HyPer as a tool to determine the reductive activity in cellular compartments

Abstract

A multitude of cellular metabolic and regulatory processes rely on controlled thiol reduction and oxidation mechanisms. Due to our aerobic environment, research preferentially focuses on oxidation processes, leading to limited tools tailored for investigating cellular reduction. Here, we advocate for repurposing HyPer1, initially designed as a fluorescent probe for H₂O₂ levels, as a tool to measure the reductive power in various cellular compartments. The response of HyPer1 depends on kinetics between thiol oxidation and reduction in its OxyR sensing domain. Here, we focused on the reduction half-reaction of HyPer1. We showed that HyPer1 primarily relies on Trx/TrxR-mediated reduction in the cytosol and nucleus, characterized by a second order rate constant of 5.8 × 10² M⁻¹s⁻¹. On the other hand, within the mitochondria, HyPer1 is predominantly reduced by glutathione (GSH). The GSH-mediated reduction rate constant is 1.8 M⁻¹s⁻¹. Using human leukemia K-562 cells after a brief oxidative exposure, we quantified the compartmentalized Trx/TrxR and GSH-dependent reductive activity using HyPer1. Notably, the recovery period for mitochondrial HyPer1 was twice as long compared to cytosolic and nuclear HyPer1. After exploring various human cells, we revealed a potent cytosolic Trx/TrxR pathway, particularly pronounced in cancer cell lines such as K-562 and HeLa. In conclusion, our study demonstrates that HyPer1 can be harnessed as a robust tool for assessing compartmentalized reduction activity in cells following oxidative stress.

Graphical abstract

Highlights

• •AlphaFold2 predicts thiol-based interaction of HyPer with human Trx1.
• •In vitro, Trx1 and GSH reduce HyPer, whereas Trx2, Grx1 and Grx2 cannot facilitate the reduction.
• •Trx1 exhibits a 100-fold higher efficiency in reducing HyPer compared to GSH.
• •In cellulo, cyto-HyPer is reduced by Trx1, while mito-HyPer is reduced by GSH.
• •HyPer can be exploited to characterize thiol-reducing systems in living cells.

1. Introduction

The formation of disulfide bonds between cysteine residues is a post-translational modification that can provide structural support and confer conformational stability to proteins. In addition, reversible formation of disulfides can impact the activity of metabolic and signaling proteins, thus performing regulatory functions [[1], [2], [3]]. Disulfides are formed when the thiol groups of cysteine residues are oxidized in reactions mediated by hydrogen peroxide (H₂O₂) or in thiol-disulfide exchange reactions. The reduction of disulfide bonds is catalyzed by the thiol-disulfide exchange enzymes of the thioredoxin superfamily, such as thioredoxins or glutaredoxins [4].

Thioredoxin (Trx) and glutaredoxin (Grx) are ubiquitous oxidoreductases present in both prokaryotic and eukaryotic cells. Grx transfers oxidative equivalents from thiol groups of proteins to glutathione (GSH), a cysteine-containing tripeptide that is then reduced through the NADPH-dependent activity of glutathione reductase [5]. Additionally, Grx can act as an electron donor to deglutathionylate proteins. The Trx system is an alternative pathway for thiol reduction and consists of three components: Trx, which reduces oxidized proteins, Trx reductase (TrxR), which catalyzes Trx reduction, and NADPH, which serves as the source of electrons [6]. Trx1/TrxR1 and Grx1/GSH pathways are active in the cytosol and nucleus, while Trx2/TrxR2 and Grx2/GSH – inside the mitochondria.

The role of Trx/TrxR- and Grx/GSH-dependent enzymatic systems in cellular metabolic and signaling pathways is a focus of intense research, but the bottleneck is the limitations of the available analytical methods. The traditional approach to estimate the cellular disulfide-reducing activity involves cell lysis and experiments in vitro on total cell protein extracts [7,8]. Activity of Trx/TrxR or Grx/GSH pathways is determined from the analysis of the reduction kinetics of various synthetic or natural Trx and Grx targets. In addition, recently, bioimaging methods based on the use of synthetic dyes or genetically encoded biosensors have been elaborated to characterize disulfide-reducing pathways in living cells. By employing fluorescent dyes such as Thio [9,10] and RX1 [11], it became feasible to assess the levels of proteins engaged in these pathways. Additionally, through the utilization of genetically encoded biosensors based on Trx [12] or Grx [13], the oxidation status of redox-active cysteines within these oxidoreductases can be tracked across diverse biological contexts. Nevertheless, quantifying the disulfide-reducing activity dependent on Trx/TrxR or Grx/GSH within living cells remains a complex task. In this study, we have developed a protocol to address this challenge in human cells by utilizing HyPer (namely the HyPer1 variant), a genetically encoded biosensor originally designed for detecting H₂O₂.

HyPer1 is a chimeric protein that was constructed by inserting a circularly permuted yellow fluorescent protein (cpYFP) into the regulatory domain of Ec_OxyR, a bacterial transcription factor which is responsible for H₂O₂ sensing in E. coli [14]. Ec_OxyR is activated by H₂O₂ through the formation of an intramolecular disulfide bond, which is reversed by reduction by bacterial glutaredoxin Grx1 [15]. When HyPer1 is expressed in human cells, the oxidation of the Ec_OxyR domain by H₂O₂ causes its conformational change, which results in the alterations of the fluorescence properties of the cpYFP moiety [16]. Hence, monitoring HyPer1 fluorescence in various compartments of living human cells is widely used for following intracellular H₂O₂ levels in function of time [[17], [18], [19]]. In addition, since the redox-active thiol groups of HyPer1 are targets not only for H₂O₂, but also for cellular oxidoreductases, this biosensor has recently proven to be a tool for monitoring the disulfide reducing activity in cells [20]. In the current study, we aimed to elucidate which thiol-reducing system (Trx/TrxR- or Grx/GSH-dependent) is responsible for HyPer1 reduction in the cytosol, nucleus, and mitochondria of human cells and developed a protocol to quantify the activity of these systems in living cells after the induction of transient oxidative stress.

2. Materials and methods

2.1. Cell cultures

K-562 cells. Human K-562 myelogenous leukemia cell lines stably expressing mitochondrial (mito), nuclear (nuc) and cytosolic (cyto) versions of HyPer1 and SypHer were obtained as described in our previous study [21]. Briefly, K-562 cells were transduced with the respective lentiviruses (produced by Evrogen Ltd, Moscow) using 5 μg/ml polybrene 20 h after cell seeding. To obtain cell lines with the different biosensor expression levels, K-562 cells were transduced at different multiplicities of infection (MOI): MOI = 5, 10 (cyto-HyPer1, SypHer) and MOI = 1,2 or 5 (nuc-HyPer1, mito-HyPer1). Most experiments were performed on cell lines with MOI = 5 (cyto-HyPer1) and MOI = 1 (nuc- and mito-HyPer1), unless indicated otherwise. All K-562 cell lines were cultured in RPMI medium (Biolot Ltd) supplemented with 10 % fetal bovine serum (FBS) (HyClone), 1 % l-glutamine, and 1 % penicillin-streptomycin (Gibco). Cells were maintained at 37 °C, 5 % CO₂, and were routinely passaged twice a week at a split ratio of 1:5. For the experiments, cells were seeded in 3 cm dishes at a density of 200,000 cells/mL two days before the analysis.

HeLa cells. A HeLa cell line stably expressing cyto-HyPer1 was obtained in our previous study [22], using the transduction protocol described above at MOI = 5. Cells were cultivated in DMEM/F12 medium (Gibco) supplemented with 10 % FBS (HyClone), 1 % l-glutamine, and 1 % penicillin-streptomycin (Gibco, USA) at 37 °C and 5 % CO₂. Cells were routinely subcultured twice a week at a split ratio of 1:5. For the experiments, 100,000 cells were seeded per 3 cm dish 2 days before the analysis. Immediately before the measurements, cells were harvested with 0.05 % trypsin-EDTA.

Mesenchymal stem/stromal cells (MSCs). Human MSCs (derived from human desquamated endometrium, [23]) stably expressing cyto-HyPer1 were obtained in our previous study [21], using the transduction protocol described above at MOI = 5. MSCs were cultivated in DMEM/F12 medium (Gibco) supplemented with 10 % FBS (HyClone), 1 % l-glutamine, and 1 % penicillin-streptomycin (Gibco). Cells were maintained at 37 °C, 5 % CO_2, and were routinely subcultured twice a week at a split ratio of 1:3 for up to 15 passages. For the experiments, 50,000 cells were seeded per 3 cm dish 2 days before the analysis. Immediately before the measurements, cells were harvested with 0.05 % trypsin-EDTA.

Induced pluripotent stem cells (iPS). Human iPS constitutively expressing cyto-HyPer1 were derived by reprogramming of HyPer-expressing MSCs in our previous study [22]. iPS possessed clonal growth morphology, a diploid karyotype, expressed pluripotency markers Oct-3/4, Nanog, SSEA-4, Sox-2, alkaline phosphatase, and were able to differentiate into cell types of three germ layers (see Ref. [22] for the details). Cells were cultivated at 37⁰ С and normoxia (21 % О₂) with 5 % CO₂ content on Matrigel-coated (Corning) dishes in mTeSR medium (STEMCELL Technologies) with daily medium change and were routinely subcultured mechanically once a week for up to 30 passages. For the experiments, iPS clones were seeded in 24-well plates 3 days before the analysis. Immediately before the measurements, cells were harvested with 0.05 % trypsin-EDTA.

Dermal fibroblasts (Fibr). Human fibroblasts (Russian Cell Culture Collection, St.Petersburg, Russia) were cultivated in DMEM/F12 medium (Gibco) with 10 % FBS (HyClone), 1 % l-glutamine, and 1 % penicillin-streptomycin (Gibco). Cells were maintained at 37 °C, 5 % CO_2, and were routinely subcultured twice a week at a split ratio of 1:3. At the 4th passage, Fibr were transduced with the lentiviral vector encoding cyto-HyPer1 (Evrogen Ltd, Moscow) using the protocol described above at MOI = 5, and were cultivated for up to 10 passages. For the experiments, 50,000 cells were seeded per 3 cm dish 2 days before the analysis. Immediately before the measurements, cells were harvested with 0.05 % trypsin-EDTA.

2.2. Protein purification

HyPer1. XL1-Blue E. coli cells transformed with the pQE30-HyPer1 plasmid were plated on LB agar supplemented with 100 μg/ml ampicillin and grown at 37 °C for 16 h. Plates were then incubated at 25 °C for an additional 72 h, until the appearance of yellow colonies. LB medium with 100 μg/ml ampicillin was subsequently inoculated with a single yellow colony and grown at 25 °C for 36 h with shaking at 130 rpm. Cells were harvested and the pellet resuspended in 50 mM Tris, pH 8.0, 300 mM NaCl, 1.5 mM DTT, 1 mg/mL leupeptin, 0.1 mg/mL 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AESBF), 100 mg/mL DNaseI, 20 mM MgCl₂) and incubated at 4 °C for 30 min while rotating. Cells were then lysed by sonication at 70 % amplitude for 5 min with a 20 s on – 40 s off cycle (Sonics Vibra-cell). The lysate was clarified by centrifugation at 39,000×g for 30 min at 4 °C, filtered through a 0.45 mm filter and in-batch incubated with Ni²⁺- Sepharose (GE Healthcare) affinity beads equilibrated with binding buffer (50 mM Tris, pH 8.0, 300 mM NaCl, 1.5 mM DTT) for 1 h at 4 °C while rotating and subsequently packed into a Tricorn 5 column (Cytiva). The AKTA™ Pure system (Cytiva) controlled by the UNICORN 6.3.0.731 software was used for the next steps. After a wash step with binding buffer, HyPer1 was eluted with a linear gradient to 1 M imidazole over 10 column volumes. Imidazole was removed by dialysis in 50 mM Tris/HCl, pH 8.0, 300 mM NaCl, 1.5 mM DTT at 4 °C overnight in a Slide-A-Lyser® Dialysis Cassette G2 with a 10,000 MWCO (ThermoFisher). The fractions containing HyPer1 were concentrated using a Vivaspin 6 10,000 MWCO PES concentrator (Sartorius) and further purified by size exclusion chromatography using a HiLoad Superdex75 PG (16/60) column (GE Healthcare) equilibrated with 50 mM Tris, pH 8.0, 300 mM NaCl, 2 mM DTT. The purity of the protein was assessed by non-reducing SDS-PAGE and the pure fractions were pooled, concentrated, snap-frozen in liquid N₂, and stored at −80 °C.

Human Trx1, Trx2, Grx1 and Grx2. Grx1 and Grx2 (isoform 1) were subcloned into the pD441-SUMO-His vector by Gibson Assembly. Trx1 and Trx2 were already received in this vector from the lab of Elias Arnér. All constructs were transformed into the NEBTurbo (New England Biolabs) expression strain. An overnight preculture of NEBTurbo cells harbouring the plasmids in LB supplemented with 25 μg/ml kanamycin was diluted 1:200 in 1 l LB supplemented with 25 μg/ml kanamycin and grown at 37 °C with shaking at 130 rpm until OD₆₀₀ = 0.6. Protein expression was then induced with 1 mM IPTG and grown for a further 16 h at the same temperature. Cells were harvested and the pellet resuspended in 100 mM Tris, pH 7.5, 250 mM NaCl, 1 mM DTT, 1 mg/mL leupeptin, 0.1 mg/mL 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AESBF), 100 mg/mL DNaseI, 20 mM MgCl₂) and incubated at 4 °C for 30 min while rotating. Cells were then lysed by sonication at 70 % amplitude for 5 min with a 20 s on – 40 s off cycle (Sonics Vibra-cell). The lysate was clarified by centrifugation at 39,000×g for 30 min at 4 °C, filtered through a 0.45 mm filter and in-batch incubated with Ni²⁺- Sepharose (GE Healthcare) affinity beads equilibrated with binding buffer (100 mM Tris, pH 7.5, 250 mM NaCl, 25 mM imidazole, 1 mM DTT) for 1 h at 4 °C while rotating and subsequently packed into a Tricorn 5 column (Cytiva). The ÄKTA™ Pure system (Cytiva) controlled by the UNICORN 6.3.0.731 software was used for the next steps. After a wash step with binding buffer, the protein was eluted with a linear gradient to 1 M imidazole over 10 column volumes. Collected fractions were analyzed by SDS-PAGE and those containing Trx were pooled in a SnakeSkin™ dialysis tubing with 10K MWCO (ThermoScientific) and dialyzed into 10 mM Tris/HCl, pH 7.5, 75 mM NaCl, 1 mM EDTA overnight at 4 °C. His-tagged Ulp1 at a 1:200 w/w ratio and 25 mM DTT were then added and the mixture incubated for 1 h at 30 °C. The Trx-Ulp1 mixture was then buffer-exchanged into binding buffer and in-batch incubated with Ni²⁺-Sepharose (GE Healthcare) affinity beads equilibrated with binding buffer (100 mM Tris/HCl, pH 7.5, 250 mM NaCl, 25 mM imidazole, 1 mM DTT) for 1 h at 4 °C with rotating. The beads were subsequently packed into a Tricorn 5 column (Cytiva), and the flow-through collected. His-tagged Ulp1, SUMO-His and uncleaved SUMO-His-protein were eluted with a linear gradient to 1 M imidazole in binding buffer using the ÄKTA™ Pure system. The flow-through and collected fractions were analyzed on an SDS-PAGE and those containing the cleaved target proteins were pooled and concentrated using a Vivaspin 6 3000 MWCO PES concentrator (Sartorius) and further purified by size exclusion chromatography using a HiLoad Superdex75 PG (16/60) column (GE Healthcare) equilibrated with 100 mM Tris/HCl, pH 7.5, 500 mM NaCl, 2 mM EDTA. The fractions were again analyzed by SDS-PAGE and those containing pure protein were pooled and concentrated. 10 % glycerol was added and the proteins were snap-frozen in liquid N₂ and stored at −80 °C.

2.3. Cell treatments

In inhibition experiments, before harvesting, cells were incubated at standard growth conditions for 3 h with 20–1000 nM auranofin, 100–500 μM mitomycin C, and 200–1000 nM Tri-1. 500–1000 μM BSO (buthionine sulfoximine) or 10–25 μM BCNU (1,3-Bis(2-chloroethyl)-1-nitrosourea) were added to the cell medium 24 h or 1 h prior to analysis, respectively. Dual inhibition was performed by adding auranofin (100 nM, 3 h) to cells preincubated with BSO (500 μM, 3 h). When indicated, cells were also incubated with 100–1000 nM Na₂SeO₃ for 24 h, 100 μM EUK-134 for 3 h, and 10 μM menadione for 20 min.

2.4. Monitoring of HyPer1 reduction kinetics in cells

Cell sample preparation. HyPer-expressing cells were suspended in fresh DMEM/F12 medium (supplemented with FBS, l-glutamine and 10 mM HEPES) at a density of 25,000 cells/mL. Cell samples were aliquoted (600 μL per flow cytometry tube) and incubated at standard growth conditions for at least 20 min for adaptation.

Oxidative pulse. Just before the analysis of HyPer1 reduction, suspended cells were exposed to 100 μM exogenously added H₂O₂ for 5 min, after which catalase (1000 U) was added.

Flow cytometry analysis. Kinetics of cyto-HyPer1 reduction were monitored with a CytoFLEX flow cytometer (Beckman Coulter) for 10 min [22]. During the analysis, a 37 °C mixture of air and CO₂ (5 %) was supplied into the sample tube to stabilize physiological pH and temperature. Variations in the HyPer1 signal arising from the alkalization of the cell medium, which usually occurs during long-term room-air measurements, were prevented by constant CO₂ pumping. Fluorescence was simultaneously monitored at λ_ex = 488 nm/λ_em = 525 nm (corresponds to the oxidized form of HyPer1, hereafter denoted as $I_{488}$ signal) and λ_ex = 405 nm/λ_em = 525 nm (corresponds to the reduced form of HyPer1, hereafter denoted as $I_{405}$ signal) in gated HyPer1-positive cells. By varying the voltage or gain settings of the cytometer detectors, the changes in the $I_{488}$ and $I_{405}$ fluorescence signals upon HyPer1 oxidation were aligned so that $\Delta I_{488}$ was always equal to ${-\Delta I}_{405}$. Under such conditions, the sum of signals from the oxidized and reduced forms of HyPer1 $\left(I_{488}+I_{405}\right)$ does not change upon HyPer1 oxidation/reduction and depends only on the overall amount of biosensor in cells. Adjusted voltage/gain settings turned out to be the same for all cell lines used (for more details see Ref. [22]).

Measurements of fluorescence signals from completely oxidized/reduced HyPer. To calibrate the HyPer1 signal, in each experiment HyPer1 fluorescence was measured in cell samples treated with 30 mM dithiothreitol (DTT) for 10 min and 0.5 mM H₂O₂ for 5 min to obtain fully reduced and fully oxidized samples, respectively. Cell treatments were performed in standard growth conditions (37 °C and 5 % of CO₂) in flow cytometry tubes.

2.5. Kinetic data processing

To decrease the noise-to-signal ratio, the signal from the oxidized HyPer1 ($I_{488}$) was normalized to the total signal from HyPer1 ($I_{488}+I_{405}$) using the Kaluza Analysis Software (Beckman Coulter):

$$I_{488norm}=\frac{I_{488}}{I_{488}+I_{405}}$$ (1)

After that all data were exported to ASCII format and processed further using Microsoft Excel (Microsoft). The fluorescence signals were calibrated so as to convert the normalized signal $I_{488norm}$ to the proportion of oxidized HyPer1 in each cell $\left(OxD\right)$, using the following equation:

$$OxD=\frac{I_{488norm}-I_{488normDTT}}{I_{488normH2O2}-I_{488normDTT}}$$ (2)

where $I_{488normH2O2}$ and $I_{488normDTT}$ are the mean $I_{488norm}$ signals measured in cells with fully oxidized and reduced HyPer1 (cells treated with high doses of H₂O₂ and DTT, respectively). It should be noted that the value of $OxD$ here is identical to the $OxD$ parameter, introduced previously in studies on roGFP-based biosensors [24]. Calculation of $OxD$ using the formula suggested in Ref. [24] enables to avoid adjustment of the detector gains described in Section 2.4 and replaces Eq. (1) and Eq. (2) with a single equation:

$$OxD=\frac{I_{488}\times I_{405DTT}-I_{488DTT}\times I_{405}}{I_{488}\times I_{405DTT}-I_{488}\times I_{405H2O2}+I_{488H2O2}\times I_{405}-I_{488DTT}\times I_{405}}$$ (3)

Here, Eq. (3) is the same as in Ref. [24], with only one trait – indexes 488 and 405 are swapped, having in mind that the excitation maximum at 488 nm corresponds to the oxidized HyPer1 and at 405 nm – to its reduced form (in contrast to roGFP). In fact, the procedure of $OxD$ calculation using Eq. (3) obviates the adjustment of the detector gains by introducing the instrumental parameter $\left(I_{488H2O2}-I_{488DTT}\right)/(I_{405DTT}-I_{405H2O2})$ and multiplying $I_{405}$ by this parameter in Eq. (1). Despite the clear advantages of using Eq. (3), for the purposes of the current research, we used the protocol for $OxD$ derivation based on the detector gain adjustment and Eqs. (1), (2)), because in this approach the sum fluorescence signal $I_{488}+I_{405}$ does not change upon HyPer1 oxidation and reflects only the intracellular biosensor content. Gating of cells based on the sum signal was used for monitoring HyPer1 reduction in the cell populations with different HyPer1 content (see Section 2.11) and was important at the stage of the experimental protocol elaboration and validation.

2.6. Calculation of the rate constant of HyPer1 reduction

We fitted the $OxD$ vs time curves to the kinetic equations derived by Brito and Antunes [25], which we adapted to describe the reduction of HyPer1. According to the model [25], upon the removal of the oxidative load, the fraction of oxidized HyPer ($OxD$) begins to decrease and eventually reaches its background steady-state level (${OxD}_{ss}$). In the case of complete HyPer1 oxidation at the initial moment of time, the kinetics of HyPer1 reduction is described as follows (see Eq. (4) in Ref. [25]):

$$OxD\left(t\right)={OxD}_{ss}+{\mathrm{e}}^{-\left(k_{ox}+k_{rd}\right)\times t}\times \left(1-{OxD}_{ss}\right)$$ (4)

$${OxD}_{ss}=\frac{k_{ox}}{k_{rd}+k_{ox}}$$ (5)

In these equations, $k_{ox}$ and $k_{rd}$ are the pseudo first-order rate constants for the oxidation and reduction of HyPer1, respectively. $k_{rd}$ depends on the activity of the enzymatic system that controls the reduction of the redox-active thiols of HyPer1. $k_{ox}$ depends on the background intracellular level of Н₂О₂.

Our previously published results showed that under normal physiological conditions, ${OxD}_{ss}$ in the cytoplasm of different human cell cultures is practically equal to zero and $k_{ox}\ll k_{rd}$ [22]. In this case, to estimate the $k_{rd}$ upon the oxidative load removal, a simple exponential dependence (Eq. 10b, [25]) can be used:

$$OxD\left(t\right)={\mathrm{e}}^{-k_{rd}t}$$ (6)

In summary, the $k_{rd}$, the rate constant of HyPer1 reduction occurring after the removal of the oxidative load on cells, can be measured by fitting the dependence $OxD\left(t\right)$ with Eq. (6) in the absence of background HyPer1 oxidation (i.e. in normal physiological conditions) and with Eq. (3) in its presence – when the condition $k_{ox}\ll k_{rd}$ is not valid due to the disturbances of the cell redox homeostasis (e.g. in the case of decreased thiol reduction capacity and/or enhanced intracellular H₂O₂ generation).

2.7. Determination of the second-order rate constant of HyPer1 reduction by Trx1, Trx2, Grx1 Grx2 and GSH in vitro

HyPer1 was oxidized with a 10x molar excess of diamide for 10 min at RT and buffer-exchanged into 100 mM NaPO_i, pH 7.4, 150 mM NaCl, 0.1 mM diethylene triamine penta-acetic acid (DTPA) using a 5 mL HiTrap desalting column (Cytiva). Trx1, Trx2, Grx1 and Grx2 (hereafter referred to as “putative reducing protein”) were reduced with 50 mM DTT for 30 min at RT and also buffer-exchanged into the same buffer. GSH was prepared in the same buffer and the pH adjusted to 7.4 with NaOH. The protein concentrations were determined spectrophotometrically, using the Lambeert-Beer law. Each aliquot of oxidized HyPer1 and putative reducing protein or GSH was equilibrated to 25 °C for 10 min prior to the start of the assay. Oxidized HyPer1 (final concentration 1 μΜ) was mixed with increasing concentrations of the putative reducing protein or GSH in a quartz cuvette in a 1:1 vol ratio. The fluorescence intensities of λ_ex = 420 nm/λ_em = 512 nm and λ_ex = 500 nm/λ_em = 512 nm were recorded simultaneously every 3 s using a Cary Eclipse Fluorescence Spectrophotometer (Agilent) for 10 min at 25 °C. In case reduction was observed, the curves obtained for the λ_ex = 500 nm/λ_em = 512 nm were fitted to a plateau with a one-phase decay model, described by the following equation in Graphpad Prism version 9.2.0:

$$\mathrm{Y}=\text{IF}(\mathrm{X}<X_0,Y_0,\text{Plateau}+\left(Y_0-\text{Plateau}\right)\times e^{\left(-k_{obs}\left(X-X_0\right)\right)}$$ (7)

where X₀ is the time at which the decay begins, Y₀ is the fluorescence intensity at time up to X₀ and $\text{Plateau}$ is the Y value at infinite times.

To obtain the second-order rate constant, the $k_{obs}$ values obtained from model fitting were plotted against the putative reducing protein or GSH concentration and the second-order rate constant was obtained from the slope after performing a linear regression.

2.8. HED reduction assay

The HED assay was performed as described in Ref. [26]. In brief, 50 μΜ Grx1 and Grx2 were pre-reduced with 5 mM DTT at 37 °C for 1 h. The DTT was removed using PD SpinTrap G-25 columns (Cytiva) equilibrated in 100 mM NaPO_i, pH 7.0, 1 mM EDTA. The protein concentration was determined spectrophotometrically. 0.7 mM HED (Sigma-Aldrich 380474) was added to a freshly prepared mixture of 0.1 mg/ml BSA, 200 μM NADPH, 1 mM GSH, and 6 μg/ml glutathione reductase from yeast (Sigma-Aldrich G3664) in 100 mM NaPO_i, pH 7.0, 1 mM EDTA in a 96-well plate. After 3 min of incubation at RT, Grx1 and Grx2 were added to the wells at a final concentration of 1 μM, and buffer was added to the reference well. The decrease in absorbance at 340 nm was followed in function of time at RT using a SpectraMax spectrophotometer (Molecular Devices).

2.9. Insulin reduction assay

The insulin reduction assay was based on the protocol described in Ref. [27]. 50 μΜ Trx2 was pre-reduced with 5 mM DTT at 37 °C for 1 h. The DTT was removed using a PD SpinTrap G-25 column (Cytiva) equilibrated in 100 mM KPO_i, pH 7.0, 2 mM EDTA. The protein concentration was determined spectrophotometrically. Trx2 was added to wells of a clear 96-well plate containing 130 μΜ recombinant human insulin (Ref 11376497001, Roche Diagnostics) in 100 mM KPO_i, pH 7.0, 2 mM EDTA to a final concentration of 1 μΜ or 5 μΜ. An increase in turbidity was measured by following an increase in the absorbance at 650 nm using a SpectraMax spectrophotometer (Molecular Devices) at RT.

2.10. Detection of intracellular H₂O₂ with H2DCFDA

5 μM 2',7'-dichlorodihydrofluorescein diacetate (H2DCFDA, Invitrogen) was added to cells suspended in the full growth medium. Prior to staining, cells were either left untreated or exposed to a short oxidative pulse (100 μM H₂O₂ + 1000 U of catalase) or menadione (10 μM, 20 min). Cells were incubated with the dye for 20 min at 37 °C and 5 % CO₂. After that, fluorescence of dichlorofluorescein (DCF), the product of H2DCFDA oxidation, was immediately analyzed at λ_ex = 488 nm/λ_em = 525 nm with the CytoFLEX flow cytometer (Beckman Coulter). During the analysis, the HyPer-negative cell fraction was gated to measure the DCF signal.

2.11. Estimation of the level of HyPer1 expression

The level of cyto-, mito- and nuc-HyPer1 expression in K-562 cells was estimated by evaluating the HyPer1 fluorescence intensity ($I_{488}+I_{405}$, with subtraction of autofluorescence) in MFU units (mean fluorescence units). In addition, to compare the $k_{rd}$ values in cell populations with low and high cyto-HyPer1 expression within individual K-562 cell samples, gating of the $I_{488}+I_{405}$ plots was performed (see Supplement Fig. S2 C). HyPer1 content in each cell population was assessed by normalizing the mean $I_{488}+I_{405}$ signal from gated cells to the mean signal in the total cell population; the auto-fluorescence was subtracted from all the signals.

To compare the level of cyto-HyPer1 expression across various human cell lines, cyto-HyPer1 fluorescence intensity ($I_{488}+I_{405}$, with subtraction of autofluorescence) was normalized to the cell volume. The volume was determined in pL, using the Scepter™ 2.0 Cell Counter (Millipore).

2.12. Detection of intracellular GSH

5 μM of the ThiolTracker™ Violet, glutathione detection reagent (Invitrogen) was added to cells suspended in PBS containing Mg²⁺, Ca²⁺. Prior to staining, cells were either left untreated or preincubated in full growth medium with BSO for 24 h. Cells were incubated with the dye for 30 min at 37 °C and 5 % CO₂. Next, the dye was washed out and cell fluorescence was analyzed at λ_ex = 375 nm/λ_em = 525 nm with the CytoFLEX flow cytometer (Beckman Coulter). During the analysis, the HyPer-negative cell fraction was gated to measure the fluorescence signal.

2.13. Structure predicitions of HyPer1 complexes

The ColabFold interface [28] was used to construct Multiple Sequence Alignments (MSA) for HyPer1 with human Trx1 (P10599), human mitochondrial Trx2 (Q99757), human Grx1 (P35754), and human Grx2 (Q9NS18). The MSA was used as input for structure prediction with AlphaFold2 [29,30] using the following script: colabfold_batch --model-type alphafold2_multimer_v3 --num-recycle 48 --amber --use-gpu-relax. AlphaFold2-multimer-v3 is a specialized version of the AlphaFold2 model tailored for managing protein complexes or multimers. In the prediction process, 48 recycling steps were employed, serving as iterations where the model fine-tuned its predictions to enhance accuracy. The refinement stage utilized AMBER (Assisted Model Building with Energy Refinement), a force field commonly used in molecular dynamics simulations. Additionally, a relaxation process was implemented, optimizing the predicted structures further to attain more realistic and energetically favorable conformations. Visualisation was performed using PyMol software (https://pymol.org) [31].

2.14. Statistics

All experiments were repeated at least 3 times. Flow cytometry histograms and microscopy images shown in the paper correspond to the most representative experiments. Data are presented as means ± SD, when indicated. Statistical significance in the pairwise comparisons was evaluated by a t-test. A p-value <0.05 was considered statistically significant.

3. Results

3.1. Structural models show HyPer1 interaction with Trx1, but not with Trx2, Grx1, or Grx2

In search for oxidoreductase partners that likely reduce oxidized HyPer1 in human cells, we utilized AlphaFold2, an artificial intelligence (AI)-based system that predicts protein-protein complex formation using amino acid sequences. As input we used the sequences of HyPer1 and those of human Trx1, Trx2, Grx1, and Grx2. Notably, Grx1 is found in the cell cytoplasm, Trx1 operates in the cytoplasm and nucleus, while Trx2 and Grx2 (hereinafter by Grx2 we mean the Grx2 isoform 1) are exclusively located in mitochondria. Our analysis predicted distinct interactions between HyPer1 (Fig. 1A) and these oxidoreductases. AlphaFold2 predicted Trx1 binding to the linker between the Ec_OxyR domain and cpYFP in HyPer1 (Fig. 1 B). Interestingly, the predicted disulfide bond formed was not between the N-terminal Cys of the CXXC active site of Trx1 (Cys32), but rather by Cys73. Equally as interesting, the HyPer1 Cys forming the disulfide bond was not the Cys120 that is critical for the reaction with H₂O₂, but its disulfide-bond partner Cys380. Conversely, Grx1 showed no interaction with HyPer1. Trx2 and Grx2 were predicted to bind to the cpYFP domain or near the OxyR β-sheet, thus excluding interaction with the redox-active region of HyPer1 (Fig. 1 C and D, E).

Fig. 1.

Human Trx1 Cys73 forms a mixed disulfide with Cys380 of HyPer1, while Grx1, as well as mitochondrial Trx2 and Grx2, do not show similar interactions. Structures predicted with AlphaFold2 multimer are shown. A) The monomeric structure of HyPer1 predicted with an average pLDDT (predicted local distance difference test) score of 82.84 %. The transcription factor OxyR from E. coli (Ec_OxyR) and cpYFP (circular permutated yellow fluorescence protein) are indicated and colored according to the pLDDT score. B) The interaction structure of HyPer1 with Trx1. Trx1 is predicted to form a disulfide with HyPer1 via its Cys73, closely located to the typical C₃₂XXC₃₅ active site sequence motif. The nucleophilic cysteine Cys120 of HyPer1 is shown. The average pLDDT score is 79.14 % (39.45 % of all the atoms have a prediction score higher than 90 %). C) The predicted interaction between HyPer1 and Trx2 shows an interaction with the flexible region of β-barrel of cpYFP. The average pLDDT score is 76.76 %. D) The predicted interacion between HyPer1 and Grx1. The average pLDDT score is 77.64 %. E) The predicted interaction between HyPer1 and Grx2. The average pLDDT score is 75.66 %. Next to each structure the predicted alignment error (pae) diagram is provided. The fewer red lines at the bottom left and the upper right regions of these two-dimensional interaction diagrams, the more interactions are predicted. From comparising these diagrams, it becomes clear that Grx1 is not predicted to interact with HyPer1. In panels B, C, D, and E, HyPer1 is depicted in a light blue shade, while potential binding partners are shaded based on their pLDDT scores. Cysteines are in red stick presentation. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.2. Trx1 and GSH reduce oxidized HyPer1 in vitro

We proceeded to analyze the kinetics of oxidized HyPer1 reduction in vitro using various potential reductants. Given the promising predictions from AlphaFold for Trx1, we initially tested recombinant Trx1 and compared it with Trx2 (Fig. 2 A). Of the two Trxs, only Trx1 could reduce HyPer1 to its baseline state, achieving a ratio of 1 between the oxidized and reduced excitation maxima. In line with the AlphaFold2 prediction, Trx2 displayed only a slight non-specific activity (we could conclude the non-specificity due to a lack of dependency of the response on the Trx2 concentration) (Fig. S1 A), while Trx2 did prove to be active in an insulin reduction assay (Fig. S1 B). We then determined the second-order rate constant for Trx1 reduction with oxidized HyPer1 as substrate by determining the k_obs values for various Trx1 concentrations. From the plot of the k_obs against Trx1 concentration (Fig. 2 B), we derived a second-order rate constant of 5.8 (±1.04) x 10² M⁻¹s⁻¹.

Fig. 2.

Trx1 and GSH reduce oxidized HyPer1, while Trx2, Grx1 and Grx2 do not. A) 1 μΜ οxidized HyPer1 was reacted with 30 μΜ reduced Trx1 or Trx2. The ratio between the two HyPer1 excitation maxima (500 nm and 420 nm with emission at 512 nm) is presented. The result shown is the mean of 3 replicates and the dotted line indicates the SEM. B) $k_{obs}$ obtained for increasing concentrations of Trx1 vs Trx1 concentration (an example of the kinetic curve used for $k_{obs}$ calculation is presented in the inset). The equation was derived by performing a linear regression. The individual points denote the mean and the error bars indicate the SD. The dotted line marks the 95 % confidence interval. The experiment was performed in triplicates. C, D) 1 μΜ οxidized HyPer1 was reacted with 20 μΜ Grx1 and Grx2 (10 mM DTT was used as a control) (C) or increasing concentrations of GSH (D). The ratio between the two HyPer1 excitation maxima (500 nm and 420 nm) with emission at 512 nm is presented. The results shown are the mean of 3 replicates, and the dotted line indicates the SD. E) $k_{obs}$ obtained for increasing concentrations of GSH vs GSH concentration (an example of the kinetic curve used for $k_{obs}$ calculation is presented in the inset). The equation was derived by performing a linear regression. The individual points denote the mean and the error bars indicate the SD. The dotted line marks the 95 % confidence interval. The experiment was performed in triplicates.

Abbreviations: $k_{obs}$, rate constant of HyPer1 reduction; SEM, standard error of the mean; SD, standard deviation; DTT, dithiothreitol.

Even though Grx1 and Grx2 did not yield a favorable complex (Fig. 1 D, E), we decided to evaluate the ability of Grx1 and Grx2 to reduce HyPer1 in an identical experimental arrangement as for Trx1 and Trx2. As can be seen in Fig. 2C, neither Grx1 nor Grx2 reduce HyPer1, even when the concentrations significantly surpass physiological levels, while they were both active in an HED deglutathionylation assay (Fig. S1 C).

We next assessed whether GSH can reduce HyPer1. Indeed, we observed that GSH can reduce HyPer1 at concentrations similar to those encountered intracellularly (Fig. 2 D). Calculating the second-order rate constant as for Trx1, we obtained a second-order rate constant of 1.77 (±0.31) M⁻¹s⁻¹ (Fig. 2 E). Taken together, our results suggest that both Trx1 and GSH can reduce HyPer1, with Trx1 exhibiting approximately a 100-times greater efficiency. Given that the GSH concentrations we employed are in the intracellular GSH concentration range, we can assume that the reduction by GSH can become relevant in compartments lacking Trx1, such as the mitochondria.

3.3. Real-time monitoring of HyPer1 reduction in the cytosol, nucleus, and mitochondria of living cells

To investigate the pathways of HyPer1 reduction in cellulo, we performed experiments on human myelogenous leukemia K-562 cells expressing HyPer1 in different compartments: the cytoplasm (cyto-HyPer1), the nucleus (nuc-HyPer1), and the mitochondria (mito-HyPer1). The localization of the sensor was confirmed by confocal microscopy (see Fig. 3).

Fig. 3.

Confocal microscopy images of K-562 cells expressing cyto-HyPer1 (left), nuc-HyPer1 (middle), and mito-HyPer1 (right). Bar = 10 μm, MOI (multiplicity of infection) = 5.1 column.

Our first objective was to establish a procedure for examining the kinetics of HyPer1 reduction in living cells after exposure to a short pulse of exogenous H₂O₂ (Fig. 4 A). When developing the protocol, we took the following observations into account [22]: i) in human cells, under normal physiological conditions, HyPer1 is fully reduced; ii) addition of 100 μM exogenous H₂O₂ to cells typically leads to the full oxidation of HyPer1 within 5 min. To this end, to create a short oxidative pulse, we used a sequential approach of addition of exogenous H₂O₂ (100 μM) and catalase (1000 U) to the cell medium with a 5 min interval between additions. H₂O₂, penetrating inside the cells, completely oxidizes HyPer1. In turn, catalase, being incapable of penetrating the cell membrane, promptly neutralizes surplus peroxide, which includes both the dissolved peroxide within the medium and that which leaks out of the cells. The extracellular catalase together with the intracellular H₂O₂-detoxifying enzymes swiftly creates an environment favoring HyPer1 reduction. To investigate the reduction kinetics, we traced the HyPer1 fluorescence signals just after the addition of catalase. For each cell, the intensities of $I_{488}$ (EX488/FL525 signal from oxidized HyPer1) and $I_{405}$ (EX405/FL525 signal from reduced HyPer1) were measured. A simultaneous drop in the $I_{488}$ signal and increase in the $I_{405}$ in time evidenced a rapid decline in the proportion of oxidized biosensor and accumulation of its reduced form, respectively (Fig. 4 A). Both signals returned to their basal levels within approximately 10 min after the pulse.

Fig. 4.

Rate constants of HyPer1 reduction in the cytosol, nucleus and mitochondria of K-562 cells after a short exposure to exogenous H₂O₂. A) Experimental protocol with an example of the recorded fluorescence signals: drop in the $I_{488}$ signal and concomitant increase in the $I_{405}$ signal indicate rapid reduction of cyto-HyPer1 upon oxidative load removal. B) An example of fitting the kinetic curve $OxD\left(t\right)$ with the exponential law (Eq. (6)) and determining $k_{rd}$ - the rate constant of cyto-HyPer1 reduction. C-E) Validation of the experimental protocol: H2DCFDA staining of the control (untreated) cells and cells exposed to the short oxidative pulse (H₂O₂+catalase) did not reveal traces of residual H₂O₂ during the $k_{rd}$ measurements (C); rate constant of cyto-HyPer1 reduction is not affected by the variations in the cell density (2-fold increase, D), H₂O₂ concentration (25-fold increase, E) and amount of catalase added (2-fold increase, E). The value of $k_{rd}$ does not depend on the presence of EUK-134 (intracellular catalase mimetic) in cells (E). F) Comparison of the rate constant of HyPer1 reduction in the cytosol, nucleus, and mitochondria of K-562 cells. Abbreviations: $I_{488}$, EX488/FL525 signal; $I_{405}$, EX405/FL525 signal; MFU, mean fluorescence units; $OxD$, oxidized HyPer1 fraction; $k_{rd}$, rate constant of HyPer1 reduction.

To quantify the rate of HyPer1 reduction, we processed the obtained raw data (Fig. 4 B). First, we performed normalization and calibration of fluorescence signals from individual cells (Eq. (1) and (2), Materials and Methods Section 2.5), that enabled linking the intensity of cellular fluorescence ($I_{488}$ and $I_{405}$) to the proportion of oxidized HyPer1 molecules in a cell ($OxD$). After that, we fitted the obtained kinetic curve (Fig. 4 B) with an exponential decay equation (Eq. 4-6, Materials and Methods, Section 2.6) to derive the pseudo first-order rate constant of HyPer1 reduction $k_{rd}$, which reflects the activity of the enzymatic pathway responsible for biosensor reduction and actually corresponds to the $k_{obs}$ constant measured in the experiments with purified recombinant proteins.

To validate the elaborated protocol of $k_{rd}$ estimation, we ran several tests using K-562 cells expressing cyto-HyPer1. We first checked whether exogenous H₂O₂ remains in the medium and within cells during the $k_{rd}$ measurements, as the presence of residual H₂O₂ could have influenced the results obtained. To this end, we used the H2DCFDA dye. Of note, while H2DCFDA is not an H₂O₂ specific dye, and its use is often associated with various artifacts, it is known to reflect changes in intracellular peroxide levels [32]. Flow cytometric analysis of K-562 cells stained with the H2DCFDA did not reveal any difference between the cell samples pre-treated with H₂O₂/catalase and untreated cells (Fig. 4C and Fig. S2 A), indicating no residual exogenous H₂O₂. Consistent with this observation, cellular preincubation with the intracellular catalase mimetic EUK-134 prior to the experiments yielded no discernible effect on the measured $k_{rd}$ value (Fig. 4 E). This further validates the absence of residual H₂O₂ throughout the measurements. Next, we verified that alterations in cell suspension density (Fig. 4 D), or the quantity of added catalase (Fig. 4 E) did not exert any discernible impact on the dynamics of cyto-HyPer1 reduction, particularly within a 2x range of variations. Importantly, the application of very different H₂O₂ concentrations also resulted in the same $k_{rd}$ value (Fig. 4 E), which may be explained by the fast recovery of the cellular redox systems after a short oxidative pulse. Finally, to make sure that all parameters of the cell samples that can influence the accuracy of the results (such as intracellular pH, sample temperature, etc.) are stable during $k_{rd}$ measurement, we carried out the experiments with K-562 cells expressing cyto-SypHer. SypHer is an H₂O₂-insensitive variant of HyPer1 [16], which shares indistinguishable fluorescence properties and pH sensitivity with HyPer1, factors that could potentially influence the precision of $k_{rd}$ measurements, thus making it an ideal control. Notably, the introduction of H₂O₂ (100 μM) and catalase (1000 U) did not affect cyto-SypHer fluorescence (Fig. S2 B). This outcome validates our protocol of $k_{rd}$ measurement in cells expressing HyPer1.

Finally, using the established protocol, we investigated HyPer1 reduction in the cytosol, nucleus, and mitochondria of K-562 cells. Our data analysis revealed that the reduction kinetics fit a mono-exponential curve in all cellular compartments. The rate constant for cyto-HyPer1 reduction in K-562 cells calculated using Eq. (6) (Materials and Methods, Section 2.6) was determined to be 0.011 ± 0.002 s⁻¹, and this measurement was consistent across numerous independent experiments. The value of $k_{rd}$ for the nuc-HyPer1 turned out to be very similar (0.009 ± 0.002 s⁻¹), while the reduction rate of the mito-HyPer1 appears to be much lower (0.005 ± 0.001 s⁻¹, see Fig. 4 F).

3.4. Trx1 reduces HyPer1 in the cytosol and nucleus, and GSH reduces HyPer1 in the mitochondria

Next, we aimed to identify which enzymatic systems are involved in HyPer1 reduction in the cytosol, nucleus, and mitochondria. K-562 cells expressing cyto-HyPer1 were pre-incubated with auranofin [22], mitomycin C [33], or Tri-1 [34] – three different inhibitors of TrxR [35], a selenocysteine-containing enzyme that provides electrons to the Trxs. All inhibitors caused a dose-dependent drop in the rate of cyto-HyPer1 reduction (Fig. 5 A), which was estimated using Eq. (4) (Materials and Methods, Section 2.6). At high doses of inhibitors, the reduction rate decreased by approximately ten-fold. In contrast, treating K-562 cells with Na₂SeO₃, utilized to provide enough selenium for enhanced TrxR synthesis, led to an increase in the cyto-HyPer1 reduction rate (Fig. 5 A). This elevation in reduction rate provides compelling proof of the significant involvement of the Trx/TrxR pathway in the HyPer1 reduction process. To evaluate the contribution of the GSH-dependent disulfide-reducing system, we employed an inhibitor of GSH synthesis called buthionine sulfoximine (BSO), as well as the glutathione reductase inhibitor BCNU (1,3-Bis(2-chloroethyl)-1-nitrosourea). Pre-incubation of cells with BSO resulted in a two-fold decrease of intracellular GSH content (see Fig. S3), but affected the $k_{rd}$ value of cyto-HyPer1 only minimally (Fig. 5 A). BCNU also only had a negligible effect on the $k_{rd}\text{.}$ Finally, simultaneous inhibition of both Trx/TrxR and GSH systems almost completely abolished the cyto-HyPer1 reduction (Fig. 5 B). Thus, we concluded that Trx plays a pivotal role in the reduction of cytosolic HyPer1, with GSH contributing only minimally to this process.

Fig. 5.

In the cytosol and nucleus of human cells, the Trx1/TrxR1 enzymatic system plays a significant role as a primary reductant for HyPer1. Within the mitochondria, HyPer1 reduction is reliant on GSH. HyPer1 overexpression may disrupt disulfide-reducing capacity in cellular compartments. A) Inhibition of TrxR activity with Aur (20, 50, 100, 200 nM), MitC (100, 200, 500 μM), or Tri-1 (200, 500, 1000 nM) causes a dose-dependent decrease in cyto-HyPer1 reduction in K-562 cells, whilst inhibition of GSH production with BSO (0.5, 1 mM) and of GSSG reduction by BCNU (10, 25 μΜ) has no significant effect. Stimulation of TrxR production by Se supplementation (100, 500, 1000 nM) increases the rate of cyto-HyPer1 reduction. B) Simultaneous inhibition of both Trx- (Aur, 100 nM) and GSH-dependent (BSO, 0.5 mM) pathways almost completely abolishes cyto-HyPer1 reduction in K-562 cells. C, D) Inhibition of TrxR activity with Aur causes a dose-dependent decrease of nuc- but not mito-HyPer1 reduction in K-562 cells, while inhibition of GSH synthesis with BSO and of GSSG reduction by BCNU exerts an opposite effect: it affects the rate of HyPer1 reduction within mitochondria and insignificantly affects the recovery of nuc-HyPer1 (concentration of inhibitors: Aur - 0, 50, 100, 200 nM; Tri-1 - 0, 500, 1000 nM (C) and 0, 1000 nM (D); BSO - 1 mM; BCNU – 25 μΜ). E) Levels of HyPer1 expression in the compartments of K-562 cells transduced at different MOI; intracellular biosensor content is estimated based on the HyPer1 fluorescence signals. F) Dependence of the HyPer1 reduction rate on the level of biosensor expression in the cytosol, nucleus and mitochondria of K-562 cells: at low levels of HyPer1 expression, $k_{rd}$ does not depend on the intracellular HyPer1 content (solid markers, MOI ≤5 for cyto-HyPer1 and MOI ≤2 for nuc- and mito-HyPer1), whereas high expression of HyPer1 (open markers) disturbs compartmentalized disulfide-reducing capacity. Data in A and C–F are shown as mean ± SD (N > 3). A,C,D: *p < 0.05 – pairwise comparison to the control (without inhibitor) cells. Abbreviations: $OxD$, oxidized HyPer1 fraction; $k_{rd}$, rate constant of HyPer1 reduction; Cntr, control (without inhibitors) cells; Aur, auranofin; MitC, mitomycin C, BSO, buthionine sulfoximine; BCNU, 1,3-Bis(2-chloroethyl)-1-nitrosourea; MOI, multiplicity of infection; MFU, mean fluorescence units.

In the nucleus, a similar reductive HyPer1 response was observed. Inhibition analysis showed that the rate of nuc-HyPer1 reduction decreased upon TrxR inhibition (including specific TrxR1 inhibition with Tri-1) and only marginally reacted to the suppresion of GSH synthesis (Fig. 5C).

In the mitochondrial matrix, on the other hand, the HyPer1 response displayed a different pattern: the reduction rate of mito-HyPer1 remained unaffected by TrxR inactivation using both auranofin and Tri-1, while being responsive to the decline in GSH levels (Fig. 5 D). The two-fold decrease in the GSH content induced by BSO pretreatment resulted in a two-fold decrease of the mito-HyPer1 reduction rate, and treatment with BCNU affected the $k_{rd}$ value even more drastically.

Taken together, these results indicate that the Trx1/TrxR1 pathway is the main reductant of HyPer1 in the cytosol and nucleus, while GSH-mediated reduction of HyPer1 prevails in the mitochondrial matrix. It is important to point out that these findings align perfectly with the in silico and in vitro results (see Sections 1, 2, 3). Indeed, according to our in vitro results, Trx1 reduces HyPer1 with k_obs rate constants in the same range as the $k_{rd}$ values observed in the cytosol and nucleus of living cells. Our in vitro findings further established that mitochondrial Trx2 lacks the capability to reduce HyPer1. They also confirmed that GSH can act as an alternative mechanism for HyPer1 reduction when Trx1 is absent, becoming particularly crucial within the mitochondrial matrix.

3.5. The HyPer1 reduction rate depends on the HyPer1 concentration in the cell compartments at high but not low levels of HyPer1 expression

To find out whether HyPer1 expression can overload the reductive capacity of the cellular compartments, we transduced K-562 cells with cyto-, nuc-, and mito-HyPer1 at different multiplicity of infection (MOI) values (Fig. 5 E). The percentage of cells expressing HyPer1 depends on the MOI, ranging from 40 % to 95 % (see Fig. S2 C as an example). The MOI denotes the number of viral particles harboring the HyPer1 gene per cell during transduction. This value determines the final count of integrated HyPer1 gene copies within the genome. Furthermore, extended cultivation of cells expressing HyPer1 typically leads to a reduction in sensor expression. So, after a period of cell cultivation, we obtained a series of samples with varying HyPer1 content, influenced by both the MOI and the duration of cell cultivation. We measured $k_{rd}$ in these sets of samples (MOI = 1–10), and found that in cells with a high level of cyto-, nuc- and mito-HyPer1 expression (assessed by the biosensor total fluorescence signal), $k_{rd}$ negatively correlates with HyPer1 content (Fig. 5 F, open markers). However, no such correlation was observed in cases of low sensor expression (Fig. 5 F, solid markers). To further validate this observation, we gated cell populations with different cyto-HyPer1 content within one K-562 sample with low HyPer1 expression (MOI = 5, Fig. S2 C). The analysis revealed no significant disparity in $k_{rd}$ values among gated cell fractions, despite several-fold differences in sensor expression levels (Fig. S2 D).

In summary, the results we obtained indicate that elevated HyPer1 expression within cellular compartments could potentially overwhelm the Trx1-or GSH-mediated reducing capacity in cells. Conversely, the lower quantity of biosensor present in K-562 cells (achieved at MOI = 1, 2 for nuc- and mito-HyPer1, and MOI = 5 for cyto-HyPer1) does not disrupt the compartmentalized redox environment, at least within the utilized experimental conditions.

3.6. Applying the standardized protocol for k_rd measurements enables the quantification of disulfide-reducing capacity in various biological contexts

Finally, we decided to apply the developed protocol of $k_{rd}$ measurements to compare the Trx1-mediated reductive activity in the cytosol of human cells from diverse sources, such as HeLa, dermal fibroblasts (Fibr), mesenchymal stem/stromal cells (MSCs), as well as induced pluripotent stem cells (iPS) obtained by MSC reprogramming [22], all expressing cyto-HyPer1. Inhibition analysis yielded similar outcomes to those observed for the K-562 line: inhibition of TrxR activity led to a a significant decrease in the rate constant of cyto-HyPer1 reduction, while inhibiting both Trx- and GSH-controlled thiol-disulfide exchange virtually abolished cyto-HyPer1 reduction (Fig. 6 A, B). Sensitivities to inhibitors varied among cells, exemplified by HeLa cells displaying remarkable resistance to TrxR inhibition, particularly with auranofin. Notably, only concentrations of auranofin surpassing 2 μM resulted in a 90 % reduction in the $k_{rd}$ value. Interestingly, the HeLa cell line also featured a high $k_{rd}$ value (Fig. 6C).

Fig. 6.

Comparison of Trx1/TrxR1 pathway activity in different human cells. A) Kinetics of cyto-HyPer1 reduction in human MSCs in control cells and cells treated with Aur (100 nM). B) Inhibition of TrxR activity with Aur (1000 nM) decreases the rate of cyto-HyPer1 reduction in all human cell lines tested; dual TrxR/GSH inhibition (100 nM Aur+0.5 mM BSO) completely blocks biosensor reduction. C) Kinetics of cyto-HyPer1 reduction in normal human cells and in cancer (HeLa) cells. D) Rate constant of cyto-HyPer1 reduction does not correlate with the level of biosensor expression in various cell lines - the expression level was estimated by normalizing the mean total fluorescence signal from cellular HyPer1 to the mean cellular volume (see Materials and Methods, Section 2.9) Data in B, D are shown as mean ± SD (N > 3). B: *p < 0.05, derived from the pairwise comparison to the Cntr samples. Abbreviations: $k_{rd}$, rate constant of HyPer1 reduction; Cntr, control (without inhibitors) cells; Aur, auranofin; BSO, buthionine sulfoximine; MSCs, mesenchymal stem/stromal cells; Fibr, dermal fibroblasts; iPS, induced pluripotent stem cells, MFU, mean fluorescence units.

Overall, notable variations in $k_{rd}$, ranging up to several folds, were identified among distinct cell types. The highest $k_{rd}$ values, indicating the most efficient Trx1-dependent system for disulfide reduction, were identified in the cytosol of cancer cell lines - K-562 and HeLa. Conversely, normal human cells sharing a similar phenotype (MSCs and fibroblasts) displayed comparable lower $k_{rd}$ values (Fig. 6 D). Importantly, no correlation was observed between the $k_{rd}$ value measured in various cell lines under normal physiological conditions and the level of cyto-HyPer1 expression within those lines (Fig. 6 D), underscoring the robustness of the conducted comparative analysis.

Our final objective was to explore how oxidative stress induced by disrupting cellular redox homeostasis impacts the Trx1-dependent disulfide-reducing activity within the cytosol. To induce such stress in K-562 cells, we used menadione – a K3 vitamin precursor known to disturb mitochondrial metabolism and exert pro-oxidative effects in cells [36]. The pro-oxidative impact of a 20 min menadione treatment on K-562 cells was confirmed by an increased oxidation of H2DCFDA (Fig. 7A and B) and the elevation in the basal level of cyto-HyPer1 oxidation (Fig. 7C). Under basal conditions, in menadione-treated cells, approximately 40-50 % of cyto-HyPer1 was oxidized. Measurement of the $k_{rd}$ value indicated that the rate constant of HyPer1 reduction was nearly half of the biosensor reduction rate compared to untreated cells (Fig. 7 C and D). Hence, we can conclude that the disruption of K-562 cell redox homeostasis induced by menadione leads to a partial depletion of the reductive activity of Trx1/TrxR1 pathway.

Fig. 7.

Estimation of the Trx1/TrxR1 pathway activity under conditions of menadione-induced oxidative stress. A,B) H2DCFDA staining of K-562 cells treated with increasing concentrations of menadione for 20 min confirms oxidative stress induction. C,D) Decrease of the rate of cyto-HyPer reduction under conditions of menadione-stimulated oxidative stress: kinetics of HyPer reduction in the control K-562 cells and cells treated with menadione (10 μM, 20 min) (C), quantification of $k_{rd}$ (D). Data in B,D are shown as mean ± SD (N > 3); *p < 0.05, derived from the pairwise comparison to Cntr samples. Abbreviations: $k_{rd}$, rate constant of HyPer reduction; Cntr, control (without menadione) cells; Men, menadione; DCF, dichlorofluorescein.

4. Discussion

One of the championed advantages of genetically encoded biosensors for H₂O₂, as opposed to chemical dyes, is their reversibility. Indeed, the working principle of those biosensors, be it those of the roGFP- or HyPer-based family, is cysteine thiol oxidation. The disulfide bond formed then becomes susceptible to reduction by the reducing systems of the cell. Lamentably, astoundingly little attention has hitherto been given to the reductive half-reaction, with the exception of only a few independent studies [20,37,38]. We consider this a serious omission, since understanding the mechanisms that govern this reaction is crucial to correctly interpreting the biosensor signal, determining whether an increase in the biosensor signal is due to elevated H₂O₂ levels or a shortage of reducing equivalents (e.g. NADPH) [39]. Therefore, in the current study we aimed to uncover the pathways involved in the reduction process of HyPer1 in different compartments of human cells. To approach this, we created a HyPer-based method capable of precisely measuring the reductive activity within cellular compartments.

At first, we took advantage of the latest AI resources and used AlphaFold2 to find out which human oxidoreductases might be the most likely reducing partner for HyPer1. To our knowledge, this is the first time AlphaFold has been utilized to predict thiol-based interactions in human proteins, having previously only been employed to predict disulfide bonds in proteins of the plant kingdom [40]. The AlphaFold2 prediction indicated that HyPer1 interacts with the cytosolic enzyme Trx1, but not with Grx1, while mitochondrial Trx2 and Grx2 were predicted to not interact. The ability of Trx1 to reduce the active site disulfide of HyPer1 was subsequently confirmed by both our in vitro and in cellulo experiments. Importantly, the cyto-HyPer1 reduction rate measured in living cells ($k_{rd}$) was very close to the $k_{obs}$ value observed in vitro at physiologically relevant concentrations of Trx1 (10 μΜ), further supporting the validity of our findings.

Surprisingly, HyPer1-Trx1 mixed disulfide bond predicted by AlphaFold2 was between Cys73 of Trx1, which does not belong to the typical CXXC active site sequence motif of thioredoxins, and the non-peroxidatic Cys of the OxyR incorporated into HyPer1. While experimental evidence with mutants in vitro would be required to confirm this prediction, several previous studies have demonstrated that Cys73 of Trx1 can, in principle, engage in disulfide bond formation, albeit uniquely in forming intermolecular disulfide bonds with another Trx1 [41]. Experiments with C73S mutants have shown that while it is not involved in redox reactions, it can have an important function in regulating the activity of Trx1, as the formation of a Cys73-Cys73 homodimer would block the Cys32-Cys35 active site [42,43]. It should also be kept in mind that HyPer1 is an artificial, engineered protein, and therefore may participate in redox reactions not commonly observed in native proteins. The regioselectivity of the Trx1 Cys73 towards Cys280 or Cys121 can computationally be assessed by calculating the difference in the local softness between Cys73 and Cys121 vs Cys73 and Cys280. Local softness is a parameter that depends on the system polarizability and the Fukui function – a local function that would indicate where a molecule would preferentially react, which is in turn related to the electron density of the HOMO (highest occupied molecular orbital) of the nucleophilic atom and the LUMO (lowest unoccupied molecular orbital) of the electrophilic atom [44]. However, delving into this is beyond the scope of the current study.

In addition to Trx1, GSH was also found to reduce HyPer1 with a reaction rate that is 100 times slower. Taking into account that the intracellular concentration of GSH is usually several orders of magnitude higher than that of Trx1, this observation nevertheless allows us to consider the reaction of HyPer1 with GSH as one of the possible pathways for sensor reduction in living cells. Interestingly, the reducing partners of Ec_OxyR (H₂O₂-sensing domain of HyPer1) found in this study differ from those found in a bacterial cell, where Ec_OxyR is recycled by Grx1. This could be attributed to the differences in sequences and structures of Grx between bacteria and eukaryotic species. In particular, yeast studies revealed that when Ec_OxyR is ectopically expressed in S. pombe, it is primarily reduced by the Trx-, but not Grx-dependent enzymatic system [45]. Besides that, two independent studies on HyPer7 (the latest ultrasensitive addition to the HyPer family of probes) found that HyPer7 also undergoes predominant reduction by the Trx system in budding [38] as well as fission [37] yeast. H. sapiens and S. cerevisiae Trx1 sequences exhibit a 61.39 % similarity and a 47.52 % identity, including the CGPC active site sequence motif. H. sapiens Trx2, which was not functional in reducing HyPer1, has a lower identity. Hence, our finding that human Trx1 but not Trx2 reduces HyPer1 was somewhat expected, albeit it should be noted that HyPer7 utilized OxyR from another species – N. meningitidis, as its sensing domain, and is more sensitive.

Perhaps our most exciting finding is the observation that distinct reducing systems operate in reducing HyPer1 in the cytosol and mitochondria of human cells: Trx1 serves as the primary reductant in the former, while in the latter, it is reduced by GSH, but not Trx2 or Grx2. Again, as in the case of Trx1 and cyto-HyPer1, the mito-HyPer1 reduction rate measured in living cells ($k_{rd}$) was very close to the $k_{obs}$ value observed in vitro at physiologically relevant concentrations of GSH (5–10 mΜ). In general, in the light of ever accumulating evidence for a strong compartmentalization of redox processes in cellular compartments, and even subcompartments, the fact that cyto- and mito-HyPer1 are reduced by different partners does not come as a complete surprise. Indeed, it has long been known that mitochondria house their unique set of reducing enzymes, Trx2 and Grx2, which are distinct in sequence from the cytosolic counterparts Trx1 and Grx1. Additionally, there is little overlap in terms of their substrates (8/247 for Trx1 vs Trx2) [46]. Likewise, it is also not uncommon for Trxs and Grxs in general to have distinct substrates. This can be attributed to variation in their redox potentials and the presence or absence of steric tensions that could hinder the quasi-linearity of the transition state needed for an SN₂ nucleophilic attack [47]. Indeed, steric hinderance could be a plausible explanation of why HyPer1 undergoes non-enzymatic reduction in mitochondria, rather than being reduced by Trx2 or Grx2, which would have been a more efficient process.

To date, the majority of studies on redox compartmentalization, whether using biosensors or not, have mainly focused on the oxidative half-reaction. As a result, compartmentalization is primarily explained based on proximity to H₂O₂ generation sites [48] and on peroxidases that can contribute to the creation of H₂O₂ gradients through their H₂O₂ scavenging activities [[49], [50], [51], [52], [53]]. At the same time, numerous observations indicate that the reductive half-reaction is of equal importance. A very illustrative example is provided by roGFP2, which is known to equilibrate with GSH via Grxs [54]. When comparing the oxidation state of roGFP2 (or rxYFP, another redox-sensitive protein) in organelles with (the cytosol) and without (peroxisomes and the intermembrane space of mitochondria) Grxs, it was noticed to be considerably more oxidized in the organelles lacking Grxs [55]. These differences were completely abolished upon the expression of a fused version of roGFP2-Grx1 in the same organelles [56]. Without considering the reductive half-reaction, incorrect conclusions about the presence of the oxidant might have been reached. Another striking example of the need to take into account the reduction of biosensors comes from studies utilizing HyPer7 aimed at establishing a connection between the extent of H₂O₂ release from the mitochondria and the activity of the Trx system [37,52,57]. In these studies, the role of the Trx system was attributed to the reduction of the peroxiredoxins, which compete with HyPer7; the possibility of the biosensor being directly reduced was not considered. In the absence of a stimulated H₂O₂ generation, an increase in biosensor oxidation could either be interpreted as being due to reducing equivalents becoming limited, and causing the equilibrium to shift towards the oxidative half-reaction, or as an indication that a deficiency of reducing equivalents prevents the recycling of peroxidases, resulting in the accumulation of H₂O₂ that drives sensor oxidation. Whether it is possible to disentangle these two explanations if only following HyPer oxidation in cells becomes an almost philosophical question. This is why our method of following HyPer reduction is an essential complementary method to fully interpret the biosensor response.

Our results across various cell lines, where distinct $k_{rd}$ values were measured (Fig. 6 D), additionally underscore the significance of considering the reductive half-reaction and the need to develop protocols to quantify the cellular reductive activity. The obtained results indicate that cancer cells exhibit significantly higher reduction rates. This is consistent with previously published data of HyPer-based research [20] and the fact that cancer cells tend to overexpress components of the Trx system, rendering the Trx system a promising target for anticancer interventions [34,58]. The protocol established in our study could readily be adapted to high-throughput cellular screens to search for irreversible inhibitors of the Trx system directly in cells, a task typically performed in vitro at present [59].

In addition, using the elaborated protocol, we found that HyPer1 overexpression disrupts the compartmentalized disulfide-reducing activity, aligning seamlessly with previously published data [60]. Importantly, we established that the mitochondrial and nuclear HyPer1 reduction is particularly sensitive to an excess of biosensor, probably due to the smaller volume of these compartments in comparison to the cytosolic one, resulting in locally high concentrations of HyPer1.

Generally, when considering the methodological implications of the conducted research, it should be noted that the $k_{rd}$ values, which we used here to quantify the reductive activity in cellular compartments, may primarily reflect the ability of cells to recover from transient oxidative stress, rather than representing the basal activity of reductive enzymes in undisturbed cells. One might infer that as a result of the oxidative pulse, which was utilized to oxidize HyPer1, even after the intracellular H₂O₂ level returns to normal values, a significant number of oxidized proteins could persist within the cell, potentially compromising the intracellular disulfide-reducing capacity. In particular, this factor may affect the $k_{rd}$ values obtained for various cellular compartments. At the same time, based on our findings, it appears that the reductive capacity post the H₂O₂ pulse is not completely exhausted, since we observed an impact of high but not low HyPer1 loads, as well as the effects of menadione and TrxR/GSH inhibition on the $k_{rd}$ values. Moreover, the lack of dependence of $k_{rd}$ values on the intensity of oxidative pulse (i.e. peroxide concentrations used) may indicate that, under conditions of undepleted reductive activity, short exposure to H₂O₂ does not significantly disturb the cellular redox environment. This is a likely scenario given the role of several peroxidases, among which the very effective peroxiredoxins [61].

When considering potential future applications of the developed method, it is crucial to highlight that the kinetic equations used in this study (Eqs. (4), (5), (6))), initially introduced for investigating intracellular redox reactions by Brito and Antunes [25], enable the discrimination between oxidative and reductive processes. This allows for the simultaneous and unambiguous assessment of the rate constants for both HyPer1 reduction and oxidation. Therefore, the protocol we established here can also be used to characterize the reductive activity in cells with a pro-oxidative background, as illustrated in this study using menadione. Besides that, from a practical point of view, it is crucial to highlight that the parameter $OxD$ [24], used in this study for calculating $k_{rd}$, can be measured not only through flow cytometry but also using a fluorescence plate reader or microscope, which significantly broadens the potential applications of the method.

In summary, by combining in silico, in vitro, and in cellulo methods, we have for the first time systematically investigated the reduction system responsible for reducing a HyPer family probe in mammalian cells. We have shown that while the Trx/TrxR system reduces HyPer1 in the cytosol and nucleus, HyPer1 in the mitochondrial matrix is reduced directly by GSH at a cellular rate that is approximately two times slower. Our findings underscore the importance of considering the reductive half-reaction when interpreting HyPer1 response, especially when comparing it across various organelles. Future research will focus on the unraveling the reduction system responsible for mediating the reduction of HyPer7 – the most recent and sensitive member of the HyPer family of probes.

CRediT authorship contribution statement

Andrei Zhuravlev: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Daria Ezeriņa: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Julia Ivanova: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Nikita Guriev: Data curation, Investigation, Methodology, Software. Natalia Pugovkina: Investigation, Methodology. Alla Shatrova: Investigation, Methodology. Nikolay Aksenov: Investigation, Methodology. Joris Messens: Conceptualization, Formal analysis, Funding acquisition, Investigation, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing. Olga Lyublinskaya: Conceptualization, Formal analysis, Funding acquisition, Investigation, Project administration, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.