New sensors for quantitative measurement of mitochondrial Zn²⁺

Abstract

Zinc (Zn²⁺) homeostasis plays a vital role in cell function, and the dysregulation of intracellular Zn²⁺ is associated with mitochondrial dysfunction. Few tools exist to quantitatively monitor the buffered, free Zn²⁺ concentration in mitochondria of living cells ([Zn²⁺]_mito). We have validated three high dynamic range, ratiometric, genetically encoded, fluorescent Zn²⁺ sensors that we have successfully used to precisely measure and monitor [Zn²⁺]_mito in several cell types. Using one of these sensors, called mito-ZapCY1, we report observations that free Zn²⁺ is buffered at concentrations about 3 orders of magnitude lower in mitochondria than in the cytosol, and that HeLa cells expressing mito-ZapCY1 have an average [Zn²⁺]_mito of 0.14 pM, which differs significantly from other cell types. These optimized mitochondrial Zn²⁺ sensors could improve our understanding of the relationship between Zn²⁺ homeostasis and mitochondrial function.

Zn²⁺ is a micronutrient that is required for human life, and deficiency can lead to impaired cognition, immune dysfunction, diarrhea, and death, particularly in children under the age of 5 years.¹ Although Zn²⁺ is essential for cell function, accumulation of Zn²⁺ to toxic levels leads to cell death. The human genome encodes two-dozen Zn²⁺-specific transporters and many metal-buffering proteins, which are expressed in a tissue-specific manner.² Our current understanding of Zn²⁺ homeostasis is that intracellular Zn²⁺ is distributed into a large pool of structural or catalytic Zn²⁺ that is tightly bound, and two smaller pools of Zn²⁺: free Zn²⁺ and exchangeable Zn²⁺ bound loosely to small molecule or protein partners.^3,4 Zn²⁺ homeostasis can be altered in diseases, such as neurodegeneration. ⁵ In order to effectively study Zn²⁺ biology, we must be able to observe and manipulate Zn²⁺ specifically with subcellular resolution.

Cellular Zn²⁺ homeostasis affects mitochondrial function through poorly understood mechanisms. Zn²⁺ toxicity can lead to the release of cytochrome c from the intermembrane space, caspase activation, and apoptosis.^6–8 Changes in Zn²⁺ availability can affect metabolism, including oxidative phosphorylation. ^9,10 Intracellular Zn²⁺ can depolarize mitochondria and decrease mitochondrial movement.^11–13 Lastly, it is likely that mitochondria are a source and sink of Zn²⁺ in neurons and other cells.^3,14–16 To understand how Zn²⁺ homeostasis affects mitochondrial function, we must be able to measure and monitor mitochondrial Zn²⁺.

Few tools exist to observe mitochondrial Zn²⁺ homeostasis in living cells with high specificity. Small molecule fluorescent probes are arguably the most popular tools. FluoZin-3 increases in intensity upon binding Zn²⁺ and has been used to observe free Zn²⁺ in isolated mitochondria.^17,18 Positively charged probes, such as RhodZin-3, concentrate within mitochondria of intact cells due to the negative mitochondrial inner membrane potential (Δψ_m).^17,19 Consequently, it is problematic to monitor Zn²⁺ in depolarized mitochondria using such probes because a decrease in fluorescence intensity could be caused by either a decrease in Zn²⁺ or in Δψ_m. Other mitochondrial sensors consist of both a small molecule fluorophore and a protein component, which can be genetically targeted to mitochondria. This approach was used to target the fluorescent Zn²⁺ probe Zinpyr1 to mitochondria,²⁰ and to exclusively express an excitation ratiometric Zn²⁺ sensor derived from carbonic anhydrase in mitochondria.²¹ Other genetically encoded Zn²⁺ sensors, such as the eCALWY family, have not yet been targeted to mitochondria.²² Our group previously constructed mito-ZifCY1 (renamed from mito-Cys₂His₂), a genetically encoded, ratiometric, Zn²⁺-specific sensor targeted to mitochondria, but its measurements are limited by its small dynamic range.¹⁴

In this study, we show that increasing the dynamic range of genetically encoded Zn²⁺ sensors improves their precision. We make novel comparisons of the [Zn²⁺]_mito of different cell types using improved sensors. We find [Zn²⁺]_mito is about 3 orders of magnitude lower than the cytosolic free Zn²⁺ concentration and varies considerably among different cell types.

Results and Discussion

Sensor Design and Validation

The Zn²⁺ sensors constructed in this study are variants of previously published cytosolic ZapCY1 and ZifCY1 sensors (Figure 1, panel a, Supplementary Figure 1, Tables 1 and 2), which respond specifically to Zn²⁺ over other biologically relevant divalent cations, including calcium, magnesium, iron, and copper.^14,23 The conformational change upon Zn²⁺ binding changes the Forster resonance energy transfer (FRET) efficiency, and thus the sensor’s fluorescence emission reports the proportion of bound sensor. We report the magnitude of FRET as the FRET ratio (R), which is the fluorescence intensity of the acceptor fluorescence protein (FP) divided by that of the donor FP when only the donor FP is excited. Estimation of free Zn²⁺ is possible when the sensor’s affinity for Zn²⁺ is known and an in vivo sensor calibration is performed. Importantly, these sensors cannot estimate total Zn²⁺.

Figure 1.

Design of genetically encoded mitochondrial Zn²⁺ sensors mito-ZapCY1 and mito-ZifCY1. a) These sensors undergo a conformational change upon binding Zn²⁺, which leads to a change in FRET. The Zn²⁺-binding domain used in the “Zap” and “Zif” sensors consists of the first two Zn²⁺ fingers of the Saccharomyces cerevisiae protein Zap1, and the Zn²⁺ finger from the mammalian protein Zif268, respectively. The mitochondria targeting sequence (MTS) is appended to the N-terminus of the sensor. b) mito-ZapCY1 colocalizes with MitoTracker Red in living HeLa cells; Pearson’s coefficient 0.938; scale bars represent 10 μm. c) The fractional saturation of the sensor mito-ZapCY1 was determined in HeLa cells by measuring R (at rest), R_min, and R_max. d) The in situ K_D′ of mito-ZapCY1 was determined in HeLa cells at pH 7.4 and 8.0. Each point represents the average (R−R_min) of at least 3 cells in a single experiment at a specific free [Zn²⁺].

In order to identify an improved mitochondrial Zn²⁺ sensor, several new sensors were constructed with different Zn²⁺ binding domains. Details of sensor design are included in Supplemental Methods. Specifically, we sought to identify sensors that have an appropriate affinity (K_D′) for Zn²⁺ allowing detection of both decreases and increases in Zn²⁺ concentration and with an improved dynamic range (defined as the maximum R (R_max) divided by the minimum R (R_min)). In most cases, the Zn²⁺-saturated sensor reports R_max and Zn²⁺-free sensor reports R_min, but some display inverted responses. ²² Measurement of R_max and R_min was achieved by performing in situ calibrations of single HeLa cells expressing each sensor (Figure 1, panel c). In a typical in situ calibration, R is measured in living cells treated with 150 μM of the Zn²⁺ chelator N, N, N′, N′-tetrakis-(2-pyridylmethyl)-ethylenediamine (TPEN), followed by 0.75 μM pyrithione (a Zn²⁺ ionophore) and 10 μM ZnCl₂. These calibrations are performed on single cells because the fractional saturation ((R−R_min)/(R_max−R_min)) varies less than R_min and R_max from cell to cell (Supplementary Table 4). We also hoped to identify partially saturated sensors, which are best for quantitative measurements. Supplementary Table 3 reports the dynamic range and fractional saturation of all the sensors tested and reveals mito-ZapCY1 as a robust sensor with a resting fractional saturation of 8.7±5.8% and the current largest dynamic range of 3.2, significantly better than the previously reported mito-ZifCY1 sensor,¹⁴ which has a dynamic range of 1.2 (Supplementary Figure 3).

Sensors were targeted to the mitochondrial matrix by appending an N-terminal mitochondrial targeting sequence, and they display excellent co-localization with MitoTracker Red in HeLa cells (Figure 1, panel b) and other cells tested (Supplementary Figure 2).

Estimates of mitochondrial free Zn²⁺ in HeLa cells based on mito-ZapCY1 dramatically differ from those based on mito-ZifCY1. Using mito-ZifCY1 (in vitro ZifCY1 K_D′ = 1.7±0.2 μM), we previously observed a fractional saturation of ~41%, identical to what we observe in the present study (Supplementary Table 3) leading to an estimate of [Zn²⁺]_mito of 680±140 nM in HeLa cells. In contrast, measurements using mito-ZapCY1 (in vitro ZapCY1 K_D′ = 2.53 pM, pH 7.4) estimate a [Zn²⁺]_mito of 0.22 pM in HeLa cells. We hypothesized that the poor dynamic range of mito-ZifCY1 (Supplementary Figure 3) results in unreliable estimates of [Zn²⁺]_mito and set out to improve its dynamic range. Circular permutation of FPs, which involves relocating the N-and C-termini to different loops, has been shown to impact the dynamic range of FRET sensors by changing the orientation of the two FPs (Figure 2, panel a).^24–26 We screened 5 variants of circularly permuted Venus (cpV), each permuted at a different location, within the framework of mito-ZifCY1. Figure 2 presents the dynamic range and fractional saturation of these sensors and Supplementary Figure 4 presents representative calibration traces. The 5 variants yielded significantly different fractional saturation and dynamic ranges, resulting in two sensors with increased dynamic ranges, which were named mito-ZifCV1.49 and mito-ZifCV1.173, where the “49” denotes that Venus was circularly permuted at amino acid position 49. The high dynamic ranges of mito-ZifCV1.49 and mito-ZifCV1.173, which display inverted responses to Zn²⁺, result from increased FRET in the unbound state and little change in FRET in the Zn²⁺-bound state. We found that the fractional saturation decreases as the dynamic range increases, even though the Zn²⁺ binding domains are identical. It has been demonstrated that incorporation of cpV can decrease the K_D′ 2–10-fold,²⁵ but even variation of the K_D′ cannot explain the strong correlation between the fractional saturation and dynamic range, suggesting that as hypothesized, measurements made using low dynamic range sensors can be inconsistent.

Figure 2.

Circular permutation of mito-ZifCY1 dramatically increases its dynamic range. a) The N-and C-termini of cpV FPs are relocated to 5 different loops of the original Venus FP at the amino acid positions 49, 157, 173, 195, and 229. b) The average R of the unbound and bound sensors expressed in at least 10 cells from 2 or more independent experiments, acquired using identical exposure times, are summarized. c) High dynamic range sensors report lower fractional saturation.

Converting the fractional saturation to [Zn²⁺]_mito requires estimation of the K_D′. The molecular environment of mitochondrial matrix differs from cytosol, and the pH and redox balance can change significantly under different conditions. The pH in the mitochondrial matrix is typically ~8.0 without perturbation, but it can vary from about 6.5 to 8.5 under different conditions.²⁷ Using Mito-pHRed,²⁷ we estimated the mitochondrial pH in HeLa cells to be ~8.0. Therefore, we calibrated mito-ZapCY1 in mitochondria of living cells by adding Zn²⁺ buffered at different free concentrations to Ca²⁺-, Mg²⁺-, and phosphate-free imaging media in the presence of 50 μg mL⁻¹ alamethicin. Alamethicin is an antimicrobial peptide that can permeabilize the mitochondrial inner membrane to small molecules only.²⁸ Using this method, we found mito-ZapCY1 has an in situ K_D′ of 1.6 pM and 17 pM at pH 8.0 and 7.4, respectively (Figure 1, panel d). These in situ affinities are comparable to the affinities measured in vitro at different pH (Supplementary Figure 5). We estimate [Zn²⁺]_mito to be 0.14 pM in HeLa cells based on this in situ titration of mito-ZapCY1 at pH 8.0.

Comparison of Mitochondrial to Cytosolic Zn²⁺

Next, we confirmed that two high-dynamic-range sensors with different Zn²⁺ binding domains yield consistent estimates of [Zn²⁺]_mito. As shown in Figure 3, the fractional saturations of mito-ZapCY1 and mito-ZifCV1.173 in HeLa cells were 16±10% and 5.8±3.1%, respectively while the corresponding sensors in the cytosol were almost completely saturated (90±7.3% and 95±1.1%, respectively). These data indicate that under resting conditions in HeLa cells, the free Zn²⁺ is buffered at concentrations about 3 orders of magnitude lower in mitochondria than in the cytosol.

Figure 3.

The fractional saturation of Zn²⁺ sensors is lower in the mitochondria than in the cytosol. Representative calibrations of ZapCY1 and ZifCV1.173 expressed in the mitochondrial matrix or the cytosol of HeLa cells are shown. Comparison of the fractional saturation of each sensor illustrates the difference in mitochondrial and cytosolic Zn²⁺ (* p < 0.0001, Student’s T-test). Horizontal black bars represent 1000 seconds. At least 3 cells were measured in each experiment.

If, in fact, [Zn²⁺]_mito is buffered at a lower concentration than free cytosolic Zn²⁺, the addition of a Zn²⁺ ionophore, in the absence of extracellular Zn²⁺, should dissipate the Zn²⁺ gradient between these two subcellular compartments. Indeed, treatment of several cell types expressing mito-ZapCY1 with 5 μM pyrithione caused a rapid increase in mitochondrial Zn²⁺, which was reversed by the addition of 150 μM TPEN (Supplementary Figure 6). Our interpretation of these results is that pyrithione moves Zn²⁺ into mitochondria from other subcellular compartments, such as the cytosol, which buffer free Zn²⁺ at higher concentrations.

Comparison of [Zn²⁺]_mito in different cell types

We used mito-ZapCY1 to quantitatively compare [Zn²⁺]_mito in different cell types and different environmental conditions. Mito-ZapCY1 was expressed in HeLa cells, MIN6 cells (a mouse insulinoma cell line), primary cortical neurons, and HC11 cells (a mouse mammary epithelial cell line). Resting R, R_min, and R_max were measured in individual cells to determine the fractional saturation of mito-ZapCY1 in each of the different cell types. Figure 4 presents a summary of mito-ZapCY1’s fractional saturation in different cell types and pseudocolor images of cells at rest, upon TPEN treatment, and upon addition of Zn²⁺/pyrithione. Significant differences among cell types were observed, including lower fractional saturation of mito-ZapCY1 in HeLa cells (8.7±5.8%) than in MIN6 cells (41±18%) or neurons (59±20%). We also compared the fractional saturation of mito-ZapCY1 in HC11 cells grown in basal media (−prolactin) to those in lactogenic media (+prolactin) and found that [Zn²⁺]_mito is significantly higher in non-lactogenic HC11 cells than in lactogenic HC11 cells (p=0.0022, Student’s T-test). This is particularly intriguing given that HC11 cells have been shown to undergo massive redistribution of intracellular Zn²⁺ pools and alterations in Zn²⁺ transporter expression upon lactogenic stimulation.^29,30 These results suggest that [Zn²⁺]_mito is regulated differently in several cell types and under different environmental conditions. While at this point we don’t know how cells maintain different levels of mitochondrial Zn²⁺, we speculate that they will likely exhibit different levels of Zn²⁺ transporters, buffers, and other regulatory proteins, that are necessary to ensure appropriate Zn²⁺ management for each specialized cell.

Figure 4.

Quantitative comparison of [Zn²⁺]_mito in different cell types. a) Pseudocolor images of the FRET ratio of mito-ZapCY1 in HeLa cells, MIN6 cells, and a primary cortical neuron illustrate changes in the FRET ratio in response to treatment with 150 μM TPEN or with 0.75 μM pyrithione and 10 μM ZnCl₂. Scale bars represent 10 μm. b) [Zn²⁺]_mito differs significantly among cell types. Each marker shows the fractional saturation of mito-ZapCY1 in a single cell (*p<0.0022, Student’s T-test; ***p<0.0001, ANOVA, Tukey’s HSD post-hoc test).

In summary, we constructed and validated three genetically encoded, high dynamic range mitochondrial Zn²⁺ sensors. Although sensors with low dynamic range are capable of detecting relative changes in Zn²⁺, high dynamic range sensors are necessary for making consistent and quantitative comparisons of Zn²⁺ between different cell types. Using mito-ZapCY1, we estimate [Zn²⁺]_mito to be 0.14 pM in HeLa cells. A recent study reports a similar [Zn²⁺]_mito of 0.2 pM measured with a different ratiometric Zn²⁺ biosensor in the PC12 rat pheochromocytoma line.²¹ We believe that these mitochondrial Zn²⁺ sensors can be used to address the complex interplay between Zn²⁺ homeostasis and mitochondrial function.

Although FRET-based genetically encoded sensors are capable of making quantitative measurements in intact cells, it is poorly understood if and how Zn²⁺ sensor expression changes the total Zn²⁺ concentration and the Zn²⁺-buffering capacity of the cell. The fractional saturation of mito-ZapCY1 does not change as the sensor expression increases (Supplementary Figure 7), suggesting that in contrast to small molecule sensors ^4,31, these sensors do not deplete the mitochondrial Zn²⁺ pool. The lack of perturbation of resting Zn²⁺ also suggests that, similar to the cytosol, mitochondrial Zn²⁺ is buffered and the sensor concentration is much lower than the concentration of the buffer.

A potential limitation of the Zn²⁺-finger binding domains of the mito-ZapCY1 and mito-ZifCV1 sensors is their sensitivity to oxidation and changes in pH. However, we observe little perturbation due to acidification by acetic acid or treatment with H₂O₂. We performed parallel experiments with the sensors mito-pHRed and mito-RoGFP2 and observed little change in pH or oxidation in the process of sensor calibration (Supplementary Figure 8).^27,32

In conclusion, we created improved mitochondrial Zn²⁺ sensors to measure the buffered set point of free Zn²⁺ with superior accuracy and precision. These new sensors complement and expand our fluorescent toolbox for studying Zn²⁺ and other ions in complex, biological systems.

Methods

Sensor Construction

Four repeats of the coding sequence for the first 29 amino acids of the human cytochrome c oxidase subunit 8a (mitochondrial precursor; accession number NP_004065) precede the coding sequence of each Zn²⁺ sensor in the mammalian expression vector pcDNA3.1, as previously described.¹⁴ Citrine FP in “CY” sensors was replaced with cpV FPs to make “CV” sensors using 5′ SacI and 3′ EcoRI restriction sites. ZapCY1 was expressed in E. coli and purified for use in in vitro titrations as previously described ²³. Please refer to Supplemental Methods for further details.

Cell culture

HeLa cells were maintained in high glucose Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS) and penicillin/streptomycin (pen/strep). MIN6 cells were cultured in D-MEM supplemented with 10% (v/v) FBS, L-glutamine, sodium pyruvate, β-mercaptoethanol, and penicillin/streptomycin. HC11 cells were maintained, as previously published, in non-lactogenic medium, which was replaced with lactogenic medium 24–48 prior to imaging.³⁰ All cells were transfected 48–72 hours before imaging with Mirus TransIT-LT or electroporated the Neon system (Life Technologies). Primary cortical neurons were obtained, cultured, and transfected as described in Supplemental Methods. Transfected cells were stained with MitoTracker Red (Life Technologies) for co-localization studies. During all experiments except in situ titrations, cells were imaged in phosphate-free HEPES-buffered Hanks Balanced Salt Solution (HHBSS), pH 7.4.

mito-ZapCY1 titrations

Zn²⁺ solutions used for the titration of mito-ZapCY1 were buffered using Zn²⁺ chelators (EGTA, EDTA, and HEDTA) using a previously published method²³ with the following modifications. Solutions for both in vitro and in situ titrations were adjusted to the specified pH, and the free [Zn²⁺] in each Zn²⁺/chelator buffered solution was calculated for different pH. One HeLa cell imaging experiment was performed at each Zn²⁺ concentration and the average R of ≥3 cells was used to calculate K_D′. Cells were imaged in phosphate-free HHBSS, pH 7.4 throughout Zn²⁺ chelation with 150 μM TPEN, then replaced with phosphate-, Ca²⁺-, and Mg²⁺-free HHBSS, pH 7.4 or pH 8.0 with 125 μM dithiothreitol (to prevent sensor oxidation) and Zn²⁺/chelator to buffer the free [Zn²⁺]. Cells and mitochondria were permeabilized using 50 μg ml⁻¹ alamethicin. All chemicals were purchased from Sigma.

Imaging

Image acquisition and analysis were performed as previously published.²³ Microscope filter combinations for FRET and CFP: 430/24 nm excitation filter, 455 nm dichroic mirror, 535/25 nm and 470/24 nm emission filters, respectively. Images were analyzed using ImageJ and Matlab software.

Statistical Analysis

Statistical analysis was performed using the Student’s T-test or ANOVA with Tukey’s HSD post-hoc test in the KaleidaGraph program.

ABBREVIATIONS

cpV

circularly permuted Venus

FP

fluorescent protein

FRET

Förster resonance energy transfer

K_D′

dissociation constant

Δψ_m

mitochondrial inner membrane potential

TPEN

N, N, N′, N′-tetrakis-(2-pyridylmethyl)-ethylenediamine

[Zn²⁺]_mito

the buffered Zn²⁺ concentration in the mitochondrial matrix