Mitochondrial Respiratory Deficiencies Signal Up-regulation of Genes for Heat Shock Proteins* □ S

The consequences of mitochondrial dysfunction are not limited to the development of oxidative stress or initiation of apoptosis but can result in the establishment of stress tolerance. Using maize mitochondrial mutants, we show that permanent mitochondrial deficiencies trigger novel Ca 2 (cid:1) -independent signaling pathways, leading to constitutive expression of genes for molecular chaperones, heat shock proteins (HSPs) of different classes. The signaling to activate hsp genes appears to originate from a reduced mitochondrial transmembrane potential. Upon depolarization of mitochondrial membranes in transient assays, gene induction for mitochondrial HSPs occurred more rapidly than that for cytosolic HSPs. We also demon-strate that in the nematode Caenorhabditis elegans transcription of hsp genes can be induced by RNA interference of nuclear respiratory genes. In both organisms, activation of hsp genes in response to mitochondrial impairment is distinct from their responses to heat shock and is not associated with oxidative stress. Thus, mitochondria-to-nucleus signaling to express a hsp gene network is apparently a widespread retrograde mechanism to facilitate cell defense and survival. In if

In response to transient stress, impairment of mitochondrial function is a key event (1) that leads to overproduction of reactive oxygen species (ROS) 1 and, if oxidative damage is severe, to execution of programmed cell death. In contrast, mitochondrial mutants acclimated to permanent respiratory deficiencies do not develop oxidative stress and have enhanced tolerance to apoptotic stimuli and environmental stresses (2)(3)(4). It has been shown that osteosarcoma 0 cell lines devoid of mitochondrial DNA (mtDNA) are more resistant to staurosporine-induced apoptosis than are the parental ϩ lines (2). We found that maize mutants with mtDNA deletions did not undergo oxidative stress despite severe respiratory deficiencies (3). Moreover, similar mitochondrial mutants in tobacco appeared to develop higher tolerance to biotic (viral infection) and abiotic (ozone) stresses (4).
We hypothesized that the enhanced stress resistance of mitochondrial mutants might be contributed by the up-regulation of hsp genes. It was previously demonstrated that cytosolic HSP70 can preserve mitochondrial capacity to maintain membrane potential under oxidative conditions (5). Cytosolic HSP70 can also prevent mitochondria from inducing both caspase-dependent and -independent pathways of programmed cell death (6,7). One of the ␣-crystallin type small HSPs (sHSPs) appears to block cytochrome c release from mitochondria (8). Thus, some HSPs exert anti-apoptotic effects in addition to their well documented functions (9) in preventing aggregation of non-native proteins (sHSPs) and promoting ATPdependent renaturation (HSP70). In plants, along with diverse sHSPs (10), another chaperone, ClpB/HSP101, is crucial for recovery after severe damage to cellular proteins (11). Its yeast homolog, HSP104, resolves cytosolic aggregates of denatured proteins (12) in a functional complex with SSA1 (HSP70) and Ydj1 (HSP40).
To test whether permanent mitochondrial respiratory defects affect expression of hsp genes, we use two experimental models: (i) maize respiratory-deficient NCS (non-chromosomal stripe; reviewed in Ref. 13) mutants with defined deletions in mtDNA (14 -18), and (ii) the nematode Caenorhabditis elegans with loss of function of different nuclear respiratory genes mediated by RNA interference. In both cases, we show that in response to mitochondrial impairment activation of a number of hsp genes occurs, including heat-inducible, constitutive, and cryptic ones. We present evidence that the primary signaling may originate from a reduction in mitochondrial transmembrane potential and does not involve either Ca 2ϩ -or ROS-dependent pathways. Altogether, these features characterize the mitochondria-dependent expression of hsp genes as a novel physiologically important process of retrograde regulation.

EXPERIMENTAL PROCEDURES
Plant Material-Maize mutants carrying maternally inherited deletions in mtDNA were analyzed. In NCS2, the nad4 gene is partially deleted and mutant plants are deficient in the NADH ubiquinone oxidoreductase, the respiratory complex I (14). The NCS4 deletion removes the 5Ј-end of the mitochondrial gene encoding the RPS3 ribosomal protein and is associated with reduced levels of mitochondrial translation (15,16). Deletion of the 5Ј-end of the cox2 gene in NCS6 mutants causes a deficiency in cytochrome c oxidase (17,18), the respiratory complex IV. Mutant plants are heteroplasmic for the mutations, so plants with the most defective phenotypes (with very high levels of mutant mtDNA) (13) were used for analysis. Such highly mutant individuals are rare, and sufficient amounts of material for analysis could be collected from mature plants only. Control plants with normal phenotype were chosen within the same or closely related NCS families (3).
Heat shock experiments were performed with seedlings, a developmental stage at which the temperature of the whole maize plant can be * This work was supported by grants from the National Science Foundation and the Illinois-Missouri Biotechnology Alliance (to K. J. N. and T. E. E.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. reliably controlled. 4-day-old maize seedlings (genotypes A619 and B37N) were incubated at 42°C as described previously (3), and the roots were used for further analysis.
Treatments with inhibitors were performed using 4-day-old normal maize seedlings (genotype B37N) or seedlings with T type cytoplasmic male sterility (B37T) according to Ref. 3. Methomyl and chloramphenicol were applied at final concentrations of 7.5 mM and 50 g/ml, respectively. After 24-h incubations, all seedlings looked normal and retained undamaged mitochondria, because no loss of the outer mitochondrial membrane or mitochondrial matrix content was detected by Western analysis (data not shown).
Identification of Maize hsp Coding Sequences-Protein sequences were used to query plant genomic and Expressed Sequence Tag data bases with the tBLASTn algorithm. Arabidopsis homologs of the yeast SSA1-type HSP70 were identified and then used as queries to search the maize EST data bases (Maize Genome Data base, www.zmdb.iastate. edu/cgi-bin/main/ZMDB, and the TIGR Eukaryotic Gene Ortholog (EGO) data base, www.tigr.org/tdb/tgi/ego/orth_search.shtml). The maize orthologs of Arabidopsis mitochondrial HSP70s (19) were found by homology in the N-terminal protein sequences. The identification of heat-inducible and HS non-responsive maize orthologs of HSP70s was confirmed by gene-specific hybridization of maize total RNA from heatshocked seedlings. In rice and maize, mitochondrial HSP22s were searched for in the TIGR EGO data base and in the grass genome data base GRAMENE (www.gramene.org/db/searches/blast). The potential mitochondrial targeting of the candidate precursors was evaluated using the Predotar program (www.inra.fr/predotar), and phylogenetic analysis of mitochondrial HSP22s was performed by CLUSTALW alignment (clustalw.genome.ad.jp) of the predicted full-length amino acid sequences.
Northern Analysis of Maize Transcripts-Isolation of total RNAs from maize tissues and RNA blot hybridizations was performed as described previously (3). Gene-specific maize riboprobes were prepared from RT-PCR-amplified total maize (B73N) RNA, using 3Ј-or 5Ј-untranslated region-specific primers (Supplemental Table I). To obtain the hsp22B riboprobe, PCR amplification from B73N genomic DNA was used. PCR fragments were cloned into the pGEMT-Easy vector (Promega) and verified by sequencing.
To compare the methomyl and chloramphenicol effects, equal RNA amounts were loaded on the same gel, and for each hybridization probe the same x-ray film was used for quantification of hybridization signals (program Image Gauge 3.3; Fuji). Three independent experiments were performed.
RNAi Silencing in C. elegans-C. elegans (N2) RNA interference (RNAi) by feeding was performed basically as described by Kamath et al. (23). Escherichia coli strain HT115(DE3) was transformed with L4440-based constructs containing PCR fragments of four genes derived from genomic DNA (primers listed in Supplemental Table II). Adult worms were placed on nematode growth medium plates with recombinant HT115(DE3) clones induced by 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside overnight. The worms were allowed to lay eggs and were removed after 12 h. Progeny were grown for 72 h at 23°C and harvested for RNA extraction by the guanidinium thiocyanate/phenolchloroform method (24). Developmental stages were scored 10 h before harvesting. Northern hybridizations were performed as described previously (3). PCR fragments generated from N2 genomic DNA with primers listed in Supplemental Table II were used as hybridization probes.

Constitutive Expression of hsp Genes in Maize Mitochondrial
Mutants-Northern analysis showed that hsp101 and genes for cytosolic sHSPs, classes I and II (shsp-I and shsp-II), whose products are confirmed to have cytoprotective functions (11,25), are constitutively expressed in all NCS mutants under normal growth conditions (Fig. 1A). Higher levels of all transcripts were found in the translation-impaired NCS4 (rps3) mutant than in the complex I-deficient NCS2 (nad4) and complex IV-deficient NCS6 (cox2) mutants. For expression analysis of genes for cytosolic and mitochondrial HSP70s, we performed a maize Expressed Sequence Tag data base search and then identified the candidate cDNAs (Supplemental Table I) through protein sequence alignments with Arabidopsis orthologs (19) and response to heat shock (Fig. 1B). Two tested ssa1-homologous genes for cytosolic HSP70s were both highly expressed in all NCS mutants (Fig. 1A), although only one of them was found to be HS-inducible (Fig. 1B). cDNAs for mitochondrial HSP70s were identified via a signature motif in the deduced N-terminal sequences (19). Both constitutive (mthsc70) and HS-inducible (mthsp70) genes for mitochondrial HSP70s were found to be expressed in NCS mutants at significantly higher levels than in normal plants (Fig. 1, A and B). Transcript levels for CPN60, a different constitutive mitochondrial chaperonin (26), were also increased in all NCS mutants (Fig. 1A). In addition to the earlier studied hsp22A gene for the mitochondrial sHSP (22), we identified another hsp22-like expressed sequence tag (hsp22B; Supplemental Table I) that had orthologs in rice and wheat, but not in Arabidopsis ( Fig. 2A). Both genes were dramatically expressed in NCS mutants (Fig.  1A), although hsp22B did not respond to heat shock ( one of the multiple maize heat shock transcription factor genes, hsfb (Fig. 1B), that had been shown earlier to be HS-inducible (27). None of the transcripts hybridizing to this probe was found at elevated levels in any NCS mutant ( Fig. 1A and Table  I). Taken together, these results indicate that the expression of hsp genes in NCS mutants differs from a typical heat shock response (e.g. Ref. 28).
Role of the Mitochondrial Transmembrane Potential in Activation of hsp Genes-What consequences of mitochondrial mutations could initiate a signaling pathway leading to the transcriptional activation of hsp genes? We considered two effects of the loss of mitochondrial genes: (i) decrease in mitochondrial transmembrane potential ⌬ m (2,29), and (ii) failure to assemble respiratory complexes without mitochondrially encoded components, resulting in accumulation and/or proteolysis of unincorporated subunits (30). These processes can be individually induced in a transient mode by specific inhibitors that cause depolarization of the mitochondrial membrane or arrest of mitochondrial translation.
For a membrane discharge that would be highly specific to mitochondria, we chose methomyl treatment of maize seedlings with T-type cytoplasmic male sterility (31). CMS-T mtDNA encodes an aberrant membrane protein, T-URF13, that can specifically interact with methomyl to form small hydrophilic pores in the inner mitochondrial membrane (32). Indeed, strong induction of the hsp genes with two types of expression patterns was observed in response to methomyl. Induction of transcripts for mitochondrial HSPs occurred rapidly and was mostly complete within 12 h (Fig. 3A, left). For genes encoding cytosolic HSPs, there was a prolonged lag phase prior to an exponential accumulation of transcripts between 18 and 24 h of treatment (Fig. 3B, left). These results suggest that depolarization of the mitochondrial membrane can trigger divergent signaling pathways for two groups of hsp genes.
As we have shown earlier (3), neither increases in protein oxidation nor elevated levels of marker antioxidant enzymes were detected in maize NCS tissues. Absence of endogenous oxidative stress was further corroborated with the lack of induction of a few highly ROS-responsive genes, including the apx1 for cytosolic ascorbate peroxidase (33). Because no apx1 induction was detected upon methomyl treatment (Fig. 3A,  left), the activation of hsp genes was not associated with transient oxidative stress.
Chloramphenicol treatment to inhibit mitochondrial translation did not result in any detectable induction of transcripts for cytosolic HSPs (Fig. 3B, right). In addition, chloramphenicol appeared to be a much less potent inducer of genes for mitochondrial HSP22s than methomyl. In contrast, the level of cpn60 induction by chloramphenicol was considerably higher, which was consistent with previous data on cpn60 expression in response to accumulation of a misfolded mitochondrial protein (34). Thus, our results suggest that, except for cpn60, an excess of non-assembled protein subunits within mitochondria is unlikely to generate a primary signal for induction of hsp genes.
It is well known that different HSPs are abundantly expressed during pollen development in maize (27,35,36). Indeed, as seen in Fig. 2B, genes for mitochondrial HSC70 and HSP22A are actively transcribed in male florets of fertile and sterile maize lines, including CMS-T with T-URF13-impaired mitochondria and also CMS-C and CMS-S lines that develop unknown mitochondrial dysfunctions during microsporogenesis (13). However, the hsp22B gene is induced exclusively in CMS-T and the mthsc70 expression is increased in CMS-T, where mitochondrial membrane depolarization was suggested to occur naturally in anthers (37). Thus, it is likely that the same mechanism of ⌬ m -dependent hsp activation is turned on in CMS-T maize during microsporogenesis and upon methomyl treatment of vegetative tissues.
The differential kinetics of hsp induction in response to methomyl treatment indicates that there is a direct signal from a reduced ⌬ m to up-regulate the genes for mitochondrial sHSPs. On the other hand, the delayed induction of genes for cytosolic HSPs would suggest another signal generated as a secondary consequence of mitochondrial membrane depolarization. Inhibition of protein import into uncoupled mitochondria leads to accumulation of some mitochondrial precursors in cytosol (38), facilitating their aggregation. In turn, an excess of non-native proteins and their aggregates in cytosol can trigger the induction of hsp genes (9). Indeed, in NCS mutants that should have reduced ⌬ m , we have found that accumulation of the nuclearly encoded E1␣ subunit of mitochondrial pyruvate dehydrogenase in the cytosol correlates with the significant increase in levels of HSP101 protein (Fig. 4, A and B). Concomitantly in the mitochondria of NCS mutants, high steady-state amounts of HSP22(A), the major HS-responsive maize mitochondrial protein (22), were detected (Fig. 4C). In addition, analysis of mitochondrial HSP70s with the monoclonal antibodies (22) revealed the presence of a new immunoreactive protein in addition to the constitutive HSP70. It is likely that the novel HSP70 isoform represents a highly homologous inducible mitochondrial HSP70 (ϳ50% of the deduced amino acid sequence shows 92% identity to that of constitutive HSP70; data not shown). These changes in protein patterns of inducible mitochondrial HSPs are consistent with high induction of their transcripts in NCS mutants (Fig. 1A), whereas levels of constitutive mitochondrial HSPs, HSP70 and the very abundant CPN60 (Fig. 4C), do not reflect an increase in transcription of the corresponding genes.
In mtDNA-depleted mammalian cell cultures, a decrease in ⌬ m was shown to be associated with a rise in cytosolic Ca 2ϩ levels, and increased expression of some Ca 2ϩ -responsive genes was reported (29). We found no difference in transcript levels of the Ca 2ϩ -inducible adh1 gene for alcohol dehydrogenase (39) between NCS mutants and normals in the ear shoot tissues where an overexpression of the hsp genes has been observed (compare Figs. 5A and 1A). To test directly whether genes for mitochondrial and cytosolic HSPs respond to increases in cytosolic Ca 2ϩ , we treated B37T maize seedlings with the ionophore A23187, which causes fluxes of intracompartmental Ca 2ϩ into the cytosol. Unlike methomyl, which would also release Ca 2ϩ specifically from mitochondria (40), A23187 does not affect ⌬ m unless exogenous Ca 2ϩ is supplied (41,42). Adh1 is up-regulated in response to hypoxia, which was shown to be mediated by Ca 2ϩ (43). Gradual induction of adh1 is commonly observed in maize seedling roots submerged in water for up to 24 h (Fig. 5B, no additions). Both methomyl and A23187 caused much faster adh1 induction (Fig. 5B), apparently reflecting their effects on Ca 2ϩ efflux into the cytosol. However, neither hsp22A nor shsp-I responded to the A23187 treatment (Fig. 5B). These experiments suggest that a decrease in ⌬ m itself, rather than changes in cellular Ca 2ϩ fluxes, generates a primary signal to induce expression of hsp genes.
RNAi Silencing of Nuclear Respiratory Genes in C. elegans Leads to Induction of hsp Genes-We predicted that signaling from dysfunctional mitochondria to induce hsp genes would exist in organisms other than plants. It may take place not only in mitochondrial mutants but also when the expression of nuclear genes coding for mitochondrial components is impaired. To assess the presence of the mitochondria-mediated signaling to express hsp genes in nematode C. elegans, we used RNAi (44) of nuclear genes for the components of the ETC. The genes for the following proteins have been chosen as targets: B18 subunit of respiratory complex I, 14-kDa subunit of complex III, cytochrome c 1 (complex III), and coxV b subunit of complex IV. All cultures showed developmental delays (mostly L2 and L3 larvae instead of predominant L4 larvae in control culture after 74 h). Like maize mitochondrial mutants (3), none of the RNAi-treated cultures seemed to experience oxidative  2N, 4N, 6N). Immunoblots were probed with specific antibodies to maize HSPs as indicated. Other antibodies used were to maize E1␣ subunit of mitochondrial pyruvate dehydrogenase (PDH) and to Arabidopsis mitochondrial manganese superoxide dismutase (MnSOD) and cytosolic glutathione S-transferase 1 (GST1).

FIG. 5. Ca 2؉ signaling is not involved in the mitochondriamediated expression of hsp genes in maize.
A, hybridization of total RNAs from NCS mutants (lanes 2, 4, and 6) and from corresponding normals (lanes N) with probe specific to the Ca 2ϩ -inducible adh1 gene. RNA samples were similar to those shown in Fig. 1A. B, treatments of B37T maize seedlings with A23187 (10 M) and with methomyl (10 mM). Northern analysis of total RNAs from seedling roots after incubation times as indicated.
stress. No induction of the gst4 and pqm1 transcripts was detected (Fig. 6A), though these genes were shown to be the most ROS-responsive in C. elegans (45). We have found (Fig. 6A and Supplemental Table II) that RNAi inactivation of respiratory complexes I, III, and IV leads to induction of the major C. elegans heat-inducible gene (46) for a cytosolic HSP70, hsp70.6 (but not of the highly homologous hsp70.4; not shown). Moderately increased were transcripts of hsp70A and, more prominently, those of the gene for mitochondrial HSP70 (hsp70F). Among the genes for sHSPs, no induction of hsp12.6 or change in transcript levels of hsp25 was detected (data not shown).
The hsp-17 gene, whose expression had not been characterized before, was highly induced in the RNAi-treated cultures, although it did not respond to heat shock (Fig. 6, A and B). Interestingly, the homologous hsp-16.1 and hsp-16.2 genes were induced although no expression of another homologous pair of sHSP genes, hsp- 16.48 and hsp-16.41, was detected (Fig.  6A). In contrast, HS induced both types of hsp-16 genes (Fig.  6B). In the C. elegans genome, hsp16s are organized tandemly, hsp- 16.1/hsp-16.48 and hsp-16.2/hsp-16.41. In each case, the shared bidirectional promoter regions contain identical heat shock elements, which correlates with co-regulation of tan-demly organized hsp16 genes during heat shock by a single heat shock factor, HSF1 (47,48). Thus, the differential expression of hsp16 genes and activation of hsp17 gene detected in our RNAi experiments suggest that the induction of shsp genes in C. elegans carrying mitochondrial deficiencies is distinct from the HSF1-dependent HS response and may require other transcription factors and cis-recognition sequences. DISCUSSION Here we report that acclimation to permanent mitochondrial dysfunction involves steady-state activation of specific sets of hsp genes encoding cytosolic and mitochondrial molecular chaperones. Using maize NCS mutants with defined deletions in mtDNA and cultures of C. elegans with RNAi-inactivated nuclear respiratory genes, we show that mitochondria-dependent hsp gene activation in different organisms has common but distinctive characteristics.
(i) Activation of hsp genes depends on inhibition of the electron transfer chain regardless of the type of respiratory defect or its origin (loss of either mitochondrial or nuclear-coded product). In the transient in vivo assays, we have shown that in maize same sets of hsp genes can be induced by depolarization of the mitochondrial membrane. Different kinetics of ⌬ m -dependent activation suggest different induction pathways for genes encoding mitochondrial and cytosolic HSPs. It is likely that the reduced ⌬ m , a common result of the ETC inhibition, serves as a primary signal to induce hsp genes coding for mitochondrially targeted proteins. Consequently, inefficient protein import into the de-energized mitochondria might lead to the accumulation of mitochondrial precursors in the cytosol, which was indeed detected in our experiments. This excess of non-native proteins might contribute to the induction of genes for cytosolic HSPs.
(ii) In both maize and nematodes, mitochondrial deficiencies differentially activate some HS-inducible hsp genes and also cause high expression of some HS non-responsive hsp genes (including cryptic ones, like hsp22B in maize and hsp-17 in C. elegans). This pattern of hsp expression differs from a typical HS response that involves recognition of heat shock elements by an activated heat shock factor (28). It is likely that mitochondria-dependent activation of hsp genes may require additional transcription factors and cis-recognition sequences.
(iii) Expression of hsp genes under conditions of permanent respiratory deficiency is not associated with oxidative stress and is not mediated by Ca 2ϩ , which clearly distinguishes this type of hsp induction from any others known to date.
High expression of some of the tested maize hsp genes at the protein level indicates that activation of hsp genes by impaired mitochondria might be of physiological importance. The remarkable similarities in response to respiratory deficiency in plants and in animals indicate that the constitutive expression of the hsp gene network is a significant part of a program to cope with the disruption of the mitochondrial function. Permanent expression of hsp genes in respiratory-deficient individuals seems to be required for their development and may reflect different changes in genome expression patterns in comparison to the transient responses of normal organisms subjected to temporary biotic/abiotic stresses. Analysis of genome-wide expression in 0 mutants compared with wild-type cultures of S. cerevisiae treated with respiratory inhibitors confirmed that gene expression patterns upon mtDNA depletion and after transient inhibition of the ETC are different and overlap only partially (49).
Our data on maize NCS mutants suggest that mitochondriadependent hsp expression could be coordinated with another stress defensive mechanism associated with plant-specific mitochondrial alternative oxidase, AOX (3). As a terminal ubiquinol oxidase, AOX acts as a mild uncoupler of oxidative phosphorylation and as an antioxidant (50) and prevents an oxidative stress in NCS mutants (3). Thus, in NCS mutants with high constitutive expression of AOX, effective inhibition of ROS generation occurs at the expense of further de-energization of the mitochondrial membrane. In turn, this decrease in ⌬ m triggers signaling to express cytosolic and mitochondrial HSPs that would have a protective effect on cells and particularly on dysfunctional mitochondria (5)(6)(7)(8)51). Orchestrated hsp gene expression contributes to a metabolically stable acclimation that prevents oxidative stress and apoptosis in respiratory-deficient mutants (Fig. 7).
In experiments with C. elegans, we observed hsp gene induction upon RNAi inactivation of the same nuclear respiratory genes, whose RNAi silencing had been found to give a pronounced increase in lifespan (52,53). These results suggest that induction of hsp genes through the signaling from impaired mitochondria may represent one of the crucial links between mitochondrial state and longevity (reviewed in Ref. 54).
Thus, we show here that in both plants and animals, there is the mitochondria-dependent hsp gene expression that is different from typical responses to heat shock or oxidative stress and could imply novel signaling pathways and induction mechanisms. Establishing the constitutive expression of hsp genes in the absence of exogenous stress may play a vital role in cell defense and adaptation to permanent respiratory deficiency.