Introduction

Cytotoxic T lymphocytes and natural killer cells recognise and destroy virus-infected and virus-transformed cancerous cells through the exquisitely regulated secretory granule exocytosis pathway [1]. Central to this mechanism is the pore-forming effector protein, perforin (Prf1), which is stored in cytotoxic secretory granules and co-secreted with the pro-apoptotic serine proteases, granzymes, into the immune synapse formed between a cytotoxic lymphocyte and target cell [2]. The high concentration of extracellular Ca2+ (>1 mM) promotes binding of the Prf1 through its C2 domain to the target cell [3] via a single localised conformational change, which repositions 4 key hydrophobic residues and promotes anchoring of Prf1 to the plasma membrane [4]. Once bound, the neutral pH of the synapse facilitates Prf1 oligomerisation [5] into ring and arc-shaped pores [6, 7], which are required for the diffusion of granzymes into the cytoplasm of the target cell to initiate apoptosis [8]. Prf1 therefore sits at the apex of the signalling cascade that triggers apoptotic target cell death and immune defence by cytotoxic lymphocytes.

The loss of Prf1 function due to bi-allelic mutations in the PRF1 gene leads to a severe and often fatal form of immune dysregulation, familial haemophagocytic lymphohistiocytosis (FHL) [9]. If at least one of the affected alleles harbours a missense, hypomorphic mutation in PRF1, FHL may be delayed until adolescence or early adulthood, and there is also a high (up to 50%) risk of a haematological malignancy [10, 11]. A significant proportion of these missense mutations lead to Prf1 misfolding and retention in the endoplasmic reticulum (ER) [12]. Failure to traffic to the secretory granules results in failure of Prf1 secretion into the immune synapse and the loss of cytotoxic lymphocyte function. Similar to other protein misfolding diseases [13], Prf1 may partly recover its folding at permissive temperatures (for example, in vitro culture of cytotoxic lymphocytes at 30 °C), with restoration of protein trafficking and cytotoxicity [12]. However, recovery of function is typically incomplete and less likely with mutations associated with poor clinical outcomes. Over 30 of the reported pathogenic missense mutations in PRF1 map to the C2 domain, consistent with an essential role of that domain in Prf1 function; these mutations often result in Prf1 misfolding.

The C2 domain of Prf1 can coordinate up to five Ca2+ ions, including two high-affinity sites II and V [3]. Binding of Ca2+ to these two sites alone is not sufficient for membrane binding. In contrast, sites, I, III and IV bind Ca2+ at low affinity and are responsible for a conformational rearrangement of four conserved hydrophobic residues that are essential for Prf1 binding to the phospholipid bilayer [3]. The ER maintains an ionic Ca2+ concentration sufficiently high (500–700 µM) for Ca2+ binding and has neutral pH, and is the only intracellular compartment with an environment that may support Prf1 lysis. Indeed, we recently discovered that rapid export of Prf1 from the ER [14] and N-glycosylation close to the carboxy terminus [15] are essential for protecting cytotoxic lymphocytes from their own endogenous Prf1. As evidence, we showed that non-conservative mutations in a cryptic trafficking motif at the extreme C terminus of Prf1 caused its retention in the ER and death of the killer cell; however, introducing an additional mutation within the C2 domain to prevent Ca2+ binding restored cell viability without reversing the defect in trafficking. These observations clearly indicated that Ca2+ binds to the Prf1 C2 domain in the ER, but this does not compromise the viability or the effector function of the host cells. Intriguingly, it was recently found that physiological Ca2+ concentrations could stabilise Prf1 through the C2 domain [8], suggesting a role for Ca2+/C2 domain interaction beyond the regulation of Prf1 binding to the target cell membrane following exocytosis.

In this study, we discovered that Ca2+-dependent stabilisation of Prf1 through the C2 domain in the ER is required for its folding and the delivery to the secretory granules. Unexpectedly, we also identified a novel Ca2+ stabilising mutation in the C2 domain that improved the kinetics of Prf1 pore formation and also enhanced its cytotoxicity to be greater than wild-type (WT) Prf1; remarkably, it was also able to compensate for the folding and trafficking defect of a common disease-causing mutation, A90V.

Results

Ca2+ binding to the C2 domain of Prf1 in the ER is required for its folding and trafficking

We recently demonstrated that Prf1 is a thermodynamically labile protein, which acquired a more stable conformation at physiological Ca2+ concentrations; this effect of Ca2+ was abrogated by mutations in the C2 domain [4, 8]. Given the high concentration of Ca2+ in the ER and the demonstrated ability of Prf1 to bind Ca2+ there [14], we hypothesised that Ca2+ binding to the C2 domain may have an important, previously unappreciated, role in Prf1 folding and export from the ER.

First, we substituted Ca2+-binding aspartate residues, D429 and D483, each of which is essential for Ca2+ binding and Prf1 function [6, 16] and located within distinct Ca2+-binding regions CBR-1 and CBR2, respectively, with alanine (D429A and D483A) (Supplementary Figure 1) [6, 16]). These mutations completely abrogated Prf1 membrane binding and pore forming activity [16]. As Prf1 traffics from the ER through the Golgi compartment, N-linked, high mannose glycans attached at residues N204, N375 and N548 progressively undergo complex glycosylation. Accordingly, Prf1 that is retained in the ER lacks complex glycans and remains sensitive to endoglycosidase H (EndoH) [14]. We therefore assessed the EndoH resistance of WT and mutated Prf1 to determine their capacity to exit the ER.

Even though the expression levels of D429A and D483A were no different from WT, each demonstrated greater EndoH sensitivity than WT Prf1 (Fig. 1a), indicating impaired export from the ER. In addition, simultaneously introducing both Ca2+ binding site mutations (D429A and D483A) did not further increase EndoH sensitivity, suggesting that failure to bind Ca2+ was sufficient to impair the export of the protein (Fig. 1a). To further test the role of Ca2+ binding in conferring exit from the ER, we mutated D491 (D491A), a Ca2+-binding residue located in CBR3. Previous studies have revealed that a mutation at this position does not abrogate Ca2+-dependent membrane binding and pore formation at physiological concentrations of Ca2+ [16]. In contrast to D429A and D483A, D491A remained similarly resistant to EndoH as WT Prf1 (Fig. 1a).

Fig. 1
figure 1

Ca2+-dependent stabilisation of the C2 domain is necessary for efficient Prf1 folding and export from the ER. a Western immunoblot demonstrates the relative Prf1 expression in each of the transiently transfected RBL cell populations (WT, D429A, D483A, D491A) in the presence of EndoH (+) or buffer only control (−). EndoH-sensitive bands are highlighted by (*). The bar graph shows quantitative densitometry of the EndoH-treated (+) protein lysates; mutation of essential Ca2+ binding residues causes a significant increase in % EndoH-sensitive Prf1. EndoH sensitivity was calculated using the % EndoH cleaved protein versus the total amount of protein (cleaved and uncleaved) as detected by the PI-8 antibody. Shown is mean ± SEM of at least three independent experiments and statistics were determined using ANOVA and Tukey’s post-hoc analysis, *p < 0.05. b Western immunoblot using the P1-8 antibody demonstrates the relative Prf1 expression in RBL cells transiently transfected with WT and the W427A/Y430A/Y486A/W488A quadruple mutant (×4) in the presence of EndoH (EH), PNGaseF (PF) or buffer only control (Con); the EndoH sensitivity of the ×4 Prf1 was indistinguishable from WT. c Western immunoblot using the P1-8 antibody demonstrates the relative Prf1 expression in RBL cells transiently transfected with WT and the A90V mutant in the presence of EndoH (+) or buffer only control (−). EndoH-sensitive bands are highlighted by (*). The bar graph shows quantitative densitometry of the EndoH-treated (+) protein lysates; the mildly misfolded A90V Prf1 mutant caused a significant increase in % EndoH-sensitive Prf1. EndoH sensitivity of WT and A90V Prf1 was calculated as described in a. Shown is mean ± SEM of n = 16 (WT) and n = 7 (A90V) independent experiments and statistics were obtained with an unpaired two-sample for means t test, *p < 0.05. d Western immunoblot using the P1-8 antibody demonstrates the relative Prf1 expression in each of the transiently transfected RBL cell populations (WT, A90V/D429A, A90V/D483A, A90V/D491A) in the presence of EndoH (+) or buffer only control (−). EndoH-sensitive bands are highlighted by (*). Shown is mean ± SEM of at least three independent experiments. e–g Bar graphs showing quantitative densitometry of the EndoH-treated (+) protein lysates from representative westerns in c and d. To determine whether the double mutation of A90V and Ca2+ binding residues (A90V/D429A, A90V/D483A, A90V/D491A) had a synergistic effect on Prf1 stability and trafficking from the ER above normal background levels, the % EndoH-sensitive WT Prf1 was subtracted from the % EndoH sensitivity of Prf1 mutants. Dotted bars show predicted additive effect of individual mutants. Each value represents ± SEM for at least three independent experiments. The statistical significance was determined using ANOVA and Tukey’s post-hoc analysis, *p < 0.05

Full size image

Previously, we identified four hydrophobic residues (W427, Y430, Y486 and W488) that are key for interaction of the membrane proximal part of the C2 domain with the lipid membrane [4]. Mutation of these residues does not impact on Ca2+ binding [3, 4]. Accordingly, mutation of these four residues (W427A/Y430A/Y486A/W488A) did not impair mutant Prf1 export from the ER (Fig. 1b). Taken together, these data suggest that it is the ability of the Prf1 C2 domain to bind Ca2+, rather than the interaction with the ER membranes per se, which influenced Prf1 export.

In humans, hypomorphic PRF1 missense mutations that destabilise folding but retain partial function are commonly associated with haematological cancer in childhood or adolescence and/or atypical or delayed presentation with FHL [12, 17]. These mutants are invariably retained in the ER and poorly exported to the secretory granules [12]. By far, the most common mutation of this type is the A91V variant (mouse homologue A90V; Fig. 1c); this allele is carried in the heterozygous state by 8–9% of Caucasians. A91V homozygocity is strongly associated with leukaemia/lymphoma, systemic inflammatory disorders and atypical FHL [2]. As with human A91V [18,19,20], mouse A90V was partially misfolded and had a mild trafficking defect [20, 21]. To determine whether Ca2+ binding to Prf1 in the ER contributes to the stability of the A90V mutant, we combined A90V with mutations in critical Ca2+-binding residues in the C2 domain described above, A90V/D429A, A90V/D483A and A90V/D491A. We found that the loss of Ca2+ binding (A90V/D429A or A90V/D483A) markedly affected Prf1 stability and trafficking from the ER (Fig. 1d). Indeed, the EndoH sensitivity of the double mutants exceeded the sum of A90V and Ca2+ binding mutants alone (A90V + D429A and A90V + D483A; Fig. 1e, f, dotted lines). Protein expression levels were also significantly reduced in comparison to WT Prf1. By contrast, mutation of the non-essential Ca2+-binding residue D491A (A90V/D491A), did not further increase the EndoH sensitivity of A90V (Fig. 1g), indicating that the misfolding defect of A90V/C2 domain double mutants was directly associated with loss of Ca2+ binding at the C2 domain. Taken together, our results demonstrated that Ca2+ binding to the Prf1 C2 domain in the ER is required for stable folding and subsequent export of Prf1 from the ER.

The folding and trafficking of A90V are rescued by Ca2+-dependent stabilisation through the Prf1 C2 domain

The C2 domain of Prf1 differs from most others in that the overall affinity for Ca2+ is low. Unlike other C2 domains that avidly bind Ca2+ at low µM or even sub-µM concentrations, Prf1 becomes functional only at neutral pH and when free Ca2+ levels exceed 250 µM, a concentration found only in extracellular milieu (for instance, in the immunological synapse) or in the ER of intact cells.

Previously, we have used NMR-based approaches, together with mutagenesis and crystallography, to demonstrate that the canonical Ca2+-binding site III has the weakest affinity of all five Ca2+-binding sites and is only filled at high concentrations of Ca2+ [3]. These findings are consistent with our data revealing that CBR-1 must undergo significant conformational change to bring the key residues D429 and D435 into position to bind Ca2+. The rearrangement of CBR-1 further repositions the four essential hydrophobic residues such that W427 and Y430 are in close proximity to Y486 and W488. Altogether, these four hydrophobic residues are thus positioned to interact with the plasma membrane [4, 6].

We have previously suggested that the conformational change within CBR-1 is likely driven through electrostatic attraction to Ca2+ ions already bound at the high-affinity sites I and II (as D429 also interacts with the site I and II and D435 interacts with site II). On the basis of these data, we reasoned that introduction of additional negatively charged residues into CBR-1 may facilitate conformational changes within CBR-1 through provision of additional short or longer range electrostatic interactions with bound Ca2+ ions in site I and II. Accordingly, and using X-ray crystal structures that we had previously solved as a guide [3, 6], we identified T431 as one potential site to introduce a negatively charged residue into CBR-1 and generated the mutation T431D.

Previously, we showed that the thermal stability of Prf1 is markedly increased in the presence of Ca2+ [8]. We therefore compared the thermal stability of the T431D variant to that of WT Prf1 and found that the T431D substitution (Fig. 2a) caused a marked increase in melting temperature at physiologically relevant concentrations of Ca2+. Similar to WT, T431D-Prf1 stabilisation was Ca2+-dependent, as simultaneously mutating the key Ca2+-binding residue D429 (D429A/T431D double mutant) prevented this change (Fig. 2a). Taken together, these data suggested that as predicted, the mutation was in some way facilitating Ca2+ binding.

Fig. 2
figure 2

Prf1 protein is stabilised by a point mutation, T431D, in the C2 domain in the presence of Ca2+. a The thermal melting temperature of WT and mutant Prf1 in the presence of increasing Ca2+ concentrations: (left, WT and T431D-Prf1) T431D-Prf1 was significantly more stable than WT Prf1 in the presence of >1 mM Ca2+; (centre, WT and D429A/T431D-Prf1) inhibition of Ca2+ binding to the Prf1 C2 domain by D429A prevented any increase in T431D-Prf1 thermal stability; (right, WT and A90V/T431D-Prf1) the T431D mutation onto the backbone of A90V Prf1 resulted in greater thermal stability than WT Prf1. Each value represents means ± SEM for three (left and centre) or five (right) experiments, and statistics were obtained with an unpaired two-sample for means t test, *p < 0.05. b There was no significant difference in the lytic activity of recombinant WT and T431D-Prf1 (left) and A90V/T431D-Prf1 (right), as assessed using SRBCs. c Western immunoblot using the P1-8 antibody demonstrates the relative Prf1 expression in each of the transiently transfected RBL cell populations (WT, T431D, A90V and A90V/T431D) in the presence of EndoH (+) or buffer only control (−). EndoH sensitive band is highlighted by (*). The bar graph shows densitometry analysis of the EndoH-treated (+) protein lysates; the T431D mutation restored the % EndoH sensitivity of A90V to WT Prf1 levels (A90V/T431D). EndoH sensitivity of Prf1 was calculated as described in Fig. 1. WT and A90V values (dotted lines) are reproduced from Fig. 1c (WT, n = 16, A90V, n = 7); n = 3 for T431D and A90V/T431D. Statistical significance was determined using ANOVA and Tukey’s post-hoc analysis, *p < 0.05. d Western immunoblot using the P1-8 antibody demonstrates the relative Prf1 expression in each of the transiently transfected RBL cell populations (WT, D429A/T431D, A90V/D429A/T431D) in the presence of EndoH (+) or buffer only control (−). The bar graph shows densitometry of the EndoH-treated (+) protein lysates; T431D-Prf1 did not restore the % EndoH sensitivity to WT levels in the presence of the D429A Ca2+ binding mutation. EndoH sensitivity of Prf1 was calculated as described in Fig. 1a. D429A and A90V/D429A values (dotted lines) are reproduced from Fig. 1e. Each value represents ± SEM for at least three independent experiments and statistics were determined using ANOVA and Tukey’s post-hoc analysis; nsd no significant difference

Full size image

As discussed, we originally reasoned that the T431D variant may either interact directly with one of the bound Ca2+ ions or alternatively may facilitate conformational change in CBR-1. To discriminate between these possibilities we determined the apo- and Ca2+-bound structures of a Prf1 C2 domain variant that contains the T431D mutation (mutant generated on the W427A/Y430A/Y486A/W488A background [3, 4]) (Fig. 3). In the absence of Ca2+, T431 is mobile and missing from the final model, whereas D431 is clear in the electron density (Fig. 3a). In the presence of added Ca2+, however, the Ca2+ ions in sites I, II and III are conventionally coordinated by the same residues as in WT protein, and the T431D position does not directly interact with Ca2+ (Fig. 3b). These data thus favour the idea that the T431D variant instead facilitates conformational change in CBR-1, for example, through the provision of additional potential for long-range electrostatic interactions.

Fig. 3
figure 3

Crystal structure of T431 Prf1 reveals conformational changes in the calcium-binding regions (CBRs). Superposition of C2 domains crystal structures. a C2 domain of full-length Prf1 (PDB ID 3NJS, grey), C2_T431 (PDB IB 4Y1S, cyan, left) and C2_D431 (skyblue, right), no Ca2+ was added to the sample buffers or cyrstallization buffers (see Materials and methods). Also shown are high-affinity Ca2+ ions in the structures, presumably scavenged from the growth media, they are found in sites II and V of 3NJS, site V of C2_T431; no Ca2+ is found in the C2_D431 structure. All Ca2+ ions shown are coloured according to the corresponding C2 structures. Residues 433-438 of C2_T431 are missing from the electron density map and are represented as a red dashed line (left). b Crystal structures of C2 domains C2_T431 (PDB ID 4Y1T, cyan, left and bottom) and C2_D431 (skyblue, right and bottom) obtained in the presence of added Ca2+. Also shown is the C2 domain of full-length Prf1 (PDB ID 3NJS, grey, left and right) as for a. The structure of C2_D431 reveals three Ca2+ ions (CaI-III), whereas five Ca2+ ions can be observed in the structure of C2_T431. Also shown are residues D429, D483, D491, CBR-1–3 and loop 510–515 and distances between CaI and D429 and D491 (in angstrom, red)

Full size image

As we had shown Ca2+ binding in the ER regulates Prf1 folding and trafficking through stabilisation of its C2 domain, we hypothesised that T431D may potentially compensate for the misfolded phenotype of the A90V mutant. Owing to folding defects of A90V, we were unable to produce sufficient amounts of the protein to assess its melting temperature, but we found that engineering the T431D mutation onto the backbone of A90V Prf1 (A90V/T431D) resulted in greater thermal stability than WT Prf1 (Fig. 2a), and both T431D and A90V/T431D had a WT level of cytotoxic activity (Fig. 2b). Furthermore, in transiently transfected RBL cells, T431D improved EndoH sensitivity of A90V to the WT level (Fig. 2c), suggesting that under steady-state conditions both folding and trafficking of A90V were restored. This process remained dependent on essential Ca2+-binding residue/s of the C2 domain, as further addition of the D429A mutation (A90V/D429A/T431D) destabilised the protein and completely abrogated trafficking improvements provided by T431D (Fig. 2d). Furthermore, the D429A/T431D mutant was as sensitive to EndoH as the single D429A mutant, indicating once again that Ca2+ binding to the C2 domain of Prf1 in the ER was essential for the stabilising properties of T431D (Fig. 2d). From these results, we concluded that Ca2+-dependent stabilisation of the C2 domain in the ER is important for Prf1 folding, and subsequent trafficking to the secretory granules.

T431D-Prf1 has supra-physiological function

Similar to the ER, the synaptic cleft has neutral pH and high concentration of free Ca2+ (>1 mM). We therefore investigated whether enhanced stabilisation of T431D-Prf1 also affected its lytic function. Using purified recombinant WT and T431D proteins, we found no significant difference in their ability to bind to membranes (Fig. 4a), pore-size/geometry by negative stain EM (Fig. 4b) or to synergise with granzyme B to induce target cell apoptosis (Fig. 4c). However, target cell lysis at 1 mM Ca2+ was twice as fast with T431D-Prf1 as with WT (p < 0.0005) (Fig. 4d). Interestingly, at sub-physiological concentrations of Ca2+ (<0.5 mM), both T431D and A90V/T431D had relatively lower cytotoxic activity than WT (Fig. 4e). Speculatively, this may be due to the loss of CaIV, which is coordinated by D491 (Fig. 3).

Fig. 4
figure 4

Stabilisation of Prf1 with T431D increased the rate of target cell lysis. a Binding of recombinant WT and T431D-Prf1 to SRBC membranes. There was no significant difference in the amount of T431D-Prf1 bound to SRBC membranes in 1 mM Ca2+, or avidity of membrane binding, as detected by western blot after washing in Ca2+-free buffer. b Negative stain electron microscopy images of Prf1 oligomers on lipid monolayers. T431D oligomerises into arcs and rings similar to WT Prf1. Scale bar, 20 nm. c Prf1/GrzmB synergy is not affected by stabilisation of Prf1 with T431D. HeLa cells were treated with sublytic Prf1 in the absence (Prf1 only) or presence of 2, 3 and 5 μg of GrzmB and a 4 h 51Cr release assay was performed. Each value is ±SEM of three replicates, and is representative of two independent experiments. Statistics were obtained using unpaired t test, ns no significant difference. d The rate of SRBC lysis was increased almost 2-fold by stabilisation of Prf1 with T431D. Overall, 5 × 108 SRBCs/ml were incubated with purified recombinant mouse WT and T431D-Prf1 in 1 mM Ca2+ at 37 °C. The SRBC/Prf1 reaction was quenched with ice-cold EDTA after 30 s, 1, 2, 3, 4, 5, 10 and 15 mins. All values have been normalised against maximum lysis observed at 15 mins (100%); average maximum lysis was WT: 73.0 ± 10.0 SEM (n = 7) and T431D: 74.6 ± 12.2 SEM (n = 7). Dotted lines represent the time required to achieve 50% maximum cell lysis: 4.9 ± 0.4 min and 2.5 ± 0.3 min (mean ± SEM, n = 7 independent experiments; p < 0.0005 unpaired t test) for WT and T431D, respectively. e The lytic activity of recombinant WT and T431D-Prf1 (left) and A90V/T431D-Prf1 (right) in the presence of increasing Ca2+ concentrations was assessed using SRBCs. At sub-physiological Ca2+ concentrations of <0.5 mM, T431D and A90V/T431D had relatively lower cytotoxic activity than WT Prf1

Full size image

Next, we investigated the effect of the T431D mutation on Prf1 cytotoxicity in the context of an immune synapse using a standard RBL 51Cr release cytotoxicity assay [16]. The results were surprising: T431D-expressing RBL cells had >5-fold superior cytotoxicity in comparison to killer cells expressing WT Prf1 (Fig. 5a), in that over five times fewer killer cells expressing T431D-Prf1 were required to produce an equivalent level of 51Cr release (cell death). We also considered the possibility that overexpression of Prf1 in this experiment would potentially result in saturation of the cell death readout. As Prf1 cDNA was cloned into pIRES-eGFP vector and transiently transfected into RBL effector cells, we sorted high and low Prf1-expressing cells (based on eGFP fluorescence intensity) and assessed their cytotoxic activity separately. The cells expressing relatively low levels of T431D had >10-fold greater cytotoxicity than the cells expressing an equivalent amount of WT Prf1 (Fig. 5b). This was not due to an increased affinity of the C2 domain for Ca2+, as T431D was only cytotoxic at physiologically relevant concentrations of Ca2+, and only displayed augmented cytotoxicity within the range of Ca2+ concentrations that corresponded with enhanced stability (Fig. 5c).

Fig. 5
figure 5

Stabilisation of Prf1 with T431D increases cellular cytotoxicity. a Cytotoxicity of transiently transfected RBL cells expressing WT and T431D-Prf1, as determined by 51Cr release assay using Jurkat T-cells as targets, at the effector/target (E:T) ratios indicated (100% WT average maximum lysis at 30:1 E/T ratio was 42 ± 5 SEM). The data were fitted to Michaelis–Menten kinetics (using Prism 7.0.c). Using Michaelis–Menten equation, it was calculated that T431D required 4.9 effector cells (while WT required 30 effector cells) to achieve 100% lysis. Therefore, killer cells expressing T431D were 6.1-times more efficient than WT cells in producing an equivalent level of 51Cr release (dotted lines). Each value represents mean ± SEM of three independent experiments. b Cytotoxicity of transiently transfected RBL cells expressing WT and T431D-Prf1 sorted for high and low GFP fluorescence, as determined by 51Cr release assay using Jurkat T-cells as targets, at the effector/target (E:T) ratios indicated (100% WT average maximum lysis at 30:1 E/T ratio was 47.1 ± 11.5 SEM). The cytotoxicity of T431D was increased ~10-fold in low GFP expressing cells when compared with WT Prf1. Each value represents mean ± SEM of three independent experiments. Western immunoblot using the P1-8 antibody demonstrated equivalent protein expression of WT and T431D-Prf1 in low and high GFP expressing RBL cells. c Cytotoxicity of transiently transfected RBL cells expressing WT and T431D-Prf1, at increasing concentrations of Ca2+, as determined by a 51Cr release assay using Jurkat T-cells as targets, at a constant effector target (E:T) ratio of 30:1. Ca2+ was quenched from DMEM media using EGTA. Each value is mean ± SEM of three replicates, and is representative of two experiments. d Cytotoxicity of transiently transfected RBL cells expressing WT, T431D, A90V and A90V/T431D-Prf1, as determined by 51Cr release assay using Jurkat T-cells as targets, at the effector/target (E:T) ratios indicated (100% WT average maximum lysis at 30:1 E/T ratio was 50 ± 6% SEM). The A90V/T431D-Prf1 had ~5-fold increase in cytotoxicity when compared with the disease-causing mutant, A90V, alone. Each value represents mean ± SEM of three independent experiments. e WT or mutant Prf1 was transiently transfected (by electroporation) into RBL cells, plated and grown for 90–240 min. At the time-points shown above the Western blot (on the left), the cells were harvested by trypsinisation and the lysates were treated with EndoH to determine a relative proportion of EndoH-resistant (upper bands) and EndoH-sensitive (lower band) Prf1. The migration of deglycosylated Prf1 was determined by treating the lysate from a sample collected at 240 min with PNGaseF (shown as ‘F’) on the Western blot. ‘UT’ indicates an untreated lysate. The ratio of upper and lower bands was determined from three independent experiments; plotted is an acquisition of EndoH resistance over time (mean ± SEM)

Full size image

Finally, we investigated whether improved folding or trafficking through Ca2+-dependent stabilisation could also rescue A90V function. We found that this was indeed so: A90V/T431D cytotoxicity was significantly improved, above that of WT Prf1 (Fig. 5d), which was likely due to a combination of improved folding (Fig. 2a–c) and more rapid pore formation (Fig. 4d). To determine whether T431D facilitated trafficking of the A90V mutant, we assessed the rate of Prf1 export from the endoplasmic reticulum. Thus RBL cells were transiently transfected with Prf1 variants, and the rate of acquisition of complex glycosylation (EndoH resistance) over 90–240 min was a measure of Prf1 export from the ER and processing in the Golgi. The experiment demonstrated that the rate of trafficking of T431D was slightly higher than that of WT, but it did not improve the rate of A90V export from the endoplasmic reticulum (Fig. 5e).

Overall, the results of this study demonstrate that Ca2+-dependent stabilisation of Prf1 through the C2 domain is required for its folding, function and proteostasis in general.

Discussion

The current study uncovered a novel and critical role for the C2 domain in maintaining Prf1 proteostasis. This was demonstrated using a dual approach, where we first improved and then abrogated Ca2+ binding to the C2 domain of Prf1. Remarkably, by engineering a new Ca2+-stabilising mutation in the C2 domain of Prf1 (T431D), we produced a ‘super-Prf1’, which had significantly higher cytotoxic activity than WT Prf1. In addition, this mutant stabilised and rescued the delivery of a partially misfolded, common pathological mutant, A90V, to the secretory granules. In a reciprocal series of experiments, the abrogation of C2 domain function through mutation of critical Ca2+ binding aspartate residues resulted in a significant decrease in the delivery of A90V Prf1 to the secretory granules. Given that A90V is not part of the C2 domain, and its location is approximately 120 Å away from the Ca2+-binding regions [6], it appeared that Ca2+-dependent stabilisation of the C2 domain was critical for the efficient folding and export of Prf1 from the ER. It is unclear why T431D-Prf1 enhances the stability and function of Prf1, but the X-ray crystal structure suggests that this may be due to additional long-range electrostatic interactions within the C2 domain. Interestingly, we found that T431D had reduced cytotoxic activity at sub-physiological Ca2+ concentrations and that under our crystallisation conditions it was missing one Ca2+ (CaIV) that was coordinated by D491. We are led to speculate that there may be a link between the two phenomena, as we showed previously that the D491A mutation had no effect on Prf1 function at physiological concentrations of Ca2+, but the activity of the mutant was impaired at low (non-physiological) Ca2+ levels [16].

An increase in Prf1 stability through Ca2+ binding to the C2 domain also provided an oligomerisation advantage, whereby T431D could lyse cells at a faster rate than WT Prf1 to confer both augmented cytotoxicity and recovery of oligomerisation deficiencies. These results suggest that the more stable T431D-Prf1 may be ‘primed’ for more efficient oligomerisation and, potentially, pore formation. This concept is supported by previous studies, which have reported that WT Prf1 becomes significantly more stable under conditions that favour Prf1 pore formation within the immunological synapse [4, 8]. Taken together, it appears that in addition to Ca2+-dependent membrane binding [16], stabilisation of the C2 domain of WT Prf1 may be a previously unrecognised step required for pore formation.

The C2 domain of Prf1 is unique, with a relatively low affinity for Ca2+ in comparison with intracellular C2 domain proteins, which respond to submicromolar fluctuations in Ca2+ concentrations to become activated. Although the neutral pH and high concentration of Ca2+ in the ER create a favourable environment for Prf1 lethal pore-forming activity, cytotoxic lymphocytes can rapidly export it through the Golgi to the acidic secretory granules [14]. In addition, the C-terminal glycosylation protects the cell from Prf1 toxicity within the endoplasmic reticulum [15]. The results of our study indicate that, paradoxically, the high level of Ca2+ in the ER promotes Prf1 stabilisation, which is necessary for efficient protein folding and trafficking. Unlike high-affinity calcium-binding sites II and V, which are likely to be constitutively occupied, low affinity Ca2+ binding sites I, III and IV act as a chemical sensor that only responds to high Ca2+ concentrations [4]. Once the protein has been folded and exported from the ER, it likely becomes functionally inert in the low Ca2+ environment of the Golgi and acidic pH of the secretory granules, where it is unable to oligomerise and insert into the membrane [8]. Overall, Ca2+ binding to the C2 domain has a critical role in the proteostasis of Prf1, which requires a precise balance between Ca2+ and membrane binding, folding in the ER, export to the Golgi, and cytotoxic lymphocyte toxicity.

It is well established that disease-causing Prf1 mutations are the result of protein misfolding in the ER [12]. Our discovery that Prf1 folding in the ER is Ca2+-dependent provides a rationale for the development and application of specific drug therapies for FHL caused by Prf1 deficiency. Patients with late-onset FHL2 often require aggressive cytotoxic drug therapy to treat the disease prior to bone marrow transplantation. Alternative therapies that modify the proteostasis network through a post-translational mechanism [21, 22] are significantly less aggressive and may provide a more suitable alternative to initial treatment. Such an approach would facilitate targeting of Prf1-expressing lymphocytes without damaging other cell types, and will be the focus of ongoing studies.

Materials and methods

Expression and binding of recombinant Prf1

Mouse Prf1 mutants were expressed and purified using a baculovirus expression system as previously described [23]. Lytic activity were assessed using sheep erythrocytes (SRBC), and cell death resulting from the synergistic effects of Prf1 and granzyme B (GzmB) was assessed by 51Cr release assays [24]. Membrane binding was assessed by incubating Prf1 and SRBCs in 10 mM Hepes, 150 mM NaCl, pH 7 with or without 1 mM Ca2+ at 4 C for 15 min, followed by four subsequent washes in Ca2+-free buffer. The thermal stability of purified recombinant mouse WT and mutant Prf1 was assessed by unfolding temperature analysis using SYPRO Orange [4].

Transient transfection of cell lines and cytotoxicity assays

Rat basophil leukaemia cells RBL-2H3 (RBL) and Jurkat human T leukaemia target cells were maintained in culture as previously described [23]. Point mutations in mouse WT Prf1 cDNA were generated using the QuikChange site-directed mutagenesis system and cloned into the pIRES-EGFP expression plasmid (Biosciences Clonech). The WT and mutant Prf1 plasmids were transiently transfected into RBL cells, which were sorted by flow cytometry for equal mean GFP fluorescence 24 h later, with subsequent assessment of cytotoxic activity in 51Cr release assays with Jurkat T cells [23]. Prf1 RBL cells sorted by flow cytometry for equal mean GFP fluorescence were also used for Western Immunoblot (see below).

Electrophoresis and immunoblotting

Whole cell lysates, SRBC membranes and purified recombinant mouse Prf1 were prepared in NP40 lysis buffer (250 mM NaCl, 25 mM HEPES, 2.5 mM EDTA, 1% [v/v] NP40, supplemented with protease inhibitors cocktail (Roche Life Sciences)). EndoH and PNGaseF glycosidase treatment was conducted according to the manufacturer’s instructions (New England Biolabs). Protein lysates were separated on 4–12% Nu-Page Bis-Tris gradient gels (Invitrogen) and immunoblotted with rat anti-Prf1 mAb P1-8 (provided by Kyowa Kirin) followed by secondary HRP-linked anti-rat antibody. The loading control used for cell lysates was mouse anti-human actin mAb (Sigma) followed by secondary HRP-linked anti-mouse Ig (Dako). Signals were amplified by chemiluminescence and detected on X-ray film (GE Healthcare). ImageJ software (ImageJ 1.475v, NIH) was used for quantitative densitometry.

Negative stain electron microscopy sample preparation

Purified recombinant WT and T431D-Prf1 pores were formed on DMPC/cholesterol lipid monolayers as described [6]. Images were collected on a Gatan 4k×4k CCD camera (15μm per pixel) on a Tecnai F20 microscope (FEI) at 200 keV and ×67,000 magnification.

Crystallography

Recombinant Prf1 C2 domains (residues 410–535) were expressed and purified from E. coli as described before [3]. A mutant form, which carries four amino acid substitutions (W427A/Y430A/Y486A/W488A, C2_T431) was used for the current studies to generate C2_D431. C2_T431D was crystallised in 0.2 M ammonium iodide, 20% PEG 3350 in the absence of added Ca2+, or in 0.1 M MgCl2, 0.1 M Na-HEPES pH 7.5 and 10% w/v PEG 4000 in the presence of 1 mM Ca2+ (Table 1). The crystals were flash-cooled in liquid nitrogen using 25% (v/v) glycerol as the cryoprotectant. Data sets were collected at the Australian Synchrotron MX2 beamline at 100 K. The data were merged and processed using XDS [25], POINTLESS [26] and SCALA [27] or AIMLESS [28]. Five per cent of the data set was flagged as a validation set for calculation of the Rfree. Molecular replacement (MR) was carried out using WT Prf1 structure 3NSJ as a search probe [29] One molecule was found per asymmetric unit cell and an initial model was generated using PHASER. Model building was performed using COOT [30] and refinement was performed using BUSTER (Cambridge, United Kingdom: Global Phasing Ltd). Crystallographic and structural analysis was performed using the CCP4 suite [27] unless otherwise specified. The figures were generated using MacPYMOL (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC.) and the structural validation was performed using MolProbity [31]. All atomic coordinates and structural factors were deposited in the PDB under codes 5UG6 (in the absence of Ca2+) and 5UG7 (in the presence of Ca2+).

Table 1 Data collection and refinement statistics
Full size table

Statistical analysis

The statistical analyses used were: paired t test (when comparing two groups); or one-way ANOVA with Tukey’s post-hoc analysis (when comparing more than two groups). The application of each test is indicated in the figure legends. A non-linear regression analysis was applied to all SRBC lysis assays and 51Cr release killing assays for clarity.