Calcium Dynamics in the Peroxisomal Lumen of Living Cells*

We here describe the generation of novel, green fluorescent protein-based Ca2+ indicators targeted to the peroxisome lumen. We show that (i) the Ca2+ concentration of peroxisomes in living cells at rest is similar to that of the cytosol; (ii) increases in cytosolic Ca2+ concentration (elicited by either Ca2+ mobilization from stores or Ca2+ influx through plasma membrane Ca2+ channels) are followed by a slow rise in intraperoxisomal [Ca2+]; (iii) Ca2+ influx into peroxisomes is driven neither by an ATP-dependent pump nor by membrane potential nor by a H+(Na+) gradient. The peroxisomal membrane appears to play a low pass filter role, preventing the organelle from taking up shortlasting cytosolic Ca2+ transients but allowing equilibration of the peroxisomal luminal [Ca2+] with that of the cytosol during prolonged Ca2+ increases. Thus, peroxisomes appear to be an additional cytosolic Ca2+ buffer, but their influx and efflux mechanisms are unlike those of any other cellular organelle.

We here describe the generation of novel, green fluorescent protein-based Ca 2؉ indicators targeted to the peroxisome lumen. We show that (i) the Ca 2؉ concentration of peroxisomes in living cells at rest is similar to that of the cytosol; (ii) increases in cytosolic Ca 2؉ concentration (elicited by either Ca 2؉ mobilization from stores or Ca 2؉ influx through plasma membrane Ca 2؉ channels) are followed by a slow rise in intraperoxisomal [Ca 2؉ ]; (iii) Ca 2؉ influx into peroxisomes is driven neither by an ATP-dependent pump nor by membrane potential nor by a H ؉ (Na ؉ ) gradient. The peroxisomal membrane appears to play a low pass filter role, preventing the organelle from taking up shortlasting cytosolic Ca 2؉ transients but allowing equilibration of the peroxisomal luminal [Ca 2؉ ] with that of the cytosol during prolonged Ca 2؉ increases. Thus, peroxisomes appear to be an additional cytosolic Ca 2؉ buffer, but their influx and efflux mechanisms are unlike those of any other cellular organelle.
A variation in cytosolic Ca 2ϩ is a key component of the cell signaling machinery activated by receptor stimulation. Although a plethora of information is available regarding Ca 2ϩ dynamics in different subcellular compartments, a notable exception is represented by peroxisomes, single membranebound organelles diffusely distributed within the cytosol of virtually all eukaryotic cells (1). Proteins located in the peroxisomal matrix are linked to different biochemical pathways (2) such as the ␤-oxidation of fatty acids and detoxification of hydrogen peroxide. The latter pathway is exclusively localized in the peroxisomal compartment of fungi and plants, whereas in mammalian cells it is distributed between peroxisomes and mitochondria (2). Specialized peroxisomal functions, such as fatty acid degradation and synthesis of phytohormones, are found in some cells, (e.g. plants and fungi) (3). Interest in peroxisomes has increased recently due to the discovery that defects in peroxisomal biogenesis and peroxisomal enzyme deficiencies are linked to several genetic disorders in humans (4). Given that any enzymatic activity is highly sensitive to the ionic composition of the surrounding environment, it is surprising that information on the luminal ion content of peroxisomes is scarce and contradictory. In particular, no data are currently available on Ca 2ϩ concentration in the peroxisome lumen, [Ca 2ϩ ] p .
We here present a novel probe, derived from the new green fluorescent protein (GFP) 2 -based Ca 2ϩ indicators (Dcpv) (5), for monitoring [Ca 2ϩ ] p in living cells. We show that peroxisomes contribute to the sequestration of part of the Ca 2ϩ entering the cytoplasm during cell activation in a way that is unique among cellular organelles.
Cell Imaging-Cells expressing (or loaded with) the fluorescent probes were analyzed using an inverted fluorescence microscope (Zeiss Axioplan) with an immersion oil objective (ϫ63, N.A. 1.40, for fluorescent probes and ϫ40, N.A. 1.3, for fura-2 and BCECF). Excitation light was produced by a monochromator (Polychrome II; TILL Photonics, Martinsried, Germany): 400 and 480 nm for pHluorin; 340 and 380 nm for fura-2; 495 and 440 nm for BCECF. The two excitation wavelengths were rapidly alternated and the emitted light deflected by dichroic mirrors (HQ 520 LP for pHluorin and BCECF and 455DRPL for fura-2) was collected through emission filters (HQ 520 LP for pHluorin and BCECF and 480 ELFP for fura-2). For the D3-derived probe, the excitation light was 425 nm. The emitted light was collected through a beamsplitter (OES s.r.l., Padua, Italy) (emission filters HQ 480/40M for cyan fluorescent protein and HQ 535/30 M for yellow fluorescent protein) and a dichroic mirror (515 DCXR). Filters and dichroic mirrors were purchased from Omega Optical and Chroma. Images were acquired using a cooled CCD camera (Imago; TILL Photonics) attached to a 12-bit frame grabber. Synchronization of the monochromator and CCD camera was performed through a control unit run by TILLvisION v.4.0 (TILL Photonics); this software was also used for image analysis. For time course experiments, the fluorescence intensity was determined over regions of interests covering small groups of peroxisomes or cytosolic regions (devoid of identifiable structures). Exposure time and frequency of image capture varied from 30 to 500 ms and from 5 to 0.2 Hz, respectively. Cells were mounted into an open-topped chamber thermostated at 37°C and maintained in an extracellular medium containing (in mM): 135 NaCl, 5 KCl, 1 MgSO 4 , 0.4 KH 2 PO 4 , 10 glucose, 20 Hepes, pH 7.4, at 37°C. Plasma membrane permeabilization was performed by treating cells for 1 min with 100 M digitonin in an intracellular-like medium containing (in mM): 130 potassium-gluconate, 10 NaCl, 1 KH 2 PO 4 , 1 MgSO 4 , 20 Hepes, pH 7.0, at 37°C and 500 M EGTA. Experiments with permeabilized cells were performed in the same medium; where indicated, the latter was supplemented with a buffer containing (in mM): 2 EGTA, 1 H-EDTA, 1 MgCl 2 , and variable CaCl 2 concentration.
Immunocytochemistry-Cells were fixed in phosphate-buffered saline containing 4% paraformaldehyde for 20 min, permeabilized with 0.1% Triton X-100 in phosphate-buffered saline containing 0.5% bovine serum albumin and 0.15% glycine (PBG) for 20 min, and blocked with 5% non-immune goat serum in PBG for 30 min. Rabbit anti-catalase (Rockland Immunochemicals, Gilbertsville, PA) or rabbit anti-PMP70 (Sigma) antibodies were added for 1 h at 37°C. Samples were washed three times in PBG and then treated with Alexa Fluor 568 goat anti-rabbit IgG (Invitrogen) for 1 h at room temperature. Samples were washed three times in PBG and three times in phosphate-buffered saline, mounted with Mowiol (Sigma), and analyzed with a Leica TCS SP5 confocal microscope using an immersion oil objective (ϫ63, N.A. 1.40). Image acquisition was performed by sequential scanning with excitation wavelengths of 488 nm for D3cpv-(KVK)-SKL and 543 nm for Alexa Fluor 568. Emission wavelengths were collected in the 495 to 535-nm (D3cpv-(KVK)-SKL) and 580 to 630-nm (Alexa Fluor 568) ranges. Correction of the bleed-through from the green fluorescence into the Alexa Fluor 568 channel and merging of the emitted fluorescence were carried out using the Acousto Optical Beam Splitter technique and the software provided by the manufacturer of the confocal microscope.
Statistical Analysis-All the data are representative of at least five different experiments. Values are expressed as mean Ϯ S.E.

RESULTS
Peroxisome Targeting of the GFP-based Ca 2ϩ Indicator- Fig.  1, top, shows the schematic structure of D3cpv, modified by the insertion at the C-terminal of the canonical peroxisomal targeting signal, the tripeptide Ser-Lys-Leu (SKL) (6). Although this sequence is known to be efficacious in targeting several recombinant proteins to peroxisomes, the D3cpv-SKL subcellular distribution ( Fig. 1A) in HeLa cells transiently expressing the construct was indistinguishable from that of cytosolic D3cpv (Fig.  1B). Treatment of cells with digitonin, although releasing all cytosolic D3cpv (not shown), revealed that a fraction of the D3cpv-SKL was trapped in numerous small structures scattered throughout the cytoplasm (Fig. 1C). The D3cpv-SKL-positive spots coincide with peroxisomes, as revealed by their positivity after immunostaining with antibodies for markers of these organelles, catalase (Fig. 2B) or the peroxisomal membrane protein 70 (Fig. 2E). The missorting of D3cpv-SKL was observed in all other cell types investigated, GH3, Chinese hamster ovary, and SH-SY5Y. The cytoplasmic staining was not due to protein overexpression and saturation of the peroxisome protein import mechanism, because the same results were obtained when transfection was carried out with only 1/5 of the cDNA or if the cells were observed 48, 72, or 96 h after transfection (not shown). To improve the peroxisome localization, a novel construct was made where the C-terminal SKL was preceded by a threeamino acid positively charged sequence, Lys-Val-Lys (KVK). This sequence was designed to fit the requirement for improved peroxisomal targeting described by Neuberger et al. The experiment presented in Fig. 3G was aimed at determining whether the Ca 2ϩ probe was trapped within the peroxisome lumen or whether it was bound to the cytosolic surface of peroxisomes. The plasma membrane of HeLa cells was permeabilized with digitonin, and the cells were then treated with Proteinase K. The protease did not affect the D3cpv-KVK-SKL fluorescent signal, whereas, on the contrary, in cells expressing a GFP construct localized on the cytosolic surface of the outer mitochondrial membrane, TOM20-GFP (9), the enzyme abolished the fluorescence in a few seconds. Similar results were obtained in GH3 cells (not shown).
Ca 2ϩ Handling by Peroxisomes in Intact Cells-We used as a first model system GH3 cells because these cells are endowed with (i) abundant plasma membrane voltage-gated Ca 2ϩ channels and (ii) endogenous receptors (TRH receptors) coupled to inositol 1,4,5-trisphosphate production and Ca 2ϩ mobilization from stores (10). Fig. 4A shows the typical response pattern of the D3cpv-KVK-SKL fluorescence signal of three GH3 cells to depolarization with 30 mM KCl. In two cells, the fluorescence was exclusively in peroxisomes, whereas in the third cell fluorescence was diffuse throughout the whole cytosol. Cell depolarization caused an increase in the fluorescence emitted at 540 nm and a decrease of the signal at 480 nm (not shown) and thus an increase in the 540/480-nm fluorescence emission ratio (here presented as ⌬R/R 0 ), which is proportional to [Ca 2ϩ ] (Fig.  4A). The kinetics of the ⌬R/R 0 changes were, however, different in the cells where the probe was localized in the peroxisomes and in the cell with the mistargeted indicator. The cytosolic ⌬R/R 0 (continuous line) reached the peak in 1-2 s and then started to decrease slowly; the peroxisome signal, on the contrary, reached the peak in 10 -15 s and then started to decline. Addition of EGTA accelerated the drop to basal level of both the cytosolic and peroxisomal signals, the effect on the cytosol being more evident. In Fig. 4B, the fluorescence emission ratio (excitation 340/380 nm) of a parallel batch of cells loaded with the Ca 2ϩ indicator fura-2 is presented. The kinetics of the fura-2 signal were similar to that of cells expressing the missorted D3cpv-KVK-SKL probe (Fig. 4A). Similar data were obtained with cells expressing the original cytosolic D3cpv (not shown). In the experiment presented in Fig. 4C    The question then arises as to the behavior of peroxisomes, in terms of Ca 2ϩ response, to agents that cause Ca 2ϩ mobilization from intracellular stores, either elicited by TRH or by the Ca 2ϩ ionophore ionomycin, both added in the absence of extracellular Ca 2ϩ . The two agents caused neither a drop nor a rise in [Ca 2ϩ ] p (under conditions that elicited significant transient Ca 2ϩ rises, as measured with fura-2; compare Fig. 5, A and C, with Fig. 5, B and D, dotted traces). When TRH or ionomycin was added after a previous pulse of KCl (to overload Ca 2ϩ stores), the percentage of peroxisomal Ca 2ϩ increases in response to TRH and ionomycin increased significantly (20 and 53% of cells, respectively; not shown). The problem of the peroxisomal behavior in response to Ca 2ϩ -mobilizing stimuli was then further addressed in HeLa cells treated with histamine or ionomycin (Fig. 5). In all cells investigated, histamine induced a cytosolic [Ca 2ϩ ] rise, as measured with fura-2 (Fig. 5B, continuous trace), whereas in 68% of cells the peroxisome signal also increased significantly (Fig. 5A, continuous trace). In HeLa cells, addition of ionomycin in Ca 2ϩ -free medium (which resulted in a large cytosolic [Ca 2ϩ ] increase in all cells tested; Fig. 5D, continuous trace) always resulted in a rise of [Ca 2ϩ ] p (Fig. 5C, continuous trace).
Mechanism of Ca 2ϩ Transport in Peroxisomes-The Ca 2ϩ rise within peroxisomes induced by KCl-dependent depolarization in GH3 cells was indistinguishable in the presence or absence of mitochondrial uncouplers or of sarcoendoplasmic reticulum Ca 2ϩ ATPase inhibitors (not shown). Given that no reliable inhibitor of the Golgi-type pump is available, to verify the involvement of ATP-dependent uptake mechanisms we investigated the effects of ATP on the rate and extent of [Ca 2ϩ ] p rise in digitonin-permeabilized cells exposed to a medium with 500 nM Ca 2ϩ . As shown in Fig. 6A, Ca 2ϩ uptake was similar with (continuous trace) and without (dotted trace) an energy source. Notably, when an excess EGTA was added (to rapidly decrease medium [Ca 2ϩ ]), the peroxisome [Ca 2ϩ ] decreased with relatively slow kinetics (Fig. 6B) To test whether peroxisomal Ca 2ϩ influx depends on the presence of a classical Ca 2ϩ channel, digitonin-permeabilized cells were treated with 10 M La 3ϩ , a nonspecific inhibitor of several Ca 2ϩ channels. The increase in [Ca 2ϩ ] p upon increase in medium [Ca 2ϩ ] to 500 nM or 5 M was unaffected by La 3ϩ (not shown).
We then investigated whether peroxisomal Ca 2ϩ uptake may depend on a Na ϩ (H ϩ )/Ca 2ϩ antiport. Intact GH3 cells were pretreated with either NH 4 Cl (Fig. 6C, dotted trace), an agent that causes an alkalinization of organelle pH, or monensin (dashed trace), a H ϩ /Na ϩ exchange ionophore, which should collapse any gradient of either Na ϩ or H ϩ across the peroxisomal membrane, if they exist. Neither NH 4 Cl nor monensin had any appreciable effect on the [Ca 2ϩ ] p increase caused by 30 mM KCl. Similar results were obtained in HeLa cells stimulated with either histamine or ionomycin (not shown).  To verify whether there are heterogeneities among the organelles, the [Ca 2ϩ ] rise in different groups of peroxisomes was next investigated. As shown in Fig. 6D, the response to a 30-mM KCl challenge of different groups of organelles within the same GH3 cell was found to be very similar. Identical results were obtained in HeLa cells using either ionomycin or histamine as the stimulus (not shown).
Finally, the peroxisome lumenal pH was directly monitored using the targeted pH indicator pHluorin (see "Materials and Methods" and Ref. 11). Cytosolic pH was measured in parallel with BCECF (12). Fig. 7 shows that the weak acid acetate caused a reduction of both cytoplasmic (Fig. 7A) and peroxisomal (Fig.  7B) pH, whereas NH 4 Cl caused an alkalinization of both compartments. Monensin also caused an increase of pH both in the cytosol (Fig. 7C, continuos trace) and in peroxisomes (Fig. 7D,  dotted trace). When the cells were incubated in a medium where NaCl was iso-osmotically substituted with KCl (to abolish the Na ϩ gradient across the plasma membrane and in the absence of Ca 2ϩ to block Ca 2ϩ influx) and the extracellular pH was dropped to 7.0 (to reduce the pH gradient), monensin hardly modified cytosolic pH (Fig. 7C, dashed point trace) and in parallel failed to cause any significant change in peroxisomal pH (Fig. 7D, dashed trace).
Calibration of the Peroxisomal [Ca 2ϩ ]-To determine the absolute values of [Ca 2ϩ ] p , the in situ K d for Ca 2ϩ of D3cpv-KVK-SKL was determined using the passive Ca 2ϩ loading procedure previously described (13). Transfected cells were permeabilized with digitonin in an intracellular-like medium, but devoid of ATP or any mitochondrial oxidizable substrate, and variable concentration of Ca 2ϩ (see "Materials and Methods"). The percentage of the normalized 540/480-nm fluorescence emission ratio changes at steady state were then plotted as a function of medium [Ca 2ϩ ] (Fig. 8)

DISCUSSION
The most common peroxisome-targeting mechanism involves the C terminus tripeptide SKL (6). When this sequence was added to the GFP-based Ca 2ϩ indicators D1-and D3cpv (5) most of the transfected protein mislocalized to the cytosol. Inclusion of a longer targeting sequence (KVK-SKL), however, resulted in more satisfactory peroxisome localization. The expressed protein is clearly trapped in the lumen of the organelles, as demonstrated by its resistance to proteolytic cleavage and by the slower kinetics of the fluorescence signal changes in response to a sudden change in extraperoxisomal [Ca 2ϩ ].
When cytosolic [Ca 2ϩ ] was increased in GH3 cells by depolarizing the plasma membrane with high KCl, the [Ca 2ϩ ] p also raised, although with slower kinetics. The amplitude of the [Ca 2ϩ ] p increase paralleled that of the cytosol. In quantitative terms, the maximum rises of [Ca 2ϩ ] p after depolarization were lower than those calculated with the classical cytosolic indicator fura-2. Considering the inherent assumptions involved in the calibration procedures of the two probes, it can be safely concluded that [Ca 2ϩ ] p tends to equilibrate with the cytosolic [Ca 2ϩ ] and no driving force (ATP and/or Na ϩ (H ϩ ) gradients) leads to Ca 2ϩ influx into peroxisomes. In support of this conclusion, the luminal pH of peroxisomes is practically indistinguishable from that of the cytoplasm, and monensin never caused an acidification of peroxisomal lumen, demonstrating that [Na ϩ ] of peroxisomes is similar to that of cytoplasm. Our conclusion concerning the lack of any significant gradient of H ϩ across the peroxisomal membrane concurs with Jankowski et al. (11), whereas other groups have reported that the intraperoxisome pH is slightly alkaline in mammalian cells (14) or in yeasts slightly acidic (15) or alkaline (16,17).
A permeability barrier to Ca 2ϩ diffusion across the peroxisome membrane, however, does exist as demonstrated by these results: (i) the rate of peroxisome Ca 2ϩ rise in intact cells treated with KCl is substantially slower than in the cytosol, and (ii) in permeabilized cells, sudden changes in medium [Ca 2ϩ ] require several seconds to equilibrate with the organelle lumen. Surprisingly, whereas increases in cytosolic [Ca 2ϩ ] elicited in GH3 cells by Ca 2ϩ influx though voltage-gated Ca 2ϩ channels were followed by [Ca 2ϩ ] p rises, Ca 2ϩ mobilization from internal stores, as induced by stimulation of TRH receptors, almost never resulted in a significant increase in [Ca 2ϩ ] p . Even unspecific Ca 2ϩ mobilization from stores, as promoted by ionomycin added in Ca 2ϩ -free medium, was unable to induce Ca 2ϩ uptake into peroxisomes of GH3 cells. The possibility was thus considered that the poor response of the peroxisomes to Ca 2ϩ mobilization in GH3 cells reflects (i) the existence of a mechanism that prevents Ca 2ϩ uptake in peroxisomes in response to Ca 2ϩ mobilization from stores or (ii) a combination of the small and transient nature of the cytosolic Ca 2ϩ rise in response to TRH (and ionomycin) in GH3 cells and of the slow Ca 2ϩ uptake rate by peroxisomes. In other words, the small and transient rise in cytosolic [Ca 2ϩ ] (as that elicited in GH3 cells by TRH or ionomycin) can be hardly coped with by the relatively slow Ca 2ϩ uptake system of peroxisomes. The cytoplasmic [Ca 2ϩ ] rise in response to depolarization, instead, does reach the peak in 2-3 s, but it is followed by a prolonged plateau level that lasts several tens of seconds. In support of the latter explanation, the very rapid Ca 2ϩ increases due to spontaneous action potential firing (and Ca 2ϩ influx through voltage-gated Ca 2ϩ channels) often observed in GH3 cells (18) were never followed by significant increases in [Ca 2ϩ ] p .
To distinguish between these possibilities, we used a different cell type, HeLa, where Ca 2ϩ mobilization from stores in response to an inositol 1,4,5-trisphosphate-generating agonist, such as histamine, results in larger and relatively more prolonged Ca 2ϩ transients compared with GH3 cells (peak values measured with fura-2 of 1.31 M and 270 nM, back to basal levels in 120 and 50 s, in HeLa and GH3 cells, respectively). Indeed, we found that in HeLa cells the percentage of peroxisome responses to histamine application was much higher than that observed in GH3 cells in response to TRH (68 versus 1%, respectively) and the percentage of [Ca 2ϩ ] p increases in response to ionomycin was close to 100% in HeLa cells compared with Ͻ5% in GH3 cells. Thus, it may be concluded that, due to the intrinsic sluggish response to a cytosolic Ca 2ϩ rise, peroxisomes are relatively insensitive to rapid transients of cytosolic [Ca 2ϩ ] but significantly increase their Ca 2ϩ level only in response to prolonged cellular Ca 2ϩ increases. We cannot exclude, however, that peroxisomes of HeLa cells are more efficient than those of GH3 cells at taking up Ca 2ϩ . However, when in GH3 cells TRH-or ionomycin-induced cytosolic Ca 2ϩ increases are larger and more prolonged (as occurs when they are applied after KCl), the percentage of peroxisomal responses increases drastically (from 1 to 21% for TRH and from 5 to 53% with ionomycin), suggesting that the first explanation is most likely.
The final question concerns the heterogeneity of peroxisomal Ca 2ϩ responses. When groups of organelles in the same cell were compared, no significant difference, either in kinetics or in amplitude of the Ca 2ϩ responses, was ever observed. It cannot be excluded, however, that single organelles localized in the proximity of Ca 2ϩ channels of either the plasma membrane or the endoplasmic reticulum may experience larger local Ca 2ϩ rises and, accordingly, undergo larger Ca 2ϩ increases.
In conclusion, we have developed novel GFP-based Ca 2ϩ indicators that can efficiently target to the peroxisomal lumen. These allow, for the first time to our knowledge, the measurement of this parameter in intact living cells. Taken together, the present data demonstrate that peroxisomes participate in the Ca 2ϩ signaling pathway but their behavior is unlike that of any other organelle. In particular, peroxisomes do not act as Ca 2ϩ stores from which Ca 2ϩ can be mobilized upon stimulation, as the endoplasmic reticulum, the Golgi apparatus or, in some cells, acidic compartments (19). The Ca 2ϩ response of peroxisomes to a rise in cytosolic [Ca 2ϩ ] is also markedly different from that of mitochondria, in as much as their luminal Ca 2ϩ does not increase as massively as that of the latter organelles. The organelle that most resembles peroxisomes in terms of Ca 2ϩ response is the nucleus, although in the latter the kinetics of Ca 2ϩ equilibration with the cytosol are 10 -100-fold faster. Thus, because of this relatively slow Ca 2ϩ influx, very rapid and transient increases in cytosolic Ca 2ϩ may not lead to appreciable changes in [Ca 2ϩ ] p , whereas more sustained increases will always lead to an increase in [Ca 2ϩ ] p . It remains to be established whether and which reactions within the peroxisomes are affected by Ca 2ϩ .
The amount of Ca 2ϩ that is sequestered by peroxisomes will depend on (i) their number and volume (which may vary among different cells and in response to specific stimuli, e.g. peroxisome proliferator-activated receptor ␥ gene activation) and (ii) the endogenous Ca 2ϩ buffering capacity of the organelles, which is presently unknown. In addition to a potential role as a cytosolic Ca 2ϩ buffer, the increases in [Ca 2ϩ ] p may be relevant for the organelle's own functions. Thus far, potential candidates are the peroxisomal Ca 2ϩ -dependent members of the mitochondrial carrier superfamily that contains four EF-hand Ca 2ϩ binding domains (20) or a Ca 2ϩ /calmodulin-regulated catalase isoform found in plant peroxisomes (21). The search for Ca 2ϩmodulated peroxisomal proteins may now be launched on a firmer ground, given the direct demonstration of the participation of these organelles in cellular Ca 2ϩ handling.