Native and engineered sensors for Ca²⁺ and Zn²⁺: lessons from calmodulin and MTF1

Abstract

Ca²⁺ and Zn²⁺ dynamics have been identified as important drivers of physiological processes. In order for these dynamics to encode function, the cell must have sensors that transduce changes in metal concentration to specific downstream actions. Here we compare and contrast the native metal sensors: calmodulin (CaM), the quintessential Ca²⁺ sensor and metal-responsive transcription factor 1 (MTF1), a candidate Zn²⁺ sensor. While CaM recognizes and modulates the activity of hundreds of proteins through allosteric interactions, MTF1 recognizes a single DNA motif that is distributed throughout the genome regulating the transcription of many target genes. We examine how the different inorganic chemistries of these two metal ions may shape these different mechanisms transducing metal ion concentration into changing physiologic activity. In addition to native metal sensors, scientists have engineered sensors to spy on the dynamic changes of metals in cells. The inorganic chemistry of the metals shapes the possibilities in the design strategies of engineered sensors. We examine how different strategies to tune the affinities of engineered sensors mirror the strategies nature developed to sense both Ca²⁺ and Zn²⁺ in cells.

Introduction

There are many parallels between calcium and zinc: they are abundant in biological systems where they exist as divalent cations, they are redox inactive, they bind to proteins where they serve as critical cofactors, and for most forms of life they are essential micronutrients. Living organisms concentrate and buffer these ions such that ion acquisition and distribution is tightly regulated. Despite exquisitely coordinated mechanisms to maintain tight homoeostatic regulation of ion concentrations, organisms use dynamic changes in the concentrations of both labile calcium (Ca²⁺) and zinc (Zn²⁺) to drive physiological processes [1,2]. Ca²⁺ transients are important for organismal and cellular processes ranging from fertilization and division to disease and apoptosis [3] and have been well characterized in cells using a variety of Ca²⁺ indicators [4,5]. While recognition of Zn²⁺ transients in cells is only beginning to emerge, studies point to Zn²⁺ fluxes as playing a regulatory or signalling role in cells. Such Zn²⁺ dynamics include ‘zinc sparks’ upon mammalian egg fertilization and ‘zinc waves’ in immune cells [6,7].

A noted feature of Ca²⁺ transients and Zn²⁺dynamics is that they are organized into distinct patterns in space and time [1,2]. In order for these dynamic patterns to encode information, cells must be able to sense the changes in metal concentration and translate that change into a specific downstream action, with different patterns encoding different functions. For Ca²⁺, many native sensing proteins have been identified, with calmodulin (CaM) as the quintessential Ca²⁺ effector [3]. The identity of the proteins that transduce mammalian Zn²⁺ dynamics is less clear, although proteins that multimerize and become active upon Zn²⁺ binding have been identified as possible Zn²⁺ sensors [8,9].

This essay will compare and contrast CaM, the textbook calcium-sensing protein, with metal-responsive transcription factor 1 (MTF1), a candidate protein for cellular zinc sensing and signal transduction in mammalian cells [10,11]. As noted above, there are many similarities between these two ions. However, there are also notable differences in the chemistry and biology of these two important ions that hint at orthogonal signalling roles in biological organisms. In this essay, we focus on proteins that sense and transduce changes in Ca²⁺ or Zn²⁺, highlighting the fundamental inorganic and protein chemistry features of these sensors that suggest these metal sensors operate by divergent mechanisms.

In addition to native metal sensors that decode natural dynamics in Ca²⁺ or Zn²⁺, scientists have sought to engineer protein-based and small molecule metal sensors to spy on these changes [4,12]. Engineered sensors translate changes in metal concentration in live cells to changes in a fluorescence signal that can be detected by microscopy. A challenge in sensor engineering is to tune the affinity of the sensor, while maintaining the specificity, so that the fluorescence changes report only on the metal of interest, ignoring the multitude of confounding and competing factors that could be present in the cellular environment. Ideally, the binding constants of sensors are tuned such that the sensor is ~50% saturated in the resting cell in the subcellular location of interest [13,14]. Distinct approaches have been used to modify the affinities of protein-based sensors for Ca²⁺ than have been employed for Zn²⁺ sensors. For Ca²⁺ sensors, a common approach for tuning the apparent binding constant has been to manipulate the interaction of CaM (or an analogous Ca²⁺-sensing protein) with a partner binding protein [15–17]. Alternatively, the most widely used approach for tuning the apparent binding constant of Zn²⁺ sensors is to alter the metal coordination site [14,18]. Although these approaches are distinct, they lead to robust, selective metal sensors to examine the dynamics and distribution of Ca²⁺ and Zn²⁺ in cells.

While native and engineered sensors have two very different purposes – to inform the cell of dynamic changes and to let scientist glimpse the inner workings of cell biology – we propose that lessons learned from studying one might inform our study of the other. CaM modulates the cell’s response to Ca²⁺ through its structural plasticity that allows it to bind and regulate over 300 partner proteins [19]. On the other hand, binding of Zn²⁺ to six different sites induces MTF1 to recognize a single DNA motif repeated throughout the genome modifying the transcription of target genes. The engineering of sensors has mirrored these native proteins, leveraging both the allosteric flexibility of sequences near Ca²⁺ sites and the variable coordination preference of Zn²⁺ to broaden our insight into the biology of these metals.

Native metal sensors

CaM

Ca²⁺ exists in the cell in two populations: labile Ca²⁺ that is not tightly bound to proteins and protein-bound Ca²⁺. Resting cells maintain a gradient of labile Ca²⁺ from 2 mM in the extracellular space to much lower concentrations in the cell, ranging from 100 nM in the cytosol to hundreds of μM in the endoplasmic reticulum [20]. Upon stimulation of Ca²⁺ signalling, the cytosolic Ca²⁺ concentration spikes [2,10,15,20]. Labile Ca²⁺ concentrations are regulated by Ca²⁺ channels, buffering proteins that act as sinks for excess Ca²⁺, and Ca²⁺ sensors that serve as effectors by binding downstream proteins upon changes in Ca²⁺ status [10,21]. Sensing proteins coordinate Ca²⁺ with oxygen ligands, as would be predicted by hard-soft acid-base theory [22], contributed by aspartate or glutamate amino acids. The most common Ca²⁺-binding motif in proteins is called an EF hand. Based on genetic data, 2540 or approximately 70% of the known Ca²⁺-binding proteins in animals contain an EF hand motif [23]. This 30-residue helix–loop–helix structure binds Ca²⁺ through six or seven oxygen atoms from six coordinating amino acids in a pentagonal bipyramidal structure [24]. The affinity of these structures for Ca²⁺ can be tuned over a 100000-fold range through the identity and conformation of amino acids in the EF hand and side chain packing through the core of the protein [10]. This ability to tune affinity makes EF hands versatile Ca²⁺ sensors over a wide range of concentrations and may explain why EF hands are found in such diverse proteins including troponin C, a Ca²⁺ sensor in muscle cells; calcineurin, a phosphatase essential for T-cell activation; and the S100 proteins, regulatory proteins found in a variety of tissues [25].

In CaM, EF hand motifs are essential for transducing Ca²⁺ binding into global protein conformation changes that lead to interactions with partner proteins. CaM senses cytosolic Ca²⁺ through four EF hands paired in two lobes separated by a flexible linker (Figure 1A) [10,26]. When CaM is in the apo form, the helices of the EF hands are antiparallel and assume a closed geometry. As Ca²⁺ binds the EF hands, core amino acids throughout CaM rearrange to a more open conformation that exposes hydrophobic interfaces which bind other proteins [27]. It is through this allosteric interaction, and associations with the flexible linker, that CaM modulates the activity of other proteins. Interestingly, this allostery is not unidirectional; the interaction of proteins with the hydrophobic patches on both holo- and apo-CaM can affect the affinity and cooperativity of the EF hands for Ca²⁺, making CaM responsive over a wide range of Ca²⁺ concentrations from 10⁻¹² to 10⁻⁶ M [26,28,29].

Figure 1. Structures of CaM and MTF1.

(A) Crystal structure of CaM with coordinating ligands highlighted (PDB entry 4BW8). (B) Crystal structure of ZIF-268 as an example of a αββ Zn²⁺ finger fold. MTF-1 encodes six similar Zn²⁺ fingers and three transactivation domains as shown in the schematic below the structure (PDB entry 1ZAA).

A remarkable characteristic of CaM is its ability to interact in a variety of conformations with over 300 proteins and to recognize those proteins specifically. The proteins that are binding partners for CaM share little to no homology, and occupy many categories of cellular machinery from G-coupled receptors to ion channels to kinases [21]. Shukla and co-workers employed molecular dynamics (MD) to probe the molecular basis of this ability to recognize diverse partners. A proposed hypothesis for how CaM binds a diverse set of proteins is that the binding partners induce the fit on unstructured CaM. On the contrary, the folding landscapes of both holo-C-CaM and apo-C-CaM include well-populated conformations that provide the intermolecular interfaces described in crystal structures of CaM with binding partners. This evidence implies that it is not the binding partners that prescribe the interaction with CaM, but intramolecular hydrophobic interactions in the CaM sequence that initiate the interaction with binding partners [21].

Upon binding Ca²⁺, CaM directly modifies the activity of its binding partners. This is accomplished through a number of mechanisms including displacement of autoinhibitory domains, active site remodelling, dimerization and/or increased autophosphorylation [30]. Each of these mechanisms occurs with different kinetics and different thermodynamic stabilities, enabling CaM to orchestrate complex downstream effects in response to Ca²⁺ dynamics. For example, apo-CaM is constitutively bound to small-conductance Ca²⁺-activated K⁺ channels (SK channels), and when Ca²⁺ binds to the N-lobe of CaM, conformational changes in CaM cause both dimerization and opening of the SK channel subunits to allow K⁺ flow across the membrane. Ca²⁺ release upon decreased Ca²⁺ concentration closes the channel [31]. In contrast, binding of Ca²⁺-CaM to CaM-dependent kinase II (CaMKII) leads to an increase in autophosphorylation of CamKII’s neighbouring subunits, increasing the activation state of the kinase for its downstream substrates. Because CamKII must be dephosphorylated, its increased activity continues after the initial Ca²⁺ flux, perhaps creating stability between Ca²⁺ pulses [32]. Complex allosteric interactions between CaM Ca²⁺-binding sites and hydrophobic intramolecular interfaces confer plasticity on CaM’s structure, and this plasticity is instrumental in allowing CaM to regulate hundreds of proteins, thereby modulating multiple diverse pathways in cells. One of the only established Zn²⁺ sensors, MTF1, responds to Zn²⁺ dynamics in a fundamentally different way.

MTF1

Like Ca²⁺, Zn²⁺ levels are regulated by a complex network of transporters, and buffered by proteins and other ligands. Labile Zn²⁺ in the cytosol of mammalian cells is maintained at a concentration in the hundreds of pM, even though total Zn²⁺ levels exceed hundreds of μM [33]. There is not strong evidence that organelles store and release excess Zn²⁺ [14,34,35], although in certain cell types, Zn²⁺ is packaged and concentrated into vesicles [36,37]. An intriguing feature of biological Zn²⁺ sensing is the sheer number of zinc-binding proteins present in an organism, and the diversity in binding sites used to coordinate Zn²⁺. Critical analysis of the Structural Classification of Proteins (SCOP) database identified increasing abundance of zinc-binding structural domains from archea to bacteria to eukaryotes [38]. Analysis of 23 archea, 233 bacteria and 57 eukaryote species revealed that while abundance typically scales with genome size, eukaryotic zinc-binding proteins exceed the predicted power law, and hence eukaryotes devote a greater percentage of their genome to zinc-binding proteins [38]. Further bioinformatics studies estimate that up to 10% of the proteins encoded by the human genome are predicted to be zinc-binding proteins [39]. In biological systems, proteins coordinate Zn²⁺ through sulfur, nitrogen or oxygen moieties and a variety of coordination spheres from four- to six-coordinate [40]. Protein coordination sites are thought to derive their selectivity for Zn²⁺ over other cations from the lack of ligand field stabilization cost of desolvating Zn²⁺ [41]. Although many Zn²⁺-binding sites coordinate Zn²⁺ with high affinity (dissociation constants of a few pM or lower), examples of lower affinity or kinetically exchangeable sites have been reported [42].

While the number and variety of Zn²⁺-binding sites in the proteome is fascinating, the redundancy of Zn²⁺ proteins has made it difficult to clearly define how cells sense and manage Zn²⁺. Several outstanding questions remain: how and where do cells load so many diverse binding sites with Zn²⁺ specifically? How does the cell organize the expression and regulation of the many Zn²⁺ proteins? As new examples of Zn²⁺ dynamics are discovered, what proteins sense the dramatic changes in Zn²⁺ flux? A few characteristics of MTF1 make it an attractive candidate to sense and regulate a response to Zn²⁺: (i) It binds Zn²⁺ with a low enough affinity to be partially unsaturated in resting cytosolic concentrations, (ii) once fully Zn²⁺ bound it translocates from the cytosol to the nucleus and (iii) in the nucleus it binds a specific DNA motif thereby regulating the expression of Zn^2+-binding proteins and a Zn²⁺ export channel. Here we will present MTF1 as an example of a Zn²⁺ sensor, but hypothesize that other Zn²⁺ sensors may exist.

Human MTF1 contains six zinc-finger motifs, three transactivation domains and a conserved cysteine-rich cluster (Figure 1B) [11]. Each zinc finger binds one Zn²⁺ in a pseudo-tetrahedral geometry through two cysteine residues and two histidine residues. Scores of structural and biophysical studies have revealed that two to three of the zinc fingers bind Zn²⁺ with relatively high affinity and three to four of the fingers bind Zn²⁺ with low affinity [43–45]. In metal-binding studies of the complete six finger domain the K_d (Zn²⁺) was estimated to be ~30 pM, and spectroscopic studies of the individual fingers binding to Co²⁺ indicate individual finger dissociation constants vary 25-fold [45]. This range of in vitro affinities in the picomolar regime supports the hypothesis that lower affinity fingers may be responsible for the Zn^2+-sensing capabilities of MTF1 in the cytosol, while the high affinity fingers constitutively bind Zn²⁺ [11].

The function of MTF1 is to bind to and modulate the transcription of DNA in response to changes in cellular Zn²⁺. When MTF1 is replete with metal it translocates from the cytosol to the nucleus where it recognizes a DNA motif called the metal-response element (MRE) that is found in the promoter region of MTF1 target genes [11]. Zn²⁺ association with the four N-terminal fingers is necessary for tight binding of MTF1 to the MRE, while the two C-terminal fingers have been implicated in providing specificity to the protein–DNA interface [44]. Upon DNA binding, the three transactivation domains recruit transcription machinery to the promoter region to regulate transcription of downstream target genes [43].

Until recently the genes identified to be under the control of MTF1 in response to Zn²⁺ were the genes for metallothioneins, proteins that buffer Zn²⁺ in the cytosol and Znt1, a Zn²⁺ exporter [46]. To identify more genes under the control of MTF1, Hardyman and co-workers examined the differential expression of genes in normal and excess Zn²⁺ in wild-type Caco-2 and MTF1 knockdown Caco-2 cells. They found that, as expected, in the MTF1 knockdown, cells expression of the previously identified MTF1 target genes was no longer sensitive to Zn²⁺ increases. However, they also discovered that expression of a number of genes was modulated by increasing Zn²⁺ in the MTF1-depleted cells as compared with wild type. As one example, the expression of genes encoding zinc uptake transporters was decreased upon exposure to increased Zn²⁺ concentration in the MTF1 knockdown cells. These data led to the hypothesis that MTF1 controls a hierarchy of Zn²⁺ responsive proteins. When MTF1 is available, it responds to Zn²⁺ increase by amplifying transcription of Zn²⁺ buffering and export proteins that lower cellular Zn²⁺ concentrations. In the absence of this safeguard, increases in Zn²⁺ were dramatic enough to uncover the expression of other Zn^2+-sensitive genes that may be under the control of unknown transcription factors [46].

These data suggest that dynamic changes in Zn²⁺ impact the proteome of the cell through transcription of a variety of genes. While this mechanism is effective at eliciting a cellular response, it is intriguing to imagine other scenarios for Zn²⁺ communication based on the unique coordination chemistry of Zn²⁺. For instance, there is evidence for Zn²⁺ binding to be kinetically labile, and such labile sites could be exploited to sense fluxes of Zn²⁺ [42]. Alternatively, Zn²⁺ can be coordinated at the interface of proteins, modulating their activity [9,47,48]. Could this be an additional mechanism for sensing Zn²⁺ concentration changes? These scenarios stand in contrast with what is known about Ca²⁺ sensing and coordination. Coordination sites that are specific to Ca²⁺ function amid a sea of Mg²⁺, which is present at much higher concentrations than Ca²⁺. This pressure requires that coordination sites sensitive to physiologic transients of Ca²⁺ be carefully tuned to coordinate Ca²⁺ [49]. Because of this competition, perhaps nature accomplishes sensing and signal transduction through the plasticity of protein conformations of a single protein rather than a library of proteins decode the cell’s response to Ca²⁺.

Engineered sensors

In order for scientists to visualize and measure dynamic changes in metals in cells, artificial sensors have been engineered to quantify metal concentrations in live cells. While both protein based and small molecule sensors have been developed, protein-based sensors allow useful comparisons with native metal sensors. One class of protein-based sensors is the family of genetically encoded sensors based on FRET. These sensors are fusions of a donor fluorescent protein (FP), a metal-sensing domain and an acceptor FP. When the sensor is metal bound, it shifts conformations leading to a change in FRET between the two FPs. A number of FRET-based sensors have been engineered for both Ca²⁺ and Zn²⁺. Scientists often seek to engineer the binding affinity such that the sensor is partially occupied by metal in the environment of interest while maintaining a large change in the fluorescence upon binding to confer a high dynamic range in the cellular milieu. In order to solve this design challenge, FRET sensors for Ca²⁺ have leveraged the allostery of CaM, while Zn²⁺ sensors have exploited the dramatic conformational restructuring and flexible coordination of zinc fingers.

A number of powerful Ca²⁺ indicators have been engineered over the past decades [4,5], but here we will pull one case study to highlight a design strategy that piggybacks off a native sensor characteristic. To create a FRET-based sensor for Ca²⁺, CaM and a fragment of the CaM-binding partner, smooth muscle myosin light chain kinase (smMLCK), were fused between a donor protein, CFP and an acceptor protein, YFP. In the resulting sensor, called cameleon, Ca²⁺ binding causes association of CaM with the smMLCK peptide altering the FRET signal between the two FPs. In the original sensors, the affinity for Ca²⁺ was tuned by mutating the EF hands. Two weaknesses of this design were that the sensor was susceptible to binding by native CaM, and the apparent dissociation constant for Ca²⁺ was weak (60 μM). To overcome these issues the interface between CaM and the peptide was engineered to include more hydrophobic bumps and holes. This re-engineering led to a series of sensors that were unperturbed by the native CaM with a range of apparent dissociation constants from ~0.1 to 49 μM [15]. This new generation of cameleons was not susceptible to binding to native CaM, making them robust sensors for application in live cells.

As with Ca²⁺ sensors a common Zn²⁺ FRET sensor design is to fuse a donor FP to a metal-sensing domain followed by acceptor FP [14,18,50]. A major difference in the design of the sensors as compared with cameleons is the ability to tune Zn²⁺ affinity altering the identity of the amino acids that coordinate Zn²⁺ without losing specificity for labile Zn²⁺ [14,18]. For example, mutating a native cysteine to a histidine in each of the two Zn²⁺-binding sites of the Zap family of sensors alters the apparent dissociation constant from 2 to 800 pM [14]. This approach is possible because Zn²⁺ is fairly amenable to different coordination geometries and ligating residues.

By examining native sensors for metals new approaches can be harnessed for engineering more robust sensors for measuring dynamics metals in cells. Here we have compared approaches to tune the affinity of engineered sensors for metals that make the use of native characteristics of both protein and metals. Ca²⁺-binding proteins are more permissive to alteration at allosteric sites, while Zn²⁺-binding sites can remain selective with changes to the identity of the coordinating moieties. As scientists continue to examine metal transients, particularly in the developing field of Zn²⁺ dynamics, it is equally important and challenging to designate the native sensors and targets of those signals. Because there are still many open questions about Zn²⁺ signalling it is useful to study Zn²⁺ through the lens of what is known about Ca²⁺. As we uncover similar patterns and statuses of Ca²⁺ and Zn²⁺ cations, it will be essential to remember the fundamental differences in the inorganic chemistry between the two metals that may shape their role in cell biology.

Summary.

• Ca²⁺ and Zn²⁺ dynamics are important drivers of physiological processes. In order for these dynamics to regulate cell physiology and function, cells must sense the changes.

• CaM and MTF1 are compared and contrasted as natural sensors of Ca²⁺ and Zn²⁺ respectively.

• Engineered protein sensors spy on these dynamic changes in metal concentrations and engineered sensor design borrows features from native sensors.

• The inorganic chemistry of metal ions shapes their cell biology and the design of engineered sensors.

Abbreviations

CaM

calmodulin

CaMKII

CaM-dependent kinase II

FP

fluorescent protein

MRE

metal-response element

MTF1

metal-responsive transcription factor 1

SK

small-conductance Ca²⁺-activated K⁺ channels

smMLCK

smooth muscle myosin light chain kinase