medical School, Boston, MA, USA. 4Department of Pathology,
tracellular copper chelator glutathione, using fatty acids, and inhibitors of mitochondrial
Brigham and Women’s Hospital, Boston, MA, USA. 5Harvard buthionine sulfoximine (BSO), sensitized cells function had a very distinct effect on the sen-
Medical School, Boston, MA, USA. 6Department of Pathology, to elesclomol-copper–induced cell death (fig. sitivity to copper ionophores as compared with
Boston Children’s Hospital, Boston, MA USA. 7Whitehead
Institute and Massachusetts Institute of Technology,
S1Q), whereas chelation of copper with tetra- sensitivity to the ferroptosis-inducing GPX4
Cambridge, MA, USA. 8Department of Medical Oncology, thiomolybdate (TTM) rescued killing (fig. S1, B inhibitor ML162 (Fig. 2B). Furthermore, in-
Dana Farber Cancer Institute, Boston, MA, USA. 9Department and C) and chelators of other metals had no hibitors of complexes I and II of the electron
of Physiology, Johns Hopkins Medical Institutes, Baltimore,
effect (fig. S1R). Lastly, a 2-hour pulse treatment transport chain (ETC) as well as inhibitors of
MD, USA. 10Department of Pathology, Dana Farber Cancer
Institute, Boston, MA, USA. 11Department of Pediatric with potent copper ionophores (elesclomol, mitochondrial pyruvate uptake attenuated
Oncology, Dana Farber Cancer Institute, Boston, MA, USA. disulfiram, and NSC319726) resulted in a ~5- cell death with no effect on ferroptosis (Fig.
12
Division of Pediatric Hematology/Oncology, Boston to 10-fold increase in levels of intracellular 2B). Importantly, the mitochondrial uncou-
Children’s Hospital, Boston, MA, USA.
*Corresponding author. Email: ptsvetko@broadinstitute.org (P.T.); copper but not zinc (fig. S2A). These results pler FCCP had no effect on copper toxicity,
golub@broadinstitute.org (T.R.G.) suggest that copper ionophore–induced cell suggesting that mitochondrial respiration, not

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 1 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

A B C ES viability D

new media
Control after drug pulse

Wash
Pulse

Fold Caspase 3/7 activaiton

CoCl2 ABC1 cells
Copper binding molecules PRISM Repurposing CuCl2
cluster (1448 compounds 489 cell lines) FeCl3 NiCl2 14 ES-Cu
FeCl2
2 hrs 0 8 24 96 Bortz ***
36 ZnCl2 12
1.2 ***
Pyrithione-zinc 1.2 hrs 10

Fold Viability
0.9

Fold viability
TMT 0 8
35 0.9 8
Thiram Disulfiram 0.6 6 ***
TSNE2

0.6 24 4
0.3 96
Elesclomol 2
34
8HQ 0.3
0.0 0
0.0
-9 -8 -7 -6 -5 -4

10
20
40
l
tro
-8.0 -7.5 -7.0 -6.5 -6.0
FR-122047 LOG10[Elesclomol], M (nM)

on
33 −50
LOG10[Elesclomol-Cu], M

C
16.5 17.0 17.5 −25 0 25
TSNE1

E F Elesclomol G Cell death inhibitors

H
Etoposide

1.2 Perturbation of other pathways

D-Boc-FMK
Z-Vad-FMK
5mM NAC
1mM NAC
Control

Pepstatin
does not rescue cell death

L-NAME
0.9 Bak/Bax KO #1

Control
Elesclomol (nM)

Nec-1
Fer-1
DPQ

TICAM1
0.6 Bak/Bax KO #2

TTM
- 1 10 40100

RIPK1

Fe
Cell lines Bak, Bax
0.3

DAI
25 Cleaved NAC
Fold viability

NCIH2030
Caspase 3 0.0
Elesclomol A549 CytC ROS
-9 -8 -7 -6 -5 -4 RIPK3
37 HCC4009
Caspase 3 Paclitaxel NCIH2030 Casp-9 Nec-1 Lipid
1.2 ML162
(ferroptosis) A549 MLKL peroxid-
Actin 0.9 HCC4009 ation
Casp-3/7

Downloaded from https://www.science.org on June 12, 2024

0.6 Bortezomib NCIH2030 Membrane Fer-1
(apoptosis) HCC4009 rupture
0.3
0.0 Apoptosis Necroptosis Ferroptosis
-9 -8 -7 -6 -5
0 0.2 0.4 0.6 0.8 1.0
LOG10, M

Fig. 1. Copper ionophore–induced cell death is nonapoptotic, nonferroptotic, etoposide for 6 hours. (F) Viability of HMC18 cells or two HMC18 clones with Bax/Bak
and non-necroptotic. (A) PRISM (profiling relative inhibition simultaneously in deleted after treatment with elesclomol-CuCl2 (1:1) (top) and paclitaxel (bottom).
mixtures) Repurposing secondary screen: Growth-inhibition estimates for 1448 drugs (G) Heatmap of viability of cells pretreated overnight with 20 mM necrostatin-1,
against 489 cell lines. TMT, tetramethylthiuram monosulfide; TSNE, t-distributed 10 mM ferrostatin-1, 1 mM N-acetylcysteine (NAC), 5 mM N-acetylcysteine, 30 mM
stochastic neighbor embedding. (B) Viability of cells (MON) after treatment with Z-VAD-FMK, 50 mM D-Boc-FMK, 20 mM TTM, 300 mM L-NAME, 1 mM pepstatin A,
elesclomol with or without 10 mM of indicated metals. (C) Viability of ABC1 cells was or 10 mM DPQ and then treated with either 30 nM elesclomol-CuCl2 (1:1), 1 mM ML162
assessed at the indicated times after elesclomol-Cu (1:1 ratio) pulse treatment and (GPX4 inhibitor), or 40 nM bortezomib for 72 hours (average of three replicates).
growth in fresh media. ES, elesclomol. (D) Caspase 3/7 cleavage in ABC1 cells (H) Schematic diagram of apoptosis, necroptosis, and ferroptosis. Inhibited pathways
16 hours after indicted treatments (fold change over control). (E) Western blot analysis are marked in red. For (B), (D), and (F), data are means ± SD, with n ≥ 3. For (D)
of G402 cells treated with the indicated concentrations of elesclomol or 25 mM and (E), media were supplemented with 1 mM CuCl2.

adenosine triphosphate (ATP) production, is per ionophore–induced cell death and mito- 1 (LIPT1), lipoyl synthase (LIAS), and dihydro-
required for copper-induced cell death (Fig. chondrial metabolism, not the ETC (Fig. 2F), lipoamide dehydrogenase (DLD)] or protein tar-
2C). Consistent with this finding, growing cells leading us to further elucidate the precise con- gets of lipoylation [the pyruvate dehydrogenase
in hypoxic conditions (1% O2) attenuated cop- nection between copper and the TCA cycle. (PDH) complex, including dihydrolipoamide
per ionophore–induced cell death, whereas S-acetyltransferase (DLAT), pyruvate dehydro-
forced stabilization of the hypoxia-inducible FDX1 and protein lipoylation are the key genase E1 subunit alpha 1 (PDHA1), and pyru-
factor (HIF) pathway with the HIF prolyl hy- regulators of copper ionophore–induced vate dehydrogenase E1 subunit beta (PDHB)]
droxylase inhibitor FG-4592 under normoxic cell death (26) (Fig. 3D). These observations were vali-
conditions (21% O2) did not (Fig. 2D and fig. To identify the specific metabolic pathways dated by an independent knockout screen that
S3, G to J), further emphasizing the role of cel- that mediate copper toxicity, we performed focused on 3000 metabolic enzymes (27) (Fig.
lular respiration in mediating copper-induced genome-wide CRISPR-Cas9 loss-of-function 3E; fig. S4, A and B; and table S2). In addition,
cell death. However, treatment with copper screens to identify the genes involved in cop- the metabolic enzyme screen showed that ge-
ionophores did not induce significant reduction per ionophore–induced death. To maximize netic suppression of complex I also rescued cells
in basal or ATP-linked respiration but did sig- the generalizability of the screen, we focused from copper-induced death (Fig. 3E; fig. S4, A
nificantly reduce the spare capacity of respi- on the intersection of two structurally distinct and B; and table S2), consistent with our finding
ration (Fig. 2E and fig. S3, K to N), suggesting copper-loaded ionophores (elesclomol and the that chemical inhibitors of the ETC block cell
that copper does not target the ETC directly but active form of disulfiram, diethyldithiocar- death mediated by copper ionophores (Fig. 2).
rather components of the TCA cycle. In support bamate) (Fig. 3, A to C, and table S2). Killing Individual gene knockout studies further
of this, metabolite profiling of cells pulse-treated by both compounds was rescued by knockout confirmed that deletion of FDX1 and LIAS
with elesclomol showed a time-dependent in- of seven genes (Fig. 3A, marked in blue), in- conferred resistance to copper-induced cell
crease in metabolite dysregulation of many TCA cluding FDX1 [which encodes a reductase death (Fig. 3, F and G, and fig. S4, C to K), fur-
cycle–associated metabolites in elesclomol- known to reduce Cu2+ to its more toxic form, ther strengthening a functional link between
sensitive ABC1 cells but not in elesclomol- Cu1+, and to be a direct target of elesclomol (5)] FDX1, the protein lipoylation machinery, and
resistant A549 cells (fig. S3, O to S, and table and six genes that encode either components copper toxicity. Moreover, FDX1 deletion re-
S1). These results establish a link between cop- of the lipoic acid pathway [lipolytransferase sulted in consistent resistance to a number of

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 2 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

A B Complex I/III MPC C

fatty acids inhibitors inhibitor antioxidants inhib cell death
1.2 Control
1.2 Rot

Fold viability
compounds:
Fold viability

Glu 0.9 AntiA

Fold viability
0.9 5nM ES 1.0
10nM ES
FCCP
0.6
Gal 0.6
20nM ES 0.5
NSC319726
0.3
0.3
ML162 0 0.0
0.0

JP4-039 1
JP4-039 10
Antimicyn A 0.1
Antimicyn A 1
Linoleic acid
Oleic acid

Elaidic acid
Pamitlic acid
α-Lipoic acid

Erastin

UK5099
α-Tocopherol
Ebselene 1
Ebselene 10
Idebenone 1
Idebenone 10
Liproxstatin-1 1
Liproxstatin-1 10

Necrostatin-1
Ferrostatin-1

control
control
control
γ-Linollenic
α-Linollenic

Telaglenastat

Z-VAD-FMK
Etomoxir
Arachidonic acid

Rotenone
-10 -9 -8 -7 -6 -5
-10 -9 -8 -7 -6 -5
LOG10[Elesclomol], M
LOG10[Elesclomol-Cu], M

D E F glucose galactose elesclomol Cu2+

OCR (pmol/min)
100 Oligo FCCP AntiA/Rot
1.2 Control pyruvate
Hypoxia 80 Control glycolysis mitochondrial uptake
0.9 metabolism
Viability

FG-4592 60
0.6 ES citrate
40 O2 H2O TCA cis-aconitate
0.3 malate cycle Glutamine
20 electron fumarate
0.0 transport succinate α-KG Glutamate
0 chain
-10 -9 -8 -7 -6 -5
LOG10[Elesclomol], M
0 20 40 60 80 GDP
GTP
BSO Glutathione
Time (min)
essential for promotes inhibits up/downregulated
cuproptosis cuproptosis cuproptosis metabolites

Downloaded from https://www.science.org on June 12, 2024

Fig. 2. Mitochondria respiration regulates copper ionophore–induced cell consumption rate (OCR) was detected after treatment with 2.5 nM elesclomol (with
death. (A) Viability of NCIH2030 cells grown in media containing either glucose or 1 mM CuCl2 in media) for 16 hours of ABC1 cells before (basal) and after the addition
galactose treated with elesclomol-Cu (ratio 1:1). (B) Viability of ABC1 cells pretreated of oligomycin (ATP-linked), the uncoupler FCCP (maximal), or the electron
with the indicated compounds (x axis) and then treated with elesclomol, NSC319726, transport inhibitor antimycin A/rotenone (baseline) (mean ± SD, n = 8).
or ML162 (y axis). The average of at least three replicates is plotted and color (F) Schematic of metabolites altered after elesclomol treatment of ABC1 cells.
coded. MPC, mitochondrial pyruvate carrier. (C) Viability of ABC1 cells pretreated with (Purple circles mark metabolites that change in abundance, as detailed in table S1.
0.1 mM rotenone (Rot), 0.1 mM antimycin A (AntiA), or 1 mM FCCP and then treated Larger circles are metabolites that are up-regulated, and smaller circles are
with elesclomol. (D) Viability of ABC1 cells grown in control (21% O2), hypoxia metabolites that are down-regulated). GDP, guanosine diphosphate; GTP,
(1% O2), or 50 mM FG-4592 (21% O2) after treatment with elesclomol. (E) The oxygen guanosine triphosphate. For (A), (C), and (D), data are means ± SD, with n ≥ 3.

copper ionophores (disulfiram, NSC319726, occur on only four enzymes, all of which involve proteins was highly correlated (p < 0.0001;
thiram, 8-HQ, and Zn-pyrithione) (fig. S5, A metabolic complexes that regulate carbon entry Fig. 4, B and C, and fig. S6, A to D). Third, we
to G) but showed no substantial effects on points to the TCA cycle (26, 30). These include determined whether FDX1 knockout affected
either caspase activation (fig. S5, H to K), the dihydrolipoamide branched chain transacylase protein lipoylation using a lipoic acid–specific
potency of apoptosis-inducers (bortezomib, E2 (DBT), glycine cleavage system protein H antibody as a measure of DLAT and DLST
paclitaxel, and topotecan), or the ferroptosis- (GCSH), dihydrolipoamide S-succinyltransferase lipoylation. FDX1 knockout resulted in com-
inducer ML162 (fig. S5, L to O). Of note, the (DLST), and DLAT, an essential component of plete loss of protein lipoylation, as measured by
strong connection between copper-mediated the PDH complex. Lipoylation of these proteins either immunoblot or immunohistochemistry
cell death and both FDX1 expression and pro- is known to be required for enzymatic function (Fig. 4D and fig. S6, E to I), and also led to a
tein lipoylation was lost at high concentra- (30) (Fig. 3D). Our findings that knockout of significant drop in cellular respiration similar
tions of elesclomol (>40 nM) (fig. S4, L to P), either FDX1 or lipoylation-related enzymes to the levels observed with the deletion of LIAS
thereby suggesting that off-target mechanisms rescues cells from copper toxicity led us to ex- itself (Fig. 4E and fig. S6J). Furthermore, me-
of cell death may occur at such concentra- plore whether FDX1 might be an upstream tabolite profiling after deletion of FDX1 led to an
tions and possibly explaining conflicting mech- regulator of protein lipoylation. To test this accumulation of pyruvate and a-ketoglutarate
anisms of action reported in the literature hypothesis, we performed three analyses. First, and depletion of succinate, as would be expected
(5, 7, 14–16, 19, 28, 29). Moreover, whereas we looked for evidence of coordinated depen- when protein lipoylation is compromised be-
the most copper-selective compounds (e.g., dencies across the Cancer Dependency Map cause of inhibition of the TCA cycle at PDH and
elesclomol, disulfiram, and NSC319726) lost (www.depmap.org), a resource of genome-wide a-ketoglutarate dehydrogenase (31) (Fig. 4F and
killing activity when cells were grown under CRISPR-Cas9 knockout screens in hundreds table S3). Also of interest, we observed an in-
glycolytic conditions, compounds with more of cancer cell lines. Genes showing similar crease in S-adenosylmethionine (SAM), a key
promiscuous metal-binding compounds (e.g., patterns of viability effects, even if subtle, sug- substrate of LIAS in the lipoic acid pathway, con-
pyrithione and 8-HQ) killed independent of gest that they have shared function or regu- sistent with FDX1 being a previously unrecog-
metabolic state. This result is consistent with lation. Notably, FDX1 and components of the nized upstream regulator of protein lipoylation.
copper’s distinctive connection to mitochon- lipoic acid pathway were highly correlated in
drial metabolism-mediated protein lipoylation. their viability effects across the cell line panel Copper directly binds and induces
(Fig. 2A and figs. S3, A to F, and S5, A to G). (p < 0.0001; Fig. 4A). the oligomerization of lipoylated DLAT
Second, we performed immunohistochem- The experiments described above establish a
FDX1 is an upstream regulator ical staining for FDX1 and lipoic acid in 208 connection between copper toxicity and pro-
of protein lipoylation human tumor specimens and semiquantitative tein lipoylation but do not establish a direct
Protein lipoylation is a highly conserved lysine light-microscopic scoring by two independent mechanistic link. We hypothesized that cop-
posttranslational modification that is known to pathologists. Expression of FDX1 and lipoylated per might directly bind to lipoylated proteins,

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 3 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

A B C
Cu-DDC Elesclomol-Cu
LA pathway 4 MTF1 LIPT1 LIPT2 4
Whole genome
single gene FDX1 LIAS LIAS
CRISPR/Cas9 KO CDKN2A PDHB FDX1
LIAS PDHB
CDKN2A PDHA1
LIPT1

-LOG10(q-value)
3 GLS 3 LIPT1

-LOG10(q-value)
DLD DLD

PDH complex MTF1 DLD

Cupric Elesclomol DLAT 2 FDX1 2 GLS
DDC Copper
PDHA1 GCSH MPC1
PDHB
MTF1 1 1
GLS
CDKN2A
n=147 10 n=35 0 0
Positive hits
-2 0 2 -2 0 2 4
Negative hits
Shared genetic FDR< 0.01
Log fold change Log fold change
regulators
D E F 1.0
ABC1 cells
Ch2-2
LA pathway

Fold viability
0.8 FDX1 KO #1
Complex I FDX1 KO #2
0.6
40nM Elesclomol-Cu LIAS KO #1
2.5 0.4 LIAS KO #2
Fatty
Acid
DLAT 0.2
Synthesis
DLD
NH
2.0 FDX1 0.0

Downloaded from https://www.science.org on June 12, 2024

-LOG10(q-value)
O
O -10 -9 -8 -7 -6
DBT
NH2
OH
octanoic acid
S NH TCA Cycle activity LOG10[Elesclomol-Cu], M
1.5 LIAS G
S O
NH2 Mitochondrial respiration
DLAT NH
Lipoylation O
S
BCAA metabolism
S 80 20nM Elesclomol OVISE cells
GCSH
NH2
LIAS S
NH
O 1.0 AAVS1 KO

% confluency
NH2 S

DLST LIPT1 60 FDX1 KO #2

S
S
DLD 0.5 40 LIAS KO #1
Inactive LIPT2 Active LIAS KO #2
essential for
complexes complexes cuproptosis 20
0.0 0
-6 -4 -2 0 2 4 6 0 20 40 60 80 100
Log fold change Hours

Fig. 3. FDX1 and lipoic acid genes are critical mediators of copper ionophore– branched-chain amino acid. (E) Summary scatter plot indicating top hits in
induced cell death. (A) Whole-genome CRIPSR-Cas9 positive selection screen the metabolism gene–focused CRISPR-Cas9 gene knockout screen of A549 cells
using two copper ionophores (Cu-DDC and elesclomol-copper) in OVISE cells treated with 40 nM of elesclomol-Cu(II) (1:1 ratio). Genes associated with the
(schematic on the left). Overlapping hits with a false discovery rate (FDR) score lipoic acid pathway are marked in blue, complex I–related genes in orange,
<0.01 were analyzed (right). Positive hits (resistance) are marked in blue, and FDX1 in purple. (F) Viability of ABC1 cells with CRIPSR-Cas9 deletion of LIAS
and negative hits (sensitizers) are marked in red. (B and C) Summary scatter or FDX1 after treatment with elesclomol in the presence of 1 mM CuCl2 in the
of the results of the screen in (A) for Cu-DDC (B) or elesclomol-Cu (C). media. (G) Growth curve measurements of OVISE cells with CRIPSR-Cas9
(D) Schematic of the lipoic acid pathway. Genes that scored in our genetic deletion of LIAS and FDX1 in the presence of 20 nM elesclomol. For (F) and
screens are marked as essential for copper-induced cell death. BCAA, (G), data are means ± SD, with n ≥ 3.

a possibility suggested by the observation that tion of DLAT that was detectable by nonde- after exposure to copper ionophores is mediated
copper binds free lipoic acid with a measured naturing gel electrophoresis (Fig. 5, B and C). at least in part by their aberrant oligomeriza-
dissociation constant of 10−17 (32). To test this Similarly, treatment of elesclomol-sensitive tion. Mass spectrometric analysis also revealed
hypothesis, we purified DLAT and DLST from cells increased levels of DLAT oligomers and in- that copper ionophore treatment leads to loss
cell lysates and found that these proteins bound soluble DLAT, whereas treatment of elesclomol- of Fe-S cluster proteins (Fig. 6, A and B) in an
to copper-charged resin but not to cobalt or insensitive cell lines or FDX1 knockout (FDX1 FDX1-dependent manner (Fig. 6C) as well as
nickel resins (Fig. 5A). When protein lipoylation KO) cells resulted in DLAT oligomerization only the induction of proteotoxic stress (Fig. 6, B
was abrogated by FDX1 deletion (Fig. 4), DLAT at much higher concentrations (Fig. 5, D and E, and C; fig. S8A; and table S4). These findings are
and DLST no longer bound copper (Fig. 5B), and fig. S7, A and B). Treatment with the strong consistent with the observation in bacteria and
suggesting that the lipoyl moiety is required reducing agent tris(2-carboxyethyl)phosphine yeast that copper can destabilize Fe-S–containing
for copper binding. (TCEP) and boiling eliminated the oligomeric proteins (33–35). Whether such loss of Fe-S clus-
We noted that copper binding to lipoylated form of DLAT (fig. S7, C to E), suggesting that ter proteins contributes to the copper ionophore
proteins did not simply lead to loss of function, the aggregates are disulfide-bond dependent. We death phenotype remains to be determined.
given that deletion of the proteins rescues confirmed these findings by immunofluores-
(not phenocopies) copper ionophore treatment. cence, observing pronounced induction of DLAT Copper-induced death mechanisms are
We proposed that copper binding to lipoylated foci by short-pulse elesclomol treatment, where- shared by genetic models of copper
TCA cycle proteins results in a toxic gain of as such foci were diminished in FDX1 KO, homeostasis dysregulation
function. Interestingly, we found that the cop- lipoylation-deficient cells (Fig. 5, F to H, and The experiments described to this point used
per binding to lipoylated TCA cycle proteins fig. S7F). These findings support a model where copper ionophores to overcome the homeosta-
resulted in lipoylation-dependent oligomeriza- the toxic gain of function of lipoylated proteins tic mechanisms that normally keep intracellular

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 4 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

Downloaded from https://www.science.org on June 12, 2024

Fig. 4. FDX1 is an upstream regulator of protein lipoylation. (A) Correlation NSCLC with correlated low (top row) and high (bottom row) expression of LA and
analysis of gene dependencies taken from the Achilles project. Presented is the gene FDX1 by IHC. (D) Immunoblot of lipoylated proteins, FDX1, DLAT, and actin from
network that correlates with FDX1 deletion using FIREWORKS (51). The correlating extracts of PSN1 cells with deletion of FDX1. (E) Basal OCR as measured in PSN1 cells
genes are marked by their described functionality (lipoic acid pathway, and with CRIPSR-Cas9 deletion of FDX1, LIAS, or AAVS1 genes (unpaired t test, ***p <
mitochondria complex I and Fe-S cluster regulation). (B) A tissue microarray of non– 0.001). (F) Plot of the average log2 fold change (FC) in metabolites between FDX1 KO
small cell lung carcinoma (NSCLC) (n = 57) was stained with lipoic acid (LA), and AAVS1 control K562 cell lines separated by functional annotations. Metabolites
and FDX immunohistochemistry (IHC) and expression was scored semiquantitatively relevant to the lipoic acid pathway are marked in orange. a-KDH, a-ketoglutarate
by two pathologists (S.C. and S.S.), showing a strong direct correlation between dehydrogenase; a-KG, a-ketoglutarate; NADH, reduced nicotinamide adenine
LA and FDX expression (mean ± SD; p < 0.0001). (C) Representative cases of dinucleotide; PPP, pentose phosphate pathway.

copper concentration low. These mechanisms creased levels of HSP70 (Fig. 6F). Importantly, mice with those of Atp7b heterozygous (Atp7b–/+)
include the copper importer SLC31A1 (CTR1) ferroptosis, necroptosis, and apoptosis inhibitors and wild-type control mice, we observed loss of
and the copper exporters ATP7A and ATP7B, did not prevent copper-induced cell death in cells lipoylated and Fe-S cluster proteins, as well as an
which are encoded by genes that are mutated overexpressing SLC31A1 (Fig. 6G and fig. S8, D increase in Hsp70 abundance (Fig. 6K and table
in the copper dysregulation syndromes Menke’s to F), whereas copper chelators, FDX1 KO, and S5). These findings in mouse models of copper
disease and Wilson’s disease, respectively (3, 36) LIAS KO each partially rescued cells from copper- toxicity suggest that copper overload results in
(Fig. 6D). To explore whether the mechanisms induced cell death (Fig. 6, G and H, and fig. S8, the same cellular effects as those induced by cop-
of copper toxicity associated with copper iono- D to G). In addition, depletion of the natural per ionophores. Taken together, our data support
phore treatment are shared by these naturally intracellular copper chaperone glutathione re- a model whereby excess copper promotes the ag-
occurring disorders of copper homeostasis, we sulted in copper-dependent cell death (Fig. 6I), gregation of lipoylated proteins and destabilization
examined three experimental models. First, we which was associated with reduced lipoylation of Fe-S cluster proteins that results in proteotoxic
overexpressed SLC31A1 in human embryonic and increased DLAT oligomerization (fig. S8, stress and ultimately cell death (Fig. 6L).
kidney (HEK) 293T and ABC1 cells, which dra- I and J) that was attenuated by FDX1 and LIAS
matically increased sensitivity to physiological KO, just as was observed with copper ionophore Discussion
concentrations of copper (Fig. 6E and fig. S8B). treatment of SCL31A1 wild-type cells (Fig. 6J Copper is a double-edged sword: It is essential
Consistent with our findings using copper ion- and fig. S8H). as a cofactor for enzymes across the animal
ophores in wild-type cells, copper supplementa- Lastly, we used a mouse model of Wilson’s kingdom, and yet even modest intracellular
tion resulted in an overall reduction in proteins disease in which Atp7b deletion leads to intra- concentrations can be toxic, resulting in cell
involved in mitochondrial respiration (fig. S8C cellular copper accumulation and cell death death (2). Genetic variation in copper homeo-
and table S4), reduced protein lipoylation, re- with increasing animal age (37, 38). Compar- stasis results in life-threatening disease (39, 40),
duced levels of Fe-S cluster proteins, and in- ing the livers of aged Atp7b-deficient (Atp7b−/−) and both copper ionophores (11, 17, 41, 42) and

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 5 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

A B C D
Input (1/10) Elution ABC1 cells A549 cells
A673 cells
Elution

KO

KO
(sensitive) (resistant)

1
Cell lysate Cell lysates

VS

VS
X1

X1
ES nM - 10 20 40 100 200
Cu Co Ni Input ES nM

AA

FD

AA

FD
AAVS1 FDX1 KO (2hr pulse) - - 10 40 100
(2hr pulse) 10 40 100

oligomers
Lip-DLST

oligomers
oligomers
DLST 250 250 250
Cu Co Ni Cu Cu
150
150 150
DLAT 100
100
Wash x 2 100 75
Wash x 2
75
Elute 75 IB : DLAT
Elute
IB : DLAT IB : DLAT

E F G H
Control 10nM ES (pulse) Control 10nM ES (pulse)
AAVS1 FDX1 KO DLAT foci number
ES nM
- 1 10 40 -

AAVS1
(2hr pulse) 1 10 40
AAVS1

Normalized foci count

AAVS1

5 ***
oligomers

FDX1 KO ***
250
4
150
***
3
100

FDX1 KO
2
FDX1 KO

75
1

Downloaded from https://www.science.org on June 12, 2024

DLAT
IB : DLAT Hoeschst
Mitotracker 0
Phalloidin ES nM - 10 40 100
(2hr pulse)

Fig. 5. Copper directly binds and promotes the oligomerization of lipoylated that were pulse-treated with the indicated concentrations of elesclomol.
DLAT. (A) The binding of the indicated proteins to copper (Cu), cobalt (Co), and (E to H) AAVS1 control or FDX1 KO ABC1 cells were pulse-treated with the
nickel (Ni) was assessed by immunoblot analysis of eluted proteins from the indicated concentrations of elesclomol for 2 hours; protein oligomerization was
indicated metal-loaded resins. (B) Copper binding was assessed by loading analyzed after 24 hours by immunoblotting (E) and both wide-field (F)
cell lysates from either ABC1 AAVS1 or FDX1 KO cells on copper-loaded resin and confocal (G) immunofluorescence imaging (green, DLAT–; red, Mitotracker;
followed by washing and analysis of the eluted proteins. Input and eluted blue, Hoechst; white, phalloidin). (H) Foci were segmented and quantified
proteins are presented. IB, immunoblot. (C) Protein content was analyzed in in each condition [n = 3, and multiple unpaired t test analysis was conducted
A673 cells that were pulse-treated (2 hours) with the indicated concentrations with the desired FDR(Q) = 1%; ***q < 0.001]. For (C) to (H), media were
of elesclomol. (D) Protein content was analyzed in ABC1 and A549 cells supplemented with 1 mM CuCl2.

copper chelators (43–46) have been suggested that we describe here in human cancer cells larly sensitive to copper ionophores and that
as anticancer agents. However, the mechanism (lipoylation and Fe-S cluster proteins) are evo- this sensitivity is explained by their high levels
by which copper overload leads to cell death has lutionarily conserved from bacteria to humans, of lipoylated TCA enzymes. Moreover, our ob-
been obscure. Here, we have shown that copper suggesting that copper-induced cell death servations that the abundance of FDX1 and
toxicity occurs by a mechanism distinct from might be also used by microorganisms where lipoylated proteins is highly correlated across
all other known mechanisms of regulated cell copper ionophores are naturally synthesized a diversity of human tumors and that cell lines
death, including apoptosis, ferroptosis, pyrop- and exhibit antimicrobial activity (48, 49). with high levels of lipoylated proteins are sensi-
tosis, and necroptosis. We therefore propose In the case of genetic disorders of copper ho- tive to copper-induced cell death suggest that
that this previously uncharacterized cell death meostasis (Wilson’s disease and Menke’s dis- copper ionophore treatment should be directed
mechanism be termed cuproptosis. ease), copper chelation is an effective form of toward tumors with such a metabolic profile.
We have shown that copper-induced cell therapy (50). In cancer, however, the exploita- Future clinical trials of copper ionophores using
death is mediated by an ancient mechanism: tion of copper toxicity has been less successful a biomarker-driven approach should therefore
protein lipoylation. Notably few mamma- (42). Copper ionophores, including elesclomol, be considered.
lian proteins are known to be lipoylated, have been tested in clinical trials, but such test-
and these are concentrated in the TCA cycle, ing occurred without the benefit of either a bio- REFERENCES AND NOTES
where lipoylation is required for enzymatic marker of the appropriate patient population 1. B. E. Kim, T. Nevitt, D. J. Thiele, Nat. Chem. Biol. 4, 176–185 (2008).
function (26, 47). Our study explains the rela- or an understanding of the drug’s mechanism 2. E. J. Ge et al., Nat. Rev. Cancer 22, 102–113 (2022).
3. S. Lutsenko, Curr. Opin. Chem. Biol. 14, 211–217 (2010).
tionship between mitochondrial metabolism of action. A phase 3 combination clinical trial of
4. T. D. Rae, P. J. Schmidt, R. A. Pufahl, V. C. Culotta,
and the sensitivity to copper-mediated cell elesclomol in patients with melanoma showed T. V. O’Halloran, Science 284, 805–808 (1999).
death: Respiring, TCA-cycle active cells have a lack of efficacy in this unselected population, 5. P. Tsvetkov et al., Nat. Chem. Biol. 15, 681–689 (2019).
increased levels of lipoylated TCA enzymes but a post hoc analysis of patients with low plas- 6. B. B. Hasinoff, A. A. Yadav, D. Patel, X. Wu, J. Inorg. Biochem.
137, 22–30 (2014).
(in particular, the PDH complex), and the lipoyl ma lactate dehydrogenase (LDH) levels showed
7. S. Tardito et al., J. Am. Chem. Soc. 133, 6235–6242 (2011).
moiety serves as a direct copper binder, re- evidence of antitumor activity (42). Low LDH 8. E. W. Hunsaker, K. J. Franz, Inorg. Chem. 58, 13528–13545 (2019).
sulting in lipoylated protein aggregation, loss reflects a higher cellular dependency on mito- 9. V. Oliveri, Coord. Chem. Rev. 422, 213474 (2020).
of Fe-S cluster–containing proteins, and induc- chondrial metabolism (as opposed to glycolysis), 10. S. M. Corsello et al., Nat. Cancer 1, 235–248 (2020).
11. D. Cen, D. Brayton, B. Shahandeh, F. L. Meyskens Jr.,
tion of HSP70, reflective of acute proteotoxic consistent with our finding that cells under- P. J. Farmer, J. Med. Chem. 47, 6914–6920 (2004).
stress. The targets of copper-induced toxicity going mitochondrial respiration are particu- 12. J. R. Kirshner et al., Mol. Cancer Ther. 7, 2319–2327 (2008).

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 6 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

A B C AAVS1 FDX1 KO
A673 cells ES nM
ES nM
Treatment Protein mass-spec Analysis - 1 10 40 - 1 10 40
(2hr pulse) - 10 40 100 200 (2hr pulse)
ES Lip-DLAT

new media
LOG2FC
(Treated/Control) Lip-DLAT

Wash
Pulse
Lip-DLST
2 hrs 24 STRING: functional Lip-DLST
hrs enrichment FDX1

Fe-S Cluster proteins

Fe-S Cluster proteins

Molecular function (GO) FDX1
LIAS LIAS
top molecular function (GO)

LOG2FC(ES100/Cont)
6 ACO-2 ACO-2
Description Enrich score FDR 4 Fe-S Cluster SDHB SDHB
4Fe-4S cluster 1.7046 downregulated genes
0.0014 (ks) 2
binding POLD1
4: DNA2 140: RTEL1 POLD1
Fe-S cluster 0.0022 (ks) 0 38: NTHL1 158: ELP3
1.30855
binding 44: POLE 171: CDK5RAP1 NDUFB8 DPYD
58: POLD1 233: FDX1
-2 78: CISD1 250: ISCA2
94: GLRX5 298: EFTDH TOMM20 NDUFB8
102: PPAT
-4
er ACTIN TOMM20
l

t
Al

us
cl

HSP70 HSP70
-S
Fe

D E F G H
Cu GFP OE SLC31A1 OE HEK293T SLC31A1
ATP7A/B SLC31A1 HEK293T
- - .5 1 2 4
****
Cu Lip-DLAT ****
1.2 0.6
1.2

Fold Viability
Lip-DLST

Fold Viability
Fold Viability

Cu-GSH 0.9
0.9 0.4
GFP SDHB
0.6
BSO SLC31A1 POLD1 0.6 0.2

Downloaded from https://www.science.org on June 12, 2024

Glu+Cys GSH 0.3
0.3
0.0 FDX1
0.0
-8 -7 -6 -5 -4 0.0
( HSP70 KO

X1
-2

AS
LOG10, [CuCl2, M]

h2
D

FD
-1
N -1
S
M

LI
- -

VA
ec

C
BC

r
TT
TOMM20

Fe

Z-
VIN

I K L
elesclomol
+ BSO No BSO Cu+
- 1 5 10 50 - 1 5 10 50 Cu2+ ATP7B SLC31A1
4 Liver
CuCl2 **
FeCl2 1.0
CoCl2 Cu+ GSH lomol]
Control Fe [elesc
S Fe S
relative intensitiy (AU)

ZnCl2 0.5 3 ATP7B -/-

MnCl2 Fe Fe
Fe-S cluster proteins Lipoylated proteins S
NiCl2 Cell Death
0
J HEK293T + BSO
2 *** *** *** *** ns *** *** *** ** ns ns ns ns Cu2+ Cu+ GSH by
cuproptosis
1.0 FDX1
Control
Fold viability

0.8 FDX1 KO DLAT

LIAS KO
1 anti-DLAT
0.6 LIAS S
S DLAT
0.4 aggregation
0.2 0
essential for promotes inhibits FeS Cluster
70
AS

PP B
FD T
X1

D 2
D T
G T
SH

D T
T
H
D 0
8

0.0
A

LA
LA
Li CO

2
FB
B
LS
H

TO PD

cuproptosis cuproptosis cuproptosis Proteins

SP

M
D
LI
SD

-6.5 -6.0 -5.5 -5.0 -4.5

U
M
A
H

LOG10, [CuCl2, M]
N

Fig. 6. Shared mechanisms in chemically and genetically induced copper- SLC31A1 was analyzed 72 hours after treatment with 2 mM CuCl2. In (G) and
dependent cell death. (A) Proteomic analysis of control and elesclomol (H), viability is presented as box-and-whisker plots representing interquartile
(100 nM) pulse-treated ABC1 cells 24 hours after treatment. Top [gene ontology ranges (boxes), medians (horizontal lines), and range (whiskers). For both,
(GO)] enriched categories are presented (table S4). (B and C) Protein content in n ≥ 5, and an ordinary one-way analysis of variance (ANOVA) with multiple
A673 cells (B) and AAVS1 and FDX1 KO ABC1 cells (C) 16 hours after 2-hour comparisons was conducted (****adjusted p < 0.0001). (I) Viability of control
pulse treatment with elesclomol. For (A) to (C), media were supplemented with and 100 mM BSO–pretreated A549 cells 48 hours after treatment with the
1 mM CuCl2. (D) Copper homeostasis schematic. (E and F) HEK293T cells indicated metals. An average of three replicates is plotted. The color scale
overexpressing green fluorescent protein (GFP) or SLC31A1 were analyzed for represents the relative viability. (J) Viability of control, FDX1 KO, and LIAS KO
viability 72 hours after supplementation with CuCl2 (E) and for protein content HEK293T cells pretreated with 20 mM BSO and then with CuCl2. (K) Protein
24 hours after treatment (F). (G) Cells overexpressing SLC31A1 were pretreated content in Atpb7b−/− mouse livers (≥5 Atpb7b−/− mice) compared with control
with 10 mM necrostatin-1, 10 mM ferrostatin-1, 100 mM a-tocopherol, 40 mM (six Atpb7b+/− mice and one wild-type mouse). Multiple unpaired t test analysis
Z-VAD-FMK, 10 mM TTM, or 40 mM bathocuproinedisulfonic acid (BCS). Viability was conducted; ***q < 0.001, **q < 0.01, and ns is not significant. AU, arbitrary
was measured after 72 hours of cells growing in the presence or absence of 3 mM units. (L) Schematic of mechanisms that promote copper-induced cell death.
CuCl2. (H) The viability of ABC1 Ch2-2, FDX1, and LIAS KO cells overexpressing For (E) and (J), data are means ± SD, with n ≥ 4.

13. S. Tardito et al., J. Med. Chem. 55, 10448–10459 (2012). 18. N. Liu, H. Huang, Q. P. Dou, J. Liu, Oncoscience 2, 457–466 (2015). 22. R. Weinlich, A. Oberst, H. M. Beere, D. R. Green, Nat. Rev. Mol.
14. M. Nagai et al., Free Radic. Biol. Med. 52, 2142–2150 (2012). 19. Z. Skrott et al., Nature 552, 194–199 (2017). Cell Biol. 18, 127–136 (2017).
15. K. Shimada et al., Cell Chem. Biol. 25, 585–594.e7 (2018). 20. D. Tang, R. Kang, T. V. Berghe, P. Vandenabeele, G. Kroemer, 23. T. Bergsbaken, S. L. Fink, B. T. Cookson, Nat. Rev. Microbiol. 7,
16. N. C. Yip et al., Br. J. Cancer 104, 1564–1574 (2011). Cell Res. 29, 347–364 (2019). 99–109 (2009).
17. D. Chen, Q. C. Cui, H. Yang, Q. P. Dou, Cancer Res. 66, 21. B. A. Carneiro, W. S. El-Deiry, Nat. Rev. Clin. Oncol. 17, 395–417 24. S. J. Dixon et al., Cell 149, 1060–1072 (2012).
10425–10433 (2006). (2020). 25. S. Elmore, Toxicol. Pathol. 35, 495–516 (2007).

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 7 of 8

 Erratum 22 April 2022. See Erratum.
RES EARCH | R E S E A R C H A R T I C L E

26. A. Solmonson, R. J. DeBerardinis, J. Biol. Chem. 293, 47. J. A. Mayr, R. G. Feichtinger, F. Tort, A. Ribes, W. Sperl, and S.S. performed the tissue microarray staining scoring and
7522–7530 (2018). J. Inherit. Metab. Dis. 37, 553–563 (2014). visualization, and A.V. assisted with the microscopy analysis. B.P. and
27. K. Birsoy et al., Cell 162, 540–551 (2015). 48. J. B. Patteson et al., Science 374, 1005–1009 (2021). N.K. performed and analyzed the metabolomics experiments. S.L.
28. N. L. Reeder et al., Antimicrob. Agents Chemother. 55, 49. N. Raffa et al., Proc. Natl. Acad. Sci. U.S.A. 118, e2015224118 provided study material and experimental advice. J.K.E. and E.F.
5753–5760 (2011). (2021). provided reagents and experimental advice. T.R.G. supervised
29. A. A. Yadav, D. Patel, X. Wu, B. B. Hasinoff, J. Inorg. Biochem. 50. A. Aggarwal, M. Bhatt, Tremor Other Hyperkinet. Mov. (N. Y.) 8, the research. P.T. and T.R.G. wrote the manuscript. Competing
126, 1–6 (2013). 525 (2018). interests: S.M.C. and T.R.G. receive research funding unrelated to
30. E. A. Rowland, C. K. Snowden, I. M. Cristea, Curr. Opin. 51. D. R. Amici et al., Life Sci. Alliance 4, e202000882 this project from Bayer HealthCare and Calico Life Sciences. T.R.G.
Chem. Biol. 42, 76–85 (2018). (2020). receives research funding related to this project from Novo Holdings,
31. M. Ni et al., Cell Rep. 27, 1376–1386.e6 (2019). recently held equity in FORMA Therapeutics, is a consultant to
32. J. Smirnova et al., Sci. Rep. 8, 1463 (2018). ACKN OWLED GMEN TS GlaxoSmithKline and Anji Pharmaceuticals, and is a founder
33. D. Brancaccio et al., J. Am. Chem. Soc. 139, 719–730 We thank J. Markley, H. Adelmann, M. Slabicki, V. Wang, and of Sherlock Biosciences. S.S. is a consultant for RareCyte, Inc.
(2017). H. Keys for constructive discussion and help with genetic screens. P.T. and T.R.G. are inventors on the patent application PCT/US21/
34. S. Chillappagari et al., J. Bacteriol. 192, 2512–2524 We thank T. Woo for help with immunohistochemistry and 19871 submitted by the Broad Institute entitled “Method of treating
(2010). immunofluorescence studies and A. Muchenditsi for help with the cancer.” J.E. is currently an employee of Kojin Therapeutics.
35. L. Macomber, J. A. Imlay, Proc. Natl. Acad. Sci. U.S.A. 106, Atp7b mice. We thank C. Lewis (Whitehead metabolomics), The other authors declare no competing interests. Data and
8344–8349 (2009). W. Salmon (Whitehead imaging), and R. Rodrigues (Harvard Medical materials availability: All data are available in the manuscript
36. T. Nevitt, H. Ohrvik, D. J. Thiele, Biochim. Biophys. Acta 1823, School–Thermo Fisher Scientific Center for Multiplexed Proteomics or the supplementary materials.
1580–1593 (2012). proteomic facility) for technical help. We thank M. O’Reilly for help
37. S. Lutsenko, Biochem. Soc. Trans. 36, 1233–1238 (2008). with the design of the figures. We thank the Image and Data Analysis SUPPLEMENTARY MATERIALS
38. A. Muchenditsi et al., Sci. Rep. 11, 5659 (2021). Core at Harvard Medical School for coding support. Funding: This
science.org/doi/10.1126/science.abf0529
39. O. Bandmann, K. H. Weiss, S. G. Kaler, Lancet Neurol. 14, work was supported by National Cancer Institute (NCI) grant 1 R35
Materials and Methods
103–113 (2015). CA242457-01 (T.R.G.), Novo Holdings (T.R.G. and P.T.), NCI grant
Figs. S1 to S8
40. E. Gaggelli, H. Kozlowski, D. Valensin, G. Valensin, Chem. Rev. K08 CA230220 (S.M.C.), National Institute of General Medical
Tables S1 to S5
106, 1995–2044 (2006). Sciences grant T32-GM007748 (S.C.), research and recruitment
References (52–63)
41. S. O’Day et al., J. Clin. Oncol. 27, 5452–5458 (2009). funding by Boston Children’s Hospital (B.P. and N.K.), NCI grant
MDAR Reproducibility Checklist
42. S. J. O’Day et al., J. Clin. Oncol. 31, 1211–1218 (2013). R01-CA194005, NCI grant U54-CA225088, and the Ludwig Center at
43. L. Cui et al., Nat. Biotechnol. 39, 357–367 (2021). Harvard (S.S.). Author contributions: P.T. conceptualized the View/request a protocol for this paper from Bio-protocol.
44. T. Tsang et al., Nat. Cell Biol. 22, 412–424 (2020). project, conducted experiments, collected data, and analyzed results.

Downloaded from https://www.science.org on June 12, 2024

45. C. I. Davis et al., Metallomics 12, 1995–2008 (2020). M.A. and M.D. assisted with experiments. L.J.-C., J.R., and M.K. 30 September 2020; resubmitted 18 January 2022
46. D. C. Brady, M. S. Crowe, D. N. Greenberg, C. M. Counter, assisted with data analysis. R.H. and R.D.S. performed the whole- Accepted 4 February 2022
Cancer Res. 77, 6240–6252 (2017). genome CRISPR-Cas9 screens under supervision from S.M.C. S.C. 10.1126/science.abf0529

Tsvetkov et al., Science 375, 1254–1261 (2022) 18 March 2022 8 of 8