Introduction

Copper (Cu) is an essential micronutrient due to its role as a cofactor in a myriad of key proteins. Cuproproteins drive diverse physiological functions, including respiration (cytochrome c oxidase), superoxide scavenging (Cu/zinc superoxide dismutase), methanotrophy (methane monooxygenase) and nitrogen cycling (nitrite reductase, azurin)1,2. The binding of Cu in these distinct structures controls their folding and function. The most common Cu-binding ligands in proteins are cysteine and histidine, which both occur in roughly two-thirds of experimentally-resolved Cu-binding sites1. Cysteine coordinates Cu with the sulfur (S) atom of a thiol while histidine ligates Cu with a nitrogen (N) atom in its imidazole ring. Oxygen (O) ligands are comparatively less common but still prevalent: glutamate and aspartate, which bind Cu with O atoms in their carboxyl side chains, occur in 7% and 6% of Cu-binding sites, respectively. Bonds between Cu and these ligands (S, N, O) differ in their intramolecular vibrational frequencies and, consequently, their energy, strength, and stability3,4,5.

Advances in mass spectrometry6,7 have facilitated the use of Cu isotopes as tracers of biological function3,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30. Copper has two stable isotopes with masses of 63 and 65 amu, respectively. In animals, these isotopes are fractionated among different tissue and fluid types3. Because fractionation patterns differ with disease state3, Cu isotope ratios has been proposed as diagnostic indicators of Alzheimer’s disease20,21, breast, ovarian and colorectal cancer27,28 as well as multiple diseases of the liver9,26, 29, 31. However, development of such diagnostics is presently hampered by numerous confounding variables including organisms’ sex8,10, age13,19, diet32, and menopausal state15,33.

Developing robust biomarkers for applications in human health, among other potential applications in the geosciences, requires a clear model of the mechanisms which drive biological Cu isotope fractionation. Previous studies have demonstrated significant variations in Cu isotope ratios in cuproproteins, which are hypothesized to be caused by differences in Cu binding site configurations17,34. Ab initio calculations based on density functional theory (DFT) predict that the identity and geometry of the coordinating amino acids at protein Cu-binding sites determine their relative affinity for 63Cu versus 65Cu4. Accordingly, the differential expression of structurally-distinct proteins would influence the isotopic composition of a cell or tissue. While numerous studies have relied on ab initio estimates to explain variations in Cu isotope ratios measured in biological samples8,9, 27, there is currently no experimental data to test or validate these theoretical predictions.

In this study, Cu2+ isotope fractionation by amino acid ligation was quantified in laboratory experiments using Donnan dialysis to separate bound versus free Cu2+ and multicollector inductively coupled plasma mass spectrometry (MC-ICP-MS) to measure isotope abundance. We conducted experiments with the metal-binding amino acids cysteine, histidine, glutamate, and aspartate. Our objective was to experimentally determine the (mass-dependent) isotope effects associated with Cu2+ ligation by these amino acids and to provide direct evidence that distinct protein metal-binding sites differentially fractionate Cu isotopes. The results of this work offer insight into the biomolecular mechanisms which underpin Cu isotope fractionation among proteins and at higher levels of biological systems.

Materials and methods

Chemical reagents

All reagents were prepared using ultrapure deionized water (18.2 MΩ·cm, Milli-Q, Millipore). High purity 4-morphlineethanesulfonic (MES) hydrate (BioUltra, Sigma-Aldrich) was used to buffer reactions. Trace analysis quality (> 99.995%) strontium nitrate (Sr(NO3)2) was used as a background electrolyte35,36. Ligands were reagent-grade (> 95%) or higher: l-cysteine (reagent grade, Sigma-Aldrich), l-histidine (BioUltra [< 99.5%], Sigma-Aldrich), l-aspartic acid (reagent grade, Sigma-Aldrich), and l-glutamic acid (reagent grade, Sigma-Aldrich). A Cu nitrate standard (Sigma-Aldrich) was used as the Cu source in all experiments. All solutions were prepared and stored in metal-free centrifuge tubes (Labcon), which were soaked for 2 days in 20% nitric acid (HNO3) and rinsed with ultrapure water prior to use. Following dialysis, samples for Cu analysis were stored in clean 2 mL microcentrifuge tubes (2 days, 20% HNO3) at 4 °C to minimize evaporation.

Cu ligation experiments

Copper stock solutions (100 µM) contained 10 mM Sr(NO3)2 as a background electrolyte and 5 mM MES hydrate as a buffer, and were adjusted to pH 6.0 using 1 M sodium hydroxide (NaOH). Copper solutions were mixed with either cysteine, histidine, aspartate, or glutamate and allowed to equilibrate overnight (≥ 16 h) under ambient laboratory conditions. All Cu ligation experiments were conducted with 100 µM amino acids except cysteine. A ratio of 100 µM Cu and 50 µM cysteine was used in cysteine experiments to avoid quantitative Cu binding. Limiting disulfide bridging, the cysteine-Cu solution was made under anoxic conditions using N2-purged ultrapure water. Previous X-ray absorption near edge structure (XANES) analyses have shown that when Cu2+ is reacted with cysteine, the oxidation state of Cu remains as 2+37.

After the complexation reaction was complete, free Cu2+ was separated from amino acid-bound Cu2+ via an equilibrium Donnan dialysis procedure adapted from Nolan et al.38 and Ryan et al.36. Briefly, two solutions were added to opposing 1.5 ml cells in Teflon micro-equilibrium dialyzers (Harvard Apparatus): (1) a Cu solution containing dissolved amino acids, and (2) an acceptor solution containing 10 mM Sr(NO3)2 and 5 mM MES hydrate, adjusted to pH 6.0. These cells were separated by a strong-acid cation-exchange membrane (Nafion-117, E.I. Dupont de Nemours) which permits passage of free but not bound Cu2+. Free Cu2+ thus equilibrates across the membrane and accumulates in the acceptor solution, while bound Cu is retained in the donor solution.

Before dialysis, Teflon dialyzers were soaked overnight in 2% HNO3 and thoroughly rinsed with ultrapure deionized water. Membranes were rinsed and stored in ultrapure deionized water for a minimum of 2 days. Immediately prior to use, membranes were pre-conditioned by two successive soaks in 10 mM Sr(NO3)2 for a minimum of 2 h each. The dialysis system was then assembled and, as a final pre-conditioning step, filled with acceptor solution adjusted to pH 6 for an additional 2 h. This solution was then removed and Cu separation was performed by adding the Cu2+ donor solution and fresh acceptor solution to the opposing cell in the dialyzer. Dialysis was carried out for approximately 27 h. No-ligand control experiments indicated that dialysis reached mass and isotopic equilibrium within 1 day (Suppl. Fig. 1; Table 1). Positive control experiments performed with the amino acid glycine, which has no side chain, confirmed that Cu-binding by carboxyl and amine terminal groups was negligible under the reaction conditions (Suppl. Fig. 2). At the end of each experiment, samples of initial (t = 0) and final (t = f) acceptor and donor solutions were collected for analysis. All experiments were replicated and no-ligand controls were run in parallel to verify that equilibrium was reached and that Cu isotope fractionation did not occur without amino acid complexation. Experimental trials with cysteine (each with one control and three replicate experiments) were carried out on two separate occasions to confirm that the results were reproducible.

Table 1 Equilibrium Cu concentrations and δ65Cua values after (27 ± 1 h) dialysis of no-ligand control i.e., donor solution containing 100% free Cu (mean ± 2 SE, n = 5)b.
Full size table

Copper concentrations were determined for acceptor, donor and no-ligand control solutions collected before (t = 0) and after (t = f) Donnan dialysis using inductively coupled plasma optical emission spectrometry (ICP-OES). Aliquots were acidified with 2% HNO3 and analyzed on an iCAP 7400 ICP-OES (ThermoFisher) at a wavelength of 325 nm. Detection limits were assessed daily and ranged from 0.12 to 1.10 µM Cu (3σ, n > 5 blanks). The mean Cu recovery in control experiments was 66.7 ± 1.7%. We attribute this loss of Cu from solution within the dialysis system to adsorption onto the membrane39. Adsorption of free Cu2+ on the membrane did not, however, fractionate Cu isotopes; the isotopic composition of the residual dissolved Cu (δ65Cu =  + 0.90 ± 0.08‰) was within error of its initial Cu isotope value (+ 0.88 ± 0.07‰; Table 1). Copper recovery was greater in experiments with cysteine, glutamate and aspartate; however, only half of Cu was recovered during trials with histidine (Suppl. Table 1), which we attribute to partial adsorption of histidine-Cu complexes on the membrane.

Cu purification and isotope analysis

Sample purification and preparation for isotope analysis was performed in a clean lab under positive pressure inside laminar-flow clean benches, using ultrapure deionized water (Milli-Q, Millipore) and high-purity acids (Aristar Ultra, BDH). Copper samples were separated using AG MP-1 anion resin (100–200 µm mesh, BioRad) in PolyPrep columns (BioRad). Samples were stored and the Cu fraction collected in Teflon beakers, which were cleaned by soaking in aqua regia (3 parts HCl to 1 part HNO3; 2 days) followed by hot 50% HNO3 and 10% hydrofluoric acid (HF; 2 days). Resin was discarded after a single use, and the emptied columns were stored in 20% hydrochloric acid (HCl).

Solution aliquots containing 15 µg Cu were transferred to acid-washed Teflon beakers and dried. Subsequently, 100 µl concentrated hydrogen peroxide (H2O2) was added and samples were refluxed for 15 min at ~ 150 °C to remove organics, then again evaporated to dryness and finally dissolved in 250 µl 6 M HCl. Copper was separated from the sample matrix following Borrok et al.7. Briefly, PolyPrep columns (2 ml stem, 10 ml reservoir) were loaded with 1.5 ml of resin. Resin beds were further rinsed with 10 ml of ultrapure deionized water and conditioned using 3 ml 6 N HCl. Samples were then loaded on the columns and washed twice with 0.5 ml 6 N HCl then once with 3 ml 6 N HCl to ensure complete removal of Sr and sodium. Copper was subsequently eluted in 29 ml 6 N HCl and placed on a hotplate to evaporate samples to complete dryness.

Copper samples were converted to nitrate salts via addition of a few drops of concentrated HNO3, dried, then redissolved in 2% HNO3 with 0.05% HF. Sample dilutions were adjusted to match the concentration (200 µg/L Cu) of the bracketing Cu standard and analyzed on a multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS; ThermoScientific Neptune Plus). A certified Cu wire reference material (HICU-1; National Research Council of Canada) was used for standard bracketing at two sample intervals40. Instrument accuracy was assessed by measuring an 1838 Cu penny standard and the Cu source solution for the experiments (Suppl. Table 2). All values were corrected by the mean difference between HICU-1 and NIST SRM 976, which was measured intermittently throughout every run. Copper isotopic composition is reported relative to the NIST SRM 976 standard:

$$\delta {}^{65}{\text{Cu}} (\textperthousand)=\left[\frac{{\frac{{}^{65}{\text{Cu}}}{{}^{63}{\text{Cu}}}}_{sample}}{{\frac{{}^{65}{\text{Cu}}}{{}^{63}{\text{Cu}}}}_{SRM976}}-1\right]\times 1000$$
(1)

Contamination was assessed by examining Cu recovered from ultrapure water and acceptor solutions (1 ml), which were subjected to column chemistry and resuspended in 1 ml dilute HNO3. Copper concentrations in these solutions averaged 2.4 (n = 3) and 4.7 (n = 4) ng ml−1, respectively. Copper background from reagent impurities, sample handling and the environment did not exceed 0.5% of total Cu in our samples. Recovery and fractionation following column chemistry was assessed using column-cleaned initial (t = 0) Cu donor solutions, each containing a different amino acid and made on a different day. Average Cu recovery for these solutions was 98 ± 5% (n = 7 columns). The mean \(\updelta\)65Cu of these solutions was + 0.88 ± 0.07‰ (n = 5 solutions, which were each individually analyzed ≥ 3 times) following column chemistry; this value is within error of the mean \(\updelta\)65Cu of the Cu stock solution, which was + 0.89 ± 0.02‰ across all instrument runs (n = 20; Suppl. Table 2). From this observation, we conclude that reagent impurities had a negligible effect on our experimental results.

Cu mass balance

The δ65Cu value of complexed Cu was calculated following a mass balance equation derived from Ryan et al.36:

$${\updelta{}^{65}{\text{Cu}}}_{complexed}=\frac{{\updelta{}^{65}{\text{Cu}}}_{total}-({\updelta{}^{65}{\text{Cu}}}_{free}\times {f}_{free})}{(1-{f}_{free})}$$
(2)

where δ65Cufree indicates the Cu isotope abundance measured directly in the acceptor solution at t = f (also referred to here as \({\updelta{}^{65}Cu}_{acceptor}\)), and ffree indicates the relative proportion of free Cu within the system, calculated as:

$${f}_{free}=\frac{{2 \times mass}_{acceptor}}{({mass}_{donor}+{mass}_{acceptor})}$$
(3)

The relative isotope abundance of all dissolved Cu (δ65Cutotal) was calculated as follows:

$${\updelta{}^{65}{\text{Cu}}}_{total}={\updelta{}^{65}{\text{Cu}}}_{acceptor} \times {f}_{acceptor}+{\updelta{}^{65}{\text{Cu}}}_{donor} \times {f}_{donor}$$
(4)

where δ65Cutotal represents the δ65Cu of the total dissolved pool following dialysis. \({\updelta{}^{65}Cu}_{acceptor}\) (equal to \({\updelta{}^{65}{\text{Cu}}}_{free}\) above) and \({\updelta{}^{65}{\text{Cu}}}_{donor}\) indicate the Cu isotope abundance at t = f in the acceptor and donor solutions, respectively, and \(\text{f}\) indicates the fractional contribution of each chamber to the sum Cu mass at t = f:

$${f}_{donor}=\frac{{mass}_{donor}}{({mass}_{donor}+{mass}_{acceptor})}$$
(5)

Finally, Cu isotope separation values between complexed and free Cu were calculated as follows:

$$\Delta {{\text{Cu}}}_{complexed-free}={\delta {}^{65}{\text{Cu}}}_{complexed}-{\delta {}^{65}{\text{Cu}}}_{free}$$
(6)

Results and discussion

Cu binding

Reaction of Cu with amino acids resulted in metal ligation and the retention of ligand-bound Cu in the chamber of the dialyzer which contained amino acids. Copper retention was greatest for cysteine (83.4 ± 04% [n = 3] and 78.5 ± 3.0% [n = 3] of total dissolved Cu2+ in two separate trials), lower for the carboxylate ligands (67.0 ± 0.6% [n = 3] for glutamate, 66.2 ± 0.8% [n = 3] for aspartate), and 57.6 ± 0.8% [n = 2] for histidine (Suppl. Table 1). Free Cu2+ migrated across the dialysis membrane and was recovered in acceptor solutions (Suppl. Fig. 1). Free Cu2+ was distributed equally between the donor and acceptor solutions at mass equilibrium, which was reached in approximately 24 h (Suppl. Fig. 1). Using mass balance equations (1-ffree, Eq. (3); Table 2), we calculated that the proportion of ligand-bound Cu was 57.1 ± 6.0% and 66.9 ± 0.8% for the two cysteine trials, respectively, 34.1 ± 1.1% for glutamate, 32.4 ± 1.6% for aspartate, and 15.3 ± 1.6% for histidine. These values correspond to partition coefficients ([Cu]free/[Cu]bound) of 0.76 ± 0.19 and 0.50 ± 0.02 for the two cysteine trials, respectively, 1.94 ± 0.10 for glutamate, 2.10 ± 0.15 for aspartate, and 5.58 ± 0.70 for histidine (Suppl. Table 4).

Table 2 Cu fractionation induced by amino acids (mean ± 2 SE).
Full size table

Cu isotope fractionation

Copper complexation by cysteine, glutamate, and aspartate resulted in Cu isotope fractionation (Table 2; Fig. 1). For cysteine, mean (± 2SE) δ65Cufree (measured in acceptor solutions at t = f) exceeded mean δ65Cucomplexed (calculated as in Eq. 2) across all experiments (n = 6), with values of + 1.38 ± 0.08‰ and + 0.89 ± 0.14‰, respectively. For the carboxylates, the opposite trend was observed: δ65Cufree was + 0.81 ± 0.04‰ (n = 3) in glutamate experiments and + 0.89 ± 0.03‰ (n = 3) in aspartate experiments while corresponding δ65Cucomplexed values were + 1.07 ± 0.06‰ (n = 3) and + 1.05 ± 0.07‰ (n = 3). In histidine experiments, δ65Cufree and δ65Cucomplexed were within error, with values of + 0.79 ± 0.08‰ (n = 2) and + 0.75 ± 0.05‰ (n = 2), respectively.

Figure 1
figure 1

(A) Cu isotope fractionation (\(\Delta\)65Cu = \(\delta\)65Cubound − \(\delta\)65Cufree) due to ligation by cysteine (Cys), histidine (His), glutamate (Glu) and aspartate (Asp). Gray points represent \(\Delta\)65Cu calculated for independent experiments; black points represent the mean \(\Delta\)65Cu of replicates (± 2SE). (B–E) Molecular structures of the amino acids investigated in this study.

Full size image

These experimental results indicate that thiol-bearing cysteine preferentially ligated 63Cu (Δ65Cucomplex-free = − 0.48 ± 0.18‰), while glutamate and aspartate, with their carboxyl (O) side chains, favored 65Cu (Δ65Cucomplex-free =  + 0.26 ± 0.04 and + 0.16 ± 0.10‰, respectively; Table 2; Fig. 1). Cooper complexation by N-bearing histidine, in contrast, resulted in negligible isotopic fractionation (Δ65Cucomplex-free = − 0.01 ± 0.04‰). Copper concentrations and δ65Cu for all samples are available in Suppl. Table 3 and averaged results are available in Suppl. Table 4. We note that the large error associated with cysteine ligation was not driven by variability among replicates in the proportion of Cu bound (see above; Table 2), but by variability in the Cu isotope abundances measured in donor and acceptor solutions across replicate experiments (see Suppl. Table 3).

Comparison with theoretical calculations

Our experimentally-derived Cu isotope fractionation effects are consistent in directionality with mass-dependent equilibrium isotope fractionation. Mass-dependent isotope fractionation occurs because chemical bonds differ in their potential energy as a function of atomic position (i.e., in their potential energy surface). When small changes in atomic position lead to significant changes in potential energy, the bond may be said to have a “steeper” potential energy surface. Heavier isotopes concentrate in the “steeper” (more stable) chemical phase to minimize the overall energy of the system41. Bonds with higher natural frequencies of fundamental vibration are associated with steeper potential energy surfaces (i.e., deeper potential energy wells)41 and thus have higher spring constants, meaning that they are more resistant to deformation5,42. These “stiff” bonds are short and strong—properties associated with lighter ligands, lower coordination numbers, and higher oxidation states41. In proteins, stiff bonds are associated with ‘hard’ Lewis bases i.e., those whose electron clouds are less easily distorted under an electric field43,44. Carboxyl groups (RO−) and water (H2O) are hard bases; imidazole is classified as ‘borderline’; thiols (RSH) are ‘soft’ or easily polarized bases45. Therefore, when Cu is in the 2+ oxidation state (and a borderline Lewis acid), the carboxylates (glutamate and aspartate) should evince greater preference for 65Cu than histidine, while cysteine should accumulate 63Cu. This pattern is clearly manifest in our experimental results (Table 2; Fig. 1).

Though consistent in directionality, experimental evidence indicates that the magnitude of Cu isotope fractionation by cysteine is less than that predicted ab initio from DFT4. Theoretical calculations from Fujii et al.4 predict that, relative to H2O, Cu complexation by cysteine should fractionate the Cu isotope pool by − 1.18‰ (Δ65Cucomplex-free) at 25 °C. Our experimentally determined Δ65Cucomplex-free of − 0.48 ± 0.18‰ is in keeping with the directionally of this value. Both theory and experiment indicate cysteine favors 63Cu; however, our measured value is less than half the predicted DFT value. This suggests that ab initio calculations may overestimate isotope fractionation due to ligation by cysteine. This discrepancy between theory and experiment can be explained by differences in the coordination geometry assumed in the DFT model compared to the Cu-amino acid complexation that occurs in real solutions. In the DFT calculations, Fujii et al.4 assumed five-fold coordination of Cu2+ bonded to H2O molecules and a single cysteine ligand. Conversely, extended X-ray absorption fine structure (EXAFS) spectroscopic measurements of Cu2+-cysteine complexes in aqueous solutions show Cu2+ coordination with three sulfur atoms in the first shell at approximately 2.28 Å37. Stiffer bonds and differences in coordination numbers in the Cu2+-cysteine that actually form in experimental systems would be expected to mute the predicted Cu isotope fractionation effect.

We have ruled out redox chemistry as a major cause of Cu isotope fractionation by cysteine complexation in our experiments. Previous XANES studies37 have shown that the oxidation state of Cu remains as Cu2+ upon binding with cysteine. Cu2+ reaction with cysteine does not appear to produce Cu1+-cysteine complexes, which can be distinguished from EXAFS spectra37. Because previous XANES and EXAFS analyses indicate that Cu2+ is not chemically reduced to Cu1+ by cysteine, we attribute our measured Cu isotope fractionation to complexation to the thiol ligand rather than a redox-driven process.

In our experiments, we observed that both glutamate and aspartate favored 65Cu with Δ65Cucomplex-free values of + 0.26 ± 0.04 and + 0.16 ± 0.10‰ (Table 2; Fig. 1), respectively which differ from the DFT derived value for glutamate of − 0.11‰ relative to H2O4. These results suggests that DFT models underestimated fractionation due to ligation by carboxylates. As with cysteine, this difference is likely attributable to variability in the coordination geometry and potentially speciation of experimental complexes relative to the simplified models used in DFT calculations. We note that our experimentally determined values are broadly consistent with observation of other organic ligands with O-binding sites. Using a similar dialysis approach, Ryan et al.36 reported Δ65Cucomplex-free values for O-rich ligands ranging from + 0.14 ± 0.11‰ (Suwanee River fulvic acid) to + 0.84 ± 0.30‰ (desferrioxamine B—a siderophore).

For histidine, our experimental results (Table 2; Fig. 1) agreed with the ab initio calculations4. These independent approaches both indicate that negligible isotopic fractionation occurs when Cu is bound to histidine under ambient conditions. Ab initio calculations4 indicate that Δ65Cucomplex-free for the aqueous Cu-histidine complexes Cu(His)(H2O)42+ and Cu(His)(H2O)32+ should yield Δ65Cucomplex-free values of − 0.05‰ and − 0.08‰, respectively. Our measured value of − 0.01 ± 0.04‰ is close to these values.

Comparison with in vivo measurements

Though our measured magnitude of Cu isotope fractionation for cysteine was smaller than predicted by DFT, our data are in excellent agreement with Cu isotope ratios measured in proteins (Table 3). The Cu isotope composition of two structurally-distinct proteins, metallothionein and Cu,zinc (Zn) superoxide dismutase, have previously been measured in vivo17,34. Metallothionein binds clusters of metal ions—as many as seven bivalent or 20 monovalent ions—using predominantly the S residues of cysteines46. Cu,Zn superoxide dismutase, in contrast, binds only two Cu atoms in identical sites using N from histidine in a tetrahedral configuration47. In human cortical tissue, metallothionein Cu is isotopically light (δ65Cu = − 0.20 ± 0.21‰) while Cu in Cu,Zn superoxide dismutase is isotopically heavy (δ65Cu =  + 0.41 ± 0.27‰)17. This pattern also holds true for the yeast Saccharomyces, with δ65Cu values for metallothionein and Cu,Zn superoxide dismutase of − 2.13‰ and − 1.60‰, respectively34. With a difference of − 0.61 ± 0.34‰ (n = 5 subjects) in humans17 and − 0.53‰ in yeast34, the consistency in the relative partitioning of Cu isotopes between these proteins across biological systems is striking (note: The absolute isotopic value of the two protein pools differs across organisms and individuals due to variability in the dietary source of Cu48, among other factors, and are thus not comparable.) Moreover, the difference between Δ65Cucomplex-free of cysteine and histidine in our experiments, − 0.47 ± 0.18‰, is within error of these values. The coherence between our in vitro Cu isotope separation value and in vivo protein data suggest that the identity of ligating residues is a major factor dictating metal fractionation in proteins. However, as this study focused exclusively on Cu2+, we cannot exclude the potential contribution of oxidation state in accounting for differences among proteins.

Table 3 Cu isotope compositions of sulfur (S) and nitrogen (N) bonding environments across levels of biological complexity.
Full size table

Isotope partitioning among proteins with S-rich and N-rich binding sites for Cu also manifests at the cellular level consistent with our experimental results: Cells and fluids enriched in 63Cu are associated with Cu proteins bearing S-rich binding sites. In healthy humans, for example, blood serum is enriched in 63Cu (δ65Cu = − 0.26 ± 0.40‰, n = 49) relative to erythrocytes (red blood cells; δ65Cu =  + 0.56 ± 0.50‰, n = 49)8. Calculations based on DFT support the hypothesis that this fractionation occurs because Cu is ferried to ceruloplasmin—the major Cu protein in blood—via an antioxidant protein (ATOX) and a chaperone protein (ATP7B). The isotope effect is expressed when ATOX and ATP7B bind Cu using two cysteine thiols in a linear configuration49. Because the transfer of Cu from ATOX/ATP7B to ceruloplasmin is quantitative, the isotopic composition of Cu in the intermediaries is transferred to ceruloplasmin. Thus, 63Cu tends to accumulate in serum while red blood cells accumulate 65Cu8,49. Albarede et al.8 also noted that Cu,Zn superoxide dismutase is more prevalent in red blood cells. The difference in δ65Cu between serum and red blood cells of − 0.82 ± 0.49‰ (Table 3), is larger but still within error of the difference we observed between Δ65Cucomplex-free values for cysteine and histidine (− 0.47 ± 0.18‰; Table 3).

Medical and environmental applications

As structurally-distinct Cu-binding sites fractionate Cu isotopes among biomolecules (Table 3), patterns in the distribution of Cu isotopes among cells, groups of cells, and their growth environment represent potential biomarkers of cellular activity. For example, like many Cu proteins, both metallothionein (S ligands) and Cu,Zn superoxide dismutase (N ligands) play important roles in mediating oxidative stress2: Cu,Zn superoxide dismutase catalyzes the disproportionation of superoxide radicals produced during oxygen metabolism47 while metallothionein is a chaperone protein involved in heavy metal detoxification50. In humans, expression of both proteins changes during carcinogenesis due to their respective roles in managing reactive oxygen species and Cu concentrations, which can rise during growth of cancerous tissue51. Consequently, an emerging body of research suggests a link between Cu isotopes and cancer9,11,27,28,29,52, and other diseases20,21,53.

Our findings support the hypothesis that changes in Cu isotope partitioning during cancer occur, in part, due to the expression of distinct Cu-binding biomolecules9,27. Serum δ65Cu is significantly lower in patients of colorectal and breast cancer27, as well as liver cancer (hepatocellular carcinoma)9,26 and cirrhosis31 relative to healthy individuals (− 0.26 ± 0.4‰, n = 49)8, with a proposed cutoff value of < − 0.35‰ indicating increased risk of mortality27. Several mechanisms have been hypothesized to explain these anomalies. Balter et al.9 ascribed low δ65Cu in the blood of hepatocellular carcinoma patients, which they related to increased concentrations, to release of metallothionein (S ligand)-bound Cu in endogenous stores. Conversely, Télouk et al.27 argued that shifts in serum δ65Cu of cancer patients occurs because lactate, which binds Cu2+ with hydroxyl (O) groups in cancer cells, accumulates 65Cu. Our experimental findings support the assumption underlying both hypotheses—that S ligands favor 63Cu and O ligands favor 65Cu. However, given the complexity of Cu trafficking and the likely importance of considerations apart from ligand identity (e.g., oxidation state, coordination geometry/ligand arrangement, metal abundance) additional work probing the relative importance of these variables is necessary to distinguish between them.

While recent work has focused on the medical potential of isotope metallomics, the work presented here is also relevant for using metal isotopes to investigate microbial metabolism in the environment. Though Cu has received less attention, several researchers have proposed biomarkers for application in the geologic record based on other transition metals, including Zn to track eukaryotes54, nickel (Ni) to trace methanogenesis55, and molybdenum (Mo) to investigate nitrogen assimilation pathways56,57. Metal isotopes are also increasingly being used to trace the sources, sinks, and cycling of metals in the ocean58,59,60. Unlike Zn61,62,63 and Ni64, marine primary producers do not appear to significantly fractionate the dissolved Cu pool58. Nevertheless, several studies have shown that prokaryotes can significantly fractionate Cu, with Δ65Cucells-media ranging from − 4.4‰22 to + 3.0‰18. Understanding why and when these biological fractionation effects occur in the environment will require further study, including consideration of when quantitative uptake may mute isotope fractionation. However, our work implicates changes in the Cu proteome of microbial communities as a potential factor driving differences in Cu isotope ratios.

Conclusion

Here, we present the first experimental results characterizing the magnitude of Cu isotope fractionation during ligation by amino acids most commonly observed at protein metal-binding sites. Our work demonstrates that amino acids differentially fractionate Cu isotopes, with light 63Cu accumulating at the S sites of cysteine and heavy 65Cu accumulating at O sites of glutamate and aspartate. Our findings are in keeping with the few studies that have successfully measured Cu isotope ratios in in vivo proteins17,34, but differ in magnitude from computational estimates (Fig. 1)4, likely due to differences in coordination environment under modeled and laboratory conditions. This work supports the hypothesis that metalloprotein biosynthesis affects the distribution of transition metal isotopes within cells and suggests that the magnitude of computed isotope separation values should be interpreted with caution.