- Split View
-
Views
-
Cite
Cite
Carlotta Dell'Agnello, Sara Leo, Alessandro Agostino, György Szabadkai, Cecilia Tiveron, Alessandra Zulian, Alessandro Prelle, Pierre Roubertoux, Rosario Rizzuto, Massimo Zeviani, Increased longevity and refractoriness to Ca2+-dependent neurodegeneration in Surf1 knockout mice, Human Molecular Genetics, Volume 16, Issue 4, 15 February 2007, Pages 431–444, https://doi.org/10.1093/hmg/ddl477
- Share Icon Share
Abstract
Leigh syndrome associated with cytochrome c oxidase (COX) deficiency is a mitochondrial disorder usually caused by mutations of SURF1, a gene encoding a putative COX assembly factor. We present here a Surf1−/− recombinant mouse obtained by inserting a loxP sequence in the open reading frame of the gene. The frequency of −/−, +/+ and +/− genotypes in newborn mice followed a mendelian distribution, indicating that the ablation of Surf1 is compatible with postnatal survival. The biochemical and assembly COX defect was present in Surf1loxP−/− mice, but milder than in humans. Surprisingly, not only these animals failed to show spontaneous neurodegeneration at any age, but they also displayed markedly prolonged lifespan, and complete protection from Ca2+-dependent neurotoxicity induced by kainic acid. Experiments on primary neuronal cultures showed markedly reduced rise of cytosolic and mitochondrial Ca2+ in Surf1loxP−/− neurons, and reduced mortality, compared to controls. The mitochondrial membrane potential was unchanged in KO versus wild-type neurons, suggesting that the effects of the ablation of Surf1 on Ca2+ homeostasis, and possibly on longevity, may be independent, at least in part, from those on COX assembly and mitochondrial bioenergetics.
INTRODUCTION
Cytochrome c oxidase (COX), the terminal enzyme of the mitochondrial respiratory chain (MRC), catalyzes the transfer of electrons from reduced cytochrome c to molecular oxygen (1). COX is composed of 13 protein subunits, the three largest being encoded by mtDNA genes, and the remaining ten are encoded by nuclear DNA genes (2). A number of accessory factors are necessary for the formation of an active holoenzyme complex (3), including those involved in synthesis of heme a, incorporation of copper atoms and assembly of the protein backbone (4). One of these factors, SURF1, is a 30 kDa hydrophobic protein embedded in the inner membrane of mitochondria. The absence, or malfunctioning, of SURF1p determines the accumulation of COX assembly intermediates, and a drastic reduction in the amount of fully assembled enzyme, in both yeast (5) and humans (6). As a consequence, profound COX deficiency (7) in multiple tissues of Surf1 mutant patients (8) leads to the development of Leigh syndrome (LSCOX), an early onset, invariably fatal mitochondrial encephalomyelopathy (9).
In vertebrates, SURF1 is part of the very tight and highly conserved surfeit gene cluster, which includes six genes (SURF1–6) (10). The reason for long-standing maintenance of such a compact physical organization is obscure, since the corresponding SURF proteins are neither functionally nor structurally related to each other. The precise function of the SURF1 gene product itself remains unknown, although the results of several studies in yeast and mammals suggest a role for SURF1 protein (SURF1p) as an auxiliary chaperone-like factor, involved in the early assembly steps of the COX protein backbone (6).
To better understand the role of SURF1p and the pathogenesis of LSCOX, we have previously created a constitutive knockout (KO) mouse model, in which exons 5–7 of the Surf1 gene were replaced by a neomycin-resistance (NEO) cassette (11). Approximately 90% of the Surf1NEO−/− mice died at E6.5–7.5. The few animals that reached birth partially recapitulated, although to a lesser extent, the biochemical findings, but failed to display the clinical and neuro-pathological features of human LSCOX.
We present here a second Surf1 KO model, based on the insertion of a loxP sequence in exon 7 of the murine Surf1 gene (Surf1loxP), leading to an aberrant, prematurely truncated and highly unstable protein. The +/+, +/− and −/− genotypes in newborn animals were in agreement with the mendelian distribution, indicating that, rather than to the ablation of Surf1 itself, the high embryonic lethality observed in the previous Surf1NEO−/− model was due to a spurious effect of the NEO cassette on the expression of neighboring genes. Similar to the previous Surf1NEO−/− mice, the Surf1loxP−/− mice displayed mild reduction of COX activity in all tissues, but no lesion resembling LSCOX encephalopathy was ever observed. However, when the sensitivity to Ca2+-dependent excitotoxicity was tested in both Surf1loxP−/− brains and neuronal cell cultures, we observed a virtually complete protection from in vivo neurodegeneration induced by exposure to high doses of kainic acid, a glutamatergic epileptogenic agonist. In addition, Surf1loxP−/− mice showed a marked increase in longevity, compared to heterozygous or homozygous wild-type (wt) littermates. These data suggest a role for Surf1p in intracellular Ca2+ homeostasis and mitochondrial control of aging.
RESULTS
Generation of Surf1loxP−/− mice
The strategy used to disrupt the mouse Surf1 gene (NM_013677) is shown in Figure 1A. Briefly, a cDNA expressing the Escherichea coli neomycin phosphotransferase (NEO), therefore conferring neomycin-resistance, flanked by two loxP sequences (loxP-NEO-loxP cassette), was inserted into a unique AccIII restriction site of Surf1 exon 7. In order to confirm homologous recombination, Southern blot analysis was performed on both extremities of the Surf1loxP-NEO-loxP recombinant allele (Fig. 1B and C). Blastocyst injection of two recombinant clones gave rise to eleven Surf1loxP-NEO-loxP chimeric mice showing germline transmission. No Surf1loxP-NEO-loxP/loxP-NEO-loxP homozygous individuals were obtained by mating Surf1loxP-NEO-loxP/+ heterozygotes to each other, due to arrest of organogenesis at E8.5–9.5 (Supplementary Materials, Fig. 1 and Table 1). The NEO cassette and one of the two loxP sequences were then excised by mating Surf1loxP-NEO-loxP/+ heterozygous animals with animals constitutively expressing the cre recombinase. The resulting recombinant allele (Surf1loxP) contains a single loxP insertion (Fig. 1D), which causes the shift of the open reading frame (ORF) from nt 674 downstream of the mouse Surf1 cDNA, and the replacement of the codon encoding N225 into a TAA stop codon (N225X) (Fig. 1A). This mutation predicts the elimination of 81 amino acids on the Surf1p carboxy terminus. No cross-reacting material was detected by western blot immunoassay using polyclonal and monoclonal anti-Surf1 antibodies in several tissues of our Surf1loxP−/− mice (Fig. 1F), indicating that the truncated Surf1loxP protein is either unstable or fails to be translated, due to mRNA decay (Fig. 1E).
Clinical and biochemical phenotype
The percentages of approximately 800 Surf1loxP−/−, +/− and +/+ newborn animals followed a mendelian distribution (26, 51 and 23%, respectively; Supplementary Material, Fig. 2A). Newborn Surf1loxP−/− individuals were significantly smaller than their littermates, but this difference progressively disappeared after weaning (Supplementary Material, Fig. 2B). No difference in the clinical phenotype was observed between Surf1loxP−/− individuals and their +/− or +/+ littermates, including the appearance of neurological symptoms, abnormal reaction to stimuli, aberrant behavior, impaired cognitive abilities and reduced fertility. There was a small but significant reduction at the rotarod test (Supplementary Material, Fig. 2C), indicating mildly decreased motor skills in Surf1loxP−/− versus wt mice.
Histochemically, we observed decreased reaction to COX and increased reaction to succinate dehydrogenase (SDH) in Surf1loxP−/− skeletal muscle (Fig. 2A), similar to, but less severe than, that observed in SURF1 mutant patients. Finally, Surf1loxP−/− brains showed normal cytoarchitecture by thionine and GFAP stainings (data not shown).
Biochemically, there was no difference in the activities of MRC complexes I, II and III, whereas the COX activity in several tissues of Surf1loxP−/− individuals was 30–50% that of control littermates (Fig. 2B). Again, this reduction was much less marked than that observed in LSCOX patients (7). Surf1loxP+/− mice showed no biochemical difference compared to Surf1+/+ littermates, as expected for a recessive trait (data not shown).
Blood lactate was higher in Surf1loxP−/− mice than in wt littermates, indicating partial block in the aerobic utilization of pyruvate (Supplementary Material, Fig. 2D).
As exemplified in Fig. 2C, fully assembled COX was variably reduced in different tissues of Surf1loxP−/− mice, to an extent compatible with the levels of residual COX activity. Early assembly COX intermediates were also present in isolated mitochondria of Surf1loxP−/− mice, similar to, but much lesser than, what is found in SURF1 mutant patients.
Taken together, these data are remarkably similar to those previously reported for the surviving Surf1NEO−/− individuals (11).
Increased longevity of Surf1loxP−/− mice
In order to evaluate whether the lack of Surf1p could determine a late-onset phenotype, Surf1loxP−/− mice and control littermates were maintained under continuous observation in the same standard breeding conditions. No neurological or other clinical symptoms were ever seen in any individual. However, the lifespan was markedly different between the two groups. A total of 25/43 Surf1loxP−/− mice died during the observation period, against 30/48 deaths recorded in the control group. The median survival was 793 days for the Surf1loxP−/− group and 654 days for the control group, the latter being the standard median survival reported for laboratory mice (12). As shown in Figure 3A, the difference in the Kaplan–Meier survival probability, calculated by the logrank test, was highly significant between the two groups (P = 0.0002), irrespective of the gender (Fig. 3B and C).
Surf1loxP−/− mice are protected from Ca2+-related excitotoxic brain damage
In order to determine whether our Surf1loxP−/− mice were more susceptible to stress-induced neuronal damage, we used kainic acid, an epileptogenic glutamate agonist that has extensively been used to test neuronal response and survival to Ca2+-mediated excitotoxicity. We injected intra peritoneum (i.p.) a total of 145 three-month old mice (73 controls and 72 Surf1loxP−/−) with 30 mg/kg of kainic acid. The mortality rate, as well as the frequency, time lapse, severity and duration of the kainate-induced seizures were similar between the KO and control groups (Table 1), suggesting that the pharmacokinetics of the drug did not differ in the two groups. Following a standard protocol (13), only animals surviving the most severe level-5 seizure were further investigated (Table 1). Whole-brain pathological examination was carried out using the neurodegeneration-sensitive FluoroJadeB (FJB) fluorochrome (14) and thionine stainings. Strong FJB-positive neurons were detected in virtually all the glutamatergic areas of the control brains, including the Cornu Ammonis (CA) areas 1–4 of the hippocampus, the cerebral cortex, the amygdala and the thalamic and olfactory nuclei. The most severely affected areas were the CA1 and CA3 regions, which stained consistently FJB-positive in all of the 24 control mice. Other brain areas were differently affected in different wt animals. However, no FJB-positive cells were ever detected in all of the 23 Surf1loxP−/− brains (Fig. 4). Differential uptake of kainic acid by Surf1loxP−/− versus control neurons was unlikely, since there was no appreciable difference in the amount of glutamate ionotropic receptor subunits, immunovisualized using specific antibodies (Supplementary Material, Fig. 3). As exemplified in Fig. 5, severe neuronal loss in glutamatergic areas was observed in kainate-treated wt brains stained with thionine, while Surf1loxP−/− brains appeared consistently identical to untreated control brains. TUNEL-positive apoptotic nuclei were abundant in CA1–4 neurons (Fig. 5G–I) and in neurons from other glutamatergic regions of treated wt brains (data not shown). No TUNEL-positive neurons were ever detected in the same regions of Surf1loxP−/− brains.
. | +/+ . | −/− . |
---|---|---|
Injected animals | 73 | 72 |
Death during crisis | 13 | 12 |
Total L5a mice analyzed | 24 | 23 |
Analyzed after 1 day | 11 | 11 |
2 days | 4 | 3 |
4 days | 3 | 2 |
9 days | 2 | 2 |
60 days | 4 | 5 |
. | +/+ . | −/− . |
---|---|---|
Injected animals | 73 | 72 |
Death during crisis | 13 | 12 |
Total L5a mice analyzed | 24 | 23 |
Analyzed after 1 day | 11 | 11 |
2 days | 4 | 3 |
4 days | 3 | 2 |
9 days | 2 | 2 |
60 days | 4 | 5 |
aL5, level-5 seizures (13).
. | +/+ . | −/− . |
---|---|---|
Injected animals | 73 | 72 |
Death during crisis | 13 | 12 |
Total L5a mice analyzed | 24 | 23 |
Analyzed after 1 day | 11 | 11 |
2 days | 4 | 3 |
4 days | 3 | 2 |
9 days | 2 | 2 |
60 days | 4 | 5 |
. | +/+ . | −/− . |
---|---|---|
Injected animals | 73 | 72 |
Death during crisis | 13 | 12 |
Total L5a mice analyzed | 24 | 23 |
Analyzed after 1 day | 11 | 11 |
2 days | 4 | 3 |
4 days | 3 | 2 |
9 days | 2 | 2 |
60 days | 4 | 5 |
aL5, level-5 seizures (13).
Glutamate-induced cytosolic Ca2+ signals and consequent neuronal cell death are reduced in Surf1loxP −/− neurons
Kainic acid treatment mimics glutamate-induced excitotoxicity, which is characterized by delayed Ca2+ deregulation (DCD) and loss of mitochondrial potential (ΔΨm) (15,16). Glutamate stimulation leads to the activation of metabotropic (mGlu) and ionotropic (AMPA, kainate and NMDA-type) plasma-membrane receptors in neurons. To further investigate the impact of Surf1 ablation on these phenomena, we used 7-day-old cortical/hippocampal primary neuronal cell cultures obtained from Surf1loxP−/− mice and controls. Activation of the metabotropic receptor leads to inositol-triphosphate (IP3)-induced Ca2+ release from the endoplasmic reticulum (ER). Metabotropic Ca2+ responses were not detected in cultured neuronal cells, as confirmed by the use of the specific mGlu agonist APB (data not shown). Activation of ionotropic receptors causes Na+-mediated cell depolarization and subsequent Ca2+ influx through voltage-dependent Ca2+ channels, as well as direct Ca2+ influx through activation of the NMDA channel. To measure cytosolic Ca2+ signals, we loaded neurons with the low-affinity Ca2+ dye fura-FF. Neurons were then challenged with 10 or 100 µm glutamate for 30 min, and cytosolic [Ca2+] ([Ca2+]c) changes were measured in single cells. As shown in Figure 6A and B, the increase in [Ca2+]c induced by 10 µm glutamate was not significantly different in Surf1loxP−/− versus Surf1+/+ neurons (ΔF/F 0.10 ± 0.01 in Surf1loxP−/− versus 0.12 ± 0.01 in Surf1+/+, n > 100 for each group, P = 0.31), but it was much lower after exposure to 100 µm glutamate (ΔF/F = 0.26 ± 0.02 in Surf1loxP−/− versus 0.45 ± 0.03 in Surf1+/+, n>140 for each group, P < 10−5).
Sustained stimulation with high doses of glutamate induces deregulation of neuronal Ca2+ homeostasis, which manifests as a secondary, delayed and irreversible [Ca2+]c increase in the supramicromolar range, eventually leading to cell death (17). We used fura-FF to detect glutamate toxicity by calculating the number of cultured neurons showing the characteristic secondary Ca2+ increase during sustained (up to 30 min) glutamate stimulation. The percentage of neuronal death after stimulation with 10 µm glutamate was reduced in Surf1loxP−/− neurons as compared to Surf1+/+ neurons, and it remained significantly lower at 100 µm glutamate stimulation (% cell death at 10µm glutamate: Surf1+/+17,1 ± 5,0% versus Surf1loxP−/−6,7 ± 2,9%; at 100 µM glutamate Surf1+/+32,5 ± 5,0% versus Surf1loxP−/−16,9 ± 2,4% P < 10−2; Fig. 6C).
Reduced mitochondrial Ca2+ uptake is responsible for reduced Ca2+ influx in Surf1loxP−/− neurons
The extent of mitochondrial Ca2+ uptake, strategically located at plasma membrane Ca2+ entry sites, has been shown to regulate Ca2+ influx through different plasma membrane channels, such as capacitative or ligand-induced Ca2+ influx channels (17). Ca2+ buffering in the sub-plasma membrane space was shown to reduce Ca2+ feedback inhibition of capacitative Ca2+ influx channels as well as of diverse subunits of NMDA channels (18,19). In order to verify that the reduction of [Ca2+]c following glutamate stimulation in Surf1loxP−/− neurons was correlated to a modification of mitochondrial Ca2+ homeostasis, determining reduced Ca2+ influx, we measured mitochondrial [Ca2+] ([Ca2+]m) in intact and permeabilized neurons, using a mitochondrially targeted low affinity aequorin probe (mitAEQmut). Neurons were transfected with the mitAEQmut probe and [Ca2+]m was measured in Surf1+/+ and Surf1loxP−/− cell populations after stimulation with glutamate at low (10 µm) and high (100 µm) concentrations. Stimulation with 100 µm glutamate induced very high increase of [Ca2+]m, leading to immediate consumption of the probe. However, mitochondrial Ca2+ transients could be measured at a lower dose of glutamate (10 µm). In these conditions (Fig. 7A), the mitochondrial Ca2+ uptake was drastically reduced in Surf1loxP−/− versus Surf1+/+ cells. Maximum [Ca2+]m was 28.95 ± 2.50 mm in Surf1loxP−/− neurons (n = 24) versus 50.95 ± 3.36 mm in Surf1+/+ neurons (n = 21), P < 10−5.
In order to verify that the reduction of mitochondrial Ca2+ uptake in Surf1loxP−/− cells is due to lack of Surf1p, rather than to reduced cytosolic Ca2+ response consequent to reduced Ca2+ influx through the plasma membrane, neurons expressing the mitAEQmut probe were treated with low-dose digitonin, which selectively permeabilizes the plasma membrane, and endogenous cytosolic Ca2+ was washed out by perfusion with Ca2+-free intracellular buffer. Mitochondrial Ca2+ uptake was then triggered by the addition of 1 µm Ca2+ to the buffer. Again, the velocity of mitochondrial Ca2+ uptake was drastically reduced in Surf1loxP−/− cells (Surf1loxP−/− 0.34 ± 0.07 µm/s n = 23 versus Surf1+/+ 1.10 ± 0.12 µm/s n = 26, P < 10−7) (Fig. 7B). This result indicates that the reduction of mitochondrial Ca2+ uptake is intrinsic to Surf1loxP−/− mitochondria. Since the inhibition of mitochondrial Ca2+ uptake has been shown to augment the feedback inhibition of plasma-membrane Ca2+ channels in neurons, the impairment of mitochondrial Ca2+ buffering in Surf1loxP−/− neurons might ultimately prevent them from Ca2+ overload, which could explain their refractoriness to DCD and cell death.
Surf1 deficiency impairs mitochondrial Ca2+ uptake without changing the mitochondrial structure and membrane potential
The mitochondrial membrane potential (ΔΨm) drives the mitochondrial Ca2+ uptake by sustaining the activity of the mitochondrial Ca2+ uniporter (20,21). Therefore, reduction of ΔΨm can decrease the driving force of Ca2+ entry and the activity of the Ca2+ uniporter machinery in the inner mitochondrial membrane, thus determining reduced Ca2+ uptake into the organelle. Since Surf1p plays a role in the formation of COX, and the latter is in turn involved in the maintenance of mitochondrial transmembrane H+ gradient and ΔΨm, the reduction of Ca2+ uptake observed in Surf1loxP−/− neurons could depend on a reduction of ΔΨm. We then measured the ΔΨm in Surf1+/+ and Surf1loxP−/− neuronal populations by steady-state loading of neurons with the ΔΨm-sensitive dye teramethyl-rhodamine-methylester (TMRM), followed by fluorimetric measurement of the intensity of the dye in mitochondria. Our results showed no difference in the steady-state distribution of TMRM in Surf1loxP−/− versus Surf1+/+ neurons (31.19 ± 0.23 in Surf1+/+ versus 32.12 ± 0.29 in Surf1loxP−/−, fluorescence intensity in arbitrary units, Fig. 8A), demonstrating that the mild reduction observed in COX activity in various tissues of Surf1loxP−/− animals failed to result in significant changes of ΔΨm.
We previously showed that mitochondrial fragmentation in epithelial cells leads to an average reduction of mitochondrial Ca2+ load upon IP3-induced Ca2+ release (22). Therefore, we next investigated whether Surf1 deficiency can change the shape of the mitochondrial network. Neurons were transfected with mitochondrially targeted DsRed probe (mtDsRed) and the structure of mitochondria in 300 Surf1loxP−/− and 300 Surf1+/+ individual cells was imaged by digital microscopy (22). As exemplified in Fig. 8B, the overall arrangement of the mitochondrial network did not appear to be disturbed in Surf1loxP−/− neurons with respect to controls.
On the basis of these results, we concluded that Surf1p deficiency in mice does not lead to reduced ΔΨm or altered mitochondrial structure, suggesting that the effect of Surf1p on mitochondrial Ca2+ uptake could be independent from its role on COX assembly and maintenance of ΔΨm.
To further confirm this hypothesis, we analyzed the effects of SURF1 overexpression on global cellular and mitochondrial Ca2+ signals in HeLa cells. HeLa cells were co-transfected with a vector expressing human SURF1 (hSURF1) (6) and with mitAEQmut or its non-targeted cytosolic variant (cytAEQ). Cells were then challenged with histamine (100 µm), which induces IP3-dependent Ca2+ release from the ER. The resulting rise in [Ca2+]c stimulates the mitochondrial Ca2+ uptake at sites located in proximity of the ER Ca2+ release sites. As shown on Figure 8C, the cytosolic Ca2+ responses remained unaltered in cells overexpressing SURF1 (peak [Ca2+]c 3.06 ± 0.08 mm in controls versus 2.86 ± 0.09 mm in hSURF1 overexpressing cells); however, mitochondrial Ca2+ uptake was significantly increased by SURF1 overexpression (peak [Ca2+]m 78.07 ± 5.32 mm in controls versus 94.51 ± 6.04 mm in overexpressing cells, P < 0.05, Fig. 8D). This increase was not due to variations of the ΔΨm, as measured by TMRM uptake (data not shown).
DISCUSSION
High embryonic lethality was a major feature of a mouse knockout model for Surf1p, an accessory assembly factor of COX, based on the replacement of a region of several kb in the midportion of the Surf1 gene with a NEO cassette (Surf1NEO). The new mouse model reported here carries a recombinant null allele, consisting in the insertion of the 35 bp loxP sequence within exon 7 of the murine Surf1 gene (Surf1loxP). The restoration of mendelian distribution of the genotypes in newborn recombinant Surf1loxP mice indicates that the embryonic lethality observed in the previous Surf1NEO model was not a consequence of the ablation of the Surf1 gene itself, but was rather due to the presence of the NEO cassette or to the elimination of regulatory elements contained in the deleted region of the Surf1 gene. This conclusion is further supported by the observation that the Surf1loxP-NEO-loxP allele, from which the Surf1loxP allele is derived, was again associated with 100% embryonic lethality, when present in homozygosity. Several recent reports (reviewed in 23) have shown that the maintenance of a NEO cassette in recombinant alleles can be associated with a number of unpredictable effects, including the creation of hypomorphic alleles, altered gene expression and embryonic lethality, mostly due to the promotion of illegitimate splicing of either the targeted gene or neighboring genes. The latter phenomenon was likely to occur in the surfeit genomic region, which is packed with six housekeeping genes, some of which share common regulatory elements (24).
Although SURF1p is a ubiquitously expressed mitochondrial protein, and its ablation leads to an early onset, invariably fatal encephalopathy in humans, no clinical disease phenotype could be observed in our Surf1loxP−/− mice at any age. This situation is similar to that reported for the Surf1NEO−/− animals that survived embryonic selection (11). The absence of neurological and extra-neurological abnormalities was associated with a biochemical phenotype that showed a specific and generalized defect of COX activity, however less severe than that observed in SURF1 mutant patients (7). The COX defect was likely too mild to cause brain failure or impairment, but was possibly sufficient to determine the modest functional and morphological alterations found in skeletal muscle of adult Surf1loxP−/− mice.
The mild biochemical and the virtually absent clinical phenotypes of Surf1loxP−/− mice suggest that, in spite of the ubiquitous expression and high evolutionary conservation of Surf1p, the function of this protein in COX assembly is either ancillary or redundant, that is, it can partly be overtaken by other unknown factors. To some extent, this may well be true also in SURF1-less mutant patients, in whom fully assembled COX is diminished, but not absent (6). The different severity of the phenotype associated with the absence of Surf1p may then depend on the efficiency and efficacy of compensatory genetic or epigenetic mechanisms acting in different organisms, notably fungi, mice and men.
In an attempt to determine whether the reduced COX activity found in Surf1loxP−/− animals could make their brain more sensitive to energy stress, we used an excitotoxic glutamate agonist, kainic acid, which triggers epileptic seizures in experimental animals. Kainic acid acts on the AMPA-kainate glutamate receptors present on the neuronal cell membrane (25). When activated, the Na+-channel component of these receptors opens up, thus determining Na+ influx and membrane depolarization. As a consequence, massive influx of Ca2+ occurs through the NMDA receptors and the VOCC channels of the plasma membrane (26). In addition to determining the epileptic discharge, the marked rise of [Ca2+]c can promote a cascade of secondary effects that may ultimately lead to cell death (27).
These experiments were aimed at challenging the OXPHOS reserve of neuronal cells provoked by epileptic discharge. The same approach has been used in the recent past to precipitate catastrophic neurodegeneration in the MItochondrial Late-Onset Neurodegeneration (MILON) mouse, a conditional TFAM knockout model, characterized by loss of mtDNA in neurons of the frontal cortex (28). To our surprise, however, the elicitation of level-5 seizures failed to cause any neuronal degeneration and neuronal loss in Surf1loxP−/− mice, while these lesions were consistently observed in glutamatergic areas of control brains. Since kainate-associated neurodegeneration is largely dependent on perturbation of Ca2+ homeostasis, we investigated the cytosolic and mitochondrial Ca2+ fluxes in primary neuronal cell cultures from Surf1loxP−/− and control mice. The results of these experiments can be summarized as follows.
First, the ablation of Surf1 drastically reduces the glutamate-induced increase of [Ca2+] in both cytosolic and mitochondrial compartments.
Second, the reduction of mitochondrial Ca2+ uptake is directly consequent to the absence of Surf1p.
Third, Surf1loxP−/− cultured neurons are much more resistant to glutamate toxicity than control neurons. How could this effect be linked to the observed reduction of mitochondrial Ca2+ uptake? One possibility is that reduced buffering capacity by Surf1loxP−/− mitochondria can determine the saturation of the Ca2+ microdomains in the contact sites between mitochondria and the plasma membrane or the ER. This effect could in turn promote the feedback closure of the Ca2+ channels in the above structures, thus inhibiting the [Ca2+]c transient rise (18). Although speculative, this hypothesis can offer a mechanistic explanation for the neuroprotection observed in vivo. A second possibility is that the ablation of Surf1 may alter the expression of nuclear genes encoding proteins engaged in Ca2+ homeostasis. As a preliminary result, quantitative PCR analysis failed to show different expression of the Ca2+–Na+ plasma membrane exchanger (NCX1) in Surf1loxP−/− versus wt brains (data not shown). More work is needed to expand this analysis to other Ca2+-related genes.
Fourth and last, the reduction of the mitochondrial [Ca2+] uptake seems not to be dependent from a decrease of the ΔΨm, as a consequence of partial defect of COX activity and mitochondrial respiration. This conclusion, which was also supported by the results of hSURF1 overexpression in HeLa cells, suggests that Surf1p could play a direct role on mitochondrial Ca2+ handling, partially or completely independent from its function as a COX assembly factor. More work is necessary to test this hypothesis, but it is interesting to observe that an increasing number of mitochondrial proteins have been established to carry out multiple functions. A well known example is cytochrome c, which acts as both a redox electron shuttle of the MRC, and as an apoptogenic messenger (29). Likewise, the apoptosis-inducing factor (AIF), a flavoprotein closely associated with the mitochondrial inner membrane, has been implicated as both a cell death-promoting molecule and a regulator of activity and protein expression of MRC complex I (30). Lastly, p66Shc is an electron-transfer redox enzyme of the mitochondrial intermembrane space, which controls the production of reactive oxygen species (ROS), and regulates Ca2+ transport by acting on plasmamembrane Ca2+-ATPases. These two independent activities can ultimately converge and synergize in mediating p66Shc-dependent apoptosis (31). Targeted disruption of p66Shc is associated with prolonged lifespan in mice, possibly related to its role on the control of ROS production.
Similar to the p66Shc−/−, our Surf1loxP−/− mouse model displays significantly increased longevity. Of note, increased longevity was also observed in a CNS-restricted conditional Surf1 knockdown (KD) model in Drosophila melanogaster, whereas the corresponding constitutive model was embryonic lethal (32). We do not have an explanation for this observation, which is in striking contrast with the early onset, invariably fatal phenotype associated with the loss of SURF1p function in humans. It is possible that the COX defect in Surf1loxP−/− mice is not severe enough to confer selective disadvantage in a ‘protected’ environment such as an animal care facility. These data are nevertheless surprising, considering that COX deficiency associated with Surf1p disruption should in principle determine an increase of ROS (33), which are proposed to play a major role in the aging process (34). An attractive possibility is that the effect on longevity is due to the role of Surf1p on mitochondrial Ca2+ uptake and cellular Ca2+ homeostasis. This effect is likely to be hidden in organisms, such as S. cerevisiae and Homo sapiens, in which lack of Surf1p leads to severe impairment of COX assembly and faulty OXPHOS, whereas it would be unmasked in other organisms, such as Mus musculus and the conditional D. melanogaster KD model, characterized by less severe impairment of COX assembly and OXPHOS phenotype.
MATERIALS AND METHODS
Creation of Surf1loxP-NEO-loxp+/− and Surf1loxP−/− recombinant mice
The list of all primers is provided in the Supplementary Material, Table 2.
For the construction of the Surf1 recombinant alleles, we used a 10 kb HindIII-EcoRI fragment containing the entire Surf1 and Surf2 genes, and part of Surf4 and Surf3 genes, cloned in BlueScript SK. A fragment of approximately 1.2 kb, composed of the gene encoding the neomycin phosphotransferase (NEO cassette) flanked by two 35-mer identical loxP sites (loxP-NEO-loxP), was inserted in the unique AccIII site contained in exon 7 of the Surf1 murine gene corresponding to nt 670 of the Surf1 cDNA. The DNA vector was verified on both strands by automated sequence analysis using the big-dye terminator kit and protocol (Applied Biosystems), on a 3100 ABI apparatus.
Gene targeting by electroporation of the HindIII-linearized vector into AB1 ES cells, derived from 129/SvEvBr ©≠/Hprt-bm2 mouse substrain (a kind gift from Alan Bradley), and generation of chimeras, were performed as described (35). Three hundred colonies that survived selection with the neomycin analogue drug G-418 (200 µg/ml) were screened for homologous recombination by PCR and Southern blot analyses.
For diagnostic PCR analysis, a 2 kb DNA fragment was amplified using a forward primer corresponding to a region inside the NEO cassette and a reverse primer corresponding to a sequence within the Surf3 gene located outside the recombinant region. For diagnostic Southern blot analysis (Fig. 1B) on the 5′ end of the recombinant region, 10 µg of ES genomic DNA was digested with EcoRI, run through a 0.8% agarose gel in 1X TE buffer, blotted on a nitrocellulose filter and hybridized with a 0.7 kb PCR fragment (probe P1 in Fig. 1) radiolabeled with [α32P]-dCTP (NEN, Boston, MA, USA) using the ‘Ready-to-go’ random priming kit (Amersham, Piscataway, NJ, USA). P1 corresponds to a region of the Surf5 gene outside but contiguous to the recombinant region. The wt Surf1 allele corresponds to a 19 kb hybridization band, whereas the Surf1loxP-NEO-loxP recombinant allele corresponds to 13 and 7 kb bands due to the presence of an EcoRI site within the NEO cassette. For diagnostic Southern blot analysis on the 3′ end, DNA was digested with EagI and HindIII, separated by electrophoresis and blotted as above. Hybridization was then carried using a 0.8 fragment corresponding to a sequence on the Surf4 gene outside but contiguous to the recombinant region (probe P2 in Fig. 1). The wt allele corresponds to a hybridization band of 9.2 kb, whereas the Surf1loxP-NEO-loxP recombinant allele corresponds to a band of 7.5 kb, again due to the presence of an extra EagI site within the NEO cassette.
Two of the five ES clones that showed homologous recombination were injected in B6D2F1©≠C57/Bl6J blastocysts (36). Chimeric pups were identified by the presence of agouti hair and, on maturity, mated with B6D2F1 (C57/Bl6J_DBA2) females to check for the contribution of the ES cells to the germline.
For the creation of Surf1loxP animals, Surf1loxP-NEO-loxP+/− mice of mixed BDF1 genetic background were mated to cre mice. Compound heterozygotes (Surf1loxP/+;cre/+) were identified by PCR and backcrossed to each other.
For genotyping, 250 ng genomic DNA extracted from tail tips was PCR amplified in 50 µl of 1×MgCl2-PCR buffer (Applied Biosystems), 200 µm dNTPs, 0.6 µm each dNTPs and 0.03 U/µl Taq Polymerase (Applied Biosystems), 5% DMSO. After an initial denaturation at 94°C for 2 min, each of the 35 PCR cycles was as follows: 94°C for 30 s, 58°C for 60 s, 72°C for 90 s. Final extension was at 72°C for 5 min. For the Surf1loxP-NEO-loxP allele, we used a single forward primer (LNL-FW) and two distinct reverse primers, one internal to the NEO cassette (LNL-RV1) specific to the recombinant allele, and another corresponding to a Surf1 region (LNL-RV2) specific to the wt allele. The recombinant allele generates a PCR fragment of 583 bp, whereas the wt allele generates a PCR fragment of 458 bp. For the Surf1loxP allele, we used the forward primer LP-FW, corresponding to a region of Surf1 exon 6; and the reverse primer LP-RV, corresponding to a region of Surf1 exon 7. The PCR produces a 305 bp wt DNA fragment and a 340 bp recombinant fragment). The cre transgene was detected by using the forward primers CRE-FW and the reverse primer CRE-RV (PCR fragment 590 bp).
Animal studies were approved by the animal welfare Ethics Committee of the National Neurological Institute in accordance with the Institutional Animal Care and Use Committee guidelines. Standard food and water were given ad libitum.
RT-PCR analysis
To evaluate the presence of recombinant Surf1loxP transcript, total RNA was extracted from brain, liver, muscle and heart of four Surf1loxP−/− and four Surf1+/+ adult animals, using the RNeasy lipid tissue kit (QIAGEN Sciences, MD, USA) following the manufacturer's protocol. Total RNA was used as a template for reverse transcription, using the ‘cDNA cycle’ kit and protocol (Invitrogen, Carlsbad, CA, USA). Total cDNA was purified and resuspended in a final volume of 20 µl and used for PCR amplification of individual cDNA fragments corresponding to GAPDH and Surf1 genes; 3 µl were then used in each 50 µl PCR reaction containing 1×MgCl2-PCR buffer, 200 µm dNTPs, 5% DMSO, 0.6 µm of each primer and 0.03 U/µl of Taq-Gold polymerase (Invitrogen, Carlsbad, CA, USA). GAPDH was detected using primers GAPDH-FW and GAPDH-RV. The wt and Surf1loxP transcripts were PCR-diagnosed using exonic primers LP-FW and LP-RV (Fig. 1E).
Western blot analysis
Western blot analysis was performed on electroblotted denaturing sodium-dodecyl sulphate polyacrylamide gel electrophoresis (SDS–PAGE), and two-dimension blue native electrophoresis (2D-BNE), as described previously (6). Approximately 100 µg non-collagenous protein was used for each sample in SDS-PAGE and 20 µg of isolated mitochondria in 2D-BNE. Chemiluminescence-based immunostaining (ECL kit, Amersham) was performed using the following antibodies: polyclonal antibody AS182–196 (6) raised against amino acid sequence 82–96 of mouse Surf1p, which is at the N-terminal of the truncated protein predicted by the Surf1loxP allele; monoclonal antibodies against subunits COX I and COX IV (Molecular Probes, Eugene, OR, USA) and Surf1p (Mitosciences LLC, Eugene, OR, USA).
Morphological analysis
For light microscopy, samples from different organs were frozen in liquid-nitrogen-cooled isopentane. Standard histological and histochemical techniques for the detection of mitochondrial alterations and muscle fiber distribution were performed on serial cryostat cross sections as previously described (37).
Biochemical analysis
Biochemical assays of individual respiratory complexes were carried out on tissue homogenates (38). Specific activities of each complex were normalized to that of CS, an indicator of the number of mitochondria. Blood lactate was measured using the ‘Lactate reagent’ kit and protocol (Sigma, St Louis, MO, USA).
Rotarod test
Motor performance tests were given to 13 wt mice and to 13 Surf1loxP−/− mice. All animals were 3 months old. The rotating rod test was performed on a Rotarod apparatus for mice (Ugo Basile) (39).
Kainate-induced seizures and brain analysis
Kainic acid was dissolved in isotonic saline (pH 7) with a drop of 1 m NaOH and administered i.p. Mice were monitored continuously for at least 3 h after injection to determine the onset and level of seizures according to Sperk et al., (13). For histological analysis of the brain, animals were treated with a lethal injection of 4% chloral hydrate before intra-cardiac perfusion with 4% paraformaldehyde in PBS. Brains were rapidly dissected and post-fixed overnight in the same solution. Serial rostro-caudal, 50 µm thick coronal brain sections were obtained using a Vibratome (Leica) and collected in 0.1 m PB at pH 7.2 with 0.01% NaN3. One every 12 sections was labeled with FJB (Histo-Chem, Jefferson, AR) according to manufacturer's instructions. Adjacent sections were stained with 0.1% thionine. Next adjacent sections were labeled by terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) (Apoptag in situ Apoptosis Detection kit; Intergen, Purchase, NY) according to the manufacturer's instructions. Fluorescent images were acquired on a confocal microscope (Radiance 2100 confocal microscope, Bio-Rad, Hercules, CA, USA) using an FITC filter and optical photographs were acquired using a Nikon Eclipse E400 microscope and a Nikon DS-U1 digital camera.
Cortical/hippocampal primary neuronal cell cultures
Cortical/hippocampal neurons were prepared from 1- to 3-day-old newborn mice, according to Pasti et al. (40), and neurons were resuspended in NeurobasalA Medium (Gibco) with supplement B-27 (Gibco), GlutaMax (Gibco) and penicilline/streptomycine (Gibco) rigorously at 37°C and plated onto glass coverslips, coated with poly-D-lysine (Sigma).
Dynamic in vivo [Ca2+] measurements with targeted aequorin probes
The construction and use of luminescent Ca2+ sensitive aequorin probes were previously described (41). All aequorin measurements were carried out in KRB containing 1 mm CaCl2 (KRB/Ca2+, Krebs-Ringer modified Buffer: 135 mm NaCl, 5 mm KCl, 1 mm MgSO4, 0.4 mm K2HPO4, 1 mm CaCl2, 15 mm glucose, 20 mm HEPES, pH 7.4). For HeLa cells KRB contained 5 mm glucose. Experiments in permeabilized neurons were performed as previously described for HeLa cells (42), except that 25 µm digitonin was used, in order to preserve mitochondrial integrity.
Imaging procedures
Cortical cultures were loaded for 20 min with 3 µm fura-2FF/AM (or fura-2) (Teflabs, Austin, TX, USA) at 37°C in the cell culture medium (Kd of fura-2FF or fura-2 for Ca2+ is 5 and 55 µm, respectively; 43). Images were acquired on an epifluorescence inverted microscope Axiovert 200 (Zeiss, Germany) equipped with a 40× fluorite objective. [Ca2+]c was monitored in single cells using two excitation light wavelengths, at 340 nm and 380 nm (Sutter Instrument Co., CA, USA). Emitted fluorescence light was selected by a 505–530 nm filter. Images were acquired by CCD camera (Roper Scientific, USA). All imaging data were collected and analyzed using the Metafluor 6.1 software. For the visualization of the mitochondrial network, Z-series of images of neurons were 3D deconvolved and reconstructed using a custom-made software (44).
Statistical analysis
Two-tailed, unpaired, unequal variance Student's t-test was used for statistical analysis. Survival probability was calculated using the Kaplan–Meier and log-rank tests.
SUPPLEMENTARY MATERIAL
Supplementary Material is available at HMG Online.
ACKNOWLEDGEMENTS
The authors are grateful to Barbara Geehan for revising the manuscript. They are also grateful to Chiara Falcone for assistance on statistical analysis and Mike Fried for the generous gift of a plasmid containing the mouse surfeit gene cluster.
This study was supported by Fondazione Telethon, Italy (Grant no. GGP030039), Fondazione Pierfranco e Luisa Mariani, MITOCIRCLE and EUMITOCOMBAT network grants from the European Union Framework Program 6. The work of R.R. was supported by grants from Telethon, Italy, the Italian Association for Cancer Research (AIRC), the Italian University Ministry (PRIN, FIRB and local research grants), the Emilia-Romagna PRRIITT program, the Ferrara Objective 2 funds and the Italian Space Agency (ASI).
Conflict of Interest statement. None declared.
REFERENCES
Author notes
These authors contributed equally to the work.
Present address: INSERM U807, University Paris 5, Faculty of Medicine Necker-Enfants Malades, Paris, France.