Elsewhere, we performed a genomewide linkage study and identified a “pure” HSP locus at chromosome 2p12 (SPG31).⁵ Two families, DUK2036 and DUK2299, yielded a combined two-point LOD score of 4.7 at marker D2S2951. Fine-mapping and haplotype analysis narrowed the locus to ~8.8 Mb between D2S139 and D2S2181 (fig. 1A). We chose nine candidate genes (CTNNA2, SUCLG1, TGOLN2, MATA2A, VAMP8, VAMP5, IMMT, VPS24, and REEP1 [MIM 609139]) on the basis of emerging pathways for spastic paraplegia and conserved protein domains contained in proteins that cause neurodegenerative diseases.⁶ Of those genes, we sequenced all exons, including 40 bp of flanking intronic and UTR sequences. We identified a single base-pair deletion, c.507delC, in family DUK2299 and a splice-site mutation, c.182-2A→G, in family DUK2036 in the receptor expression–enhancing protein 1 gene (REEP1), also known as “C2orf23” (fig. (fig.1B1B and and1C).1C). These sequence changes resulted in frameshifts leading to altered stop codons. The mutations cosegregated with the disorder in the linked pedigrees (fig. 2). REEP1 (^GenBank accession number NP_075063) belongs to a novel protein family that was identified only recently.⁷ Its subcellular location and molecular function are largely unknown. We screened for additional REEP1 mutations in a sample of 90 independent HSP-affected families of European descent, neither selected for pure HSP phenotype nor tested for mutations in other HSP genes. All individuals were studied under internal review board–approved procedures. We identified four more mutations that led to significant sequence changes in REEP1: one missense mutation (c.59C→A; Ala20Glu), one deletion (c.526delG; Gly176fs), and two 3′ UTR changes (c.606+43G→T and c.606+50G→A) (fig. (fig.1B1B and and1C1C and table 1). All mutations cosegregated with the HSP phenotype in the available pedigrees and were undetected in 365 control individuals of European descent (730 chromosomes) (fig. 2).

The splice-site mutation in linked family DUK2036 disrupted the canonical 5′ acceptor splicing signal “AG” of exon 4 (fig. (fig.1B1B and and1C).1C). REEP1 was not expressed in peripheral-blood cells, and neuronal tissue was not available from affected family members (fig. 3K). Thus, we could not demonstrate the aberrant mRNA (^GenBank accession number NM_022912), but simulation of the resulting splice efficiency with NNSPLICE revealed a reduction from 99% efficiency of the wild-type AG allele to 0% of the mutant GG allele. Missplicing of exon 4 will result in a frameshift followed by a premature stop codon. The two mutations in the 3′ UTR presented unusual changes for a Mendelian disease. However, both mutations altered the sequence of a predicted highly conserved binding site for the microRNA (miRNA) gene miR-140 (fig. (fig.1D1D and and1E).1E). miRNAs constitute a new large class of small noncoding RNA genes that target protein-coding genes for posttranscriptional repression⁸ (see the MicroRNA Registry). The c.606+43G→T mutation in REEP1 disrupted a G:U wobble base pair, and the c.606+50G→A change replaced a G:U wobble base pair with an A:U Watson-Crick pairing. It has been shown that G:U wobble base pairing has an inhibitory effect on miRNA-mediated repression of translation.⁹ Thus, both detected mutations would foster suppressive miRNA-mediated effects on translation, leading to less available REEP1 protein. Both affected nucleotides are highly conserved in the REEP1 3′ UTR as well as in miR-140 of different species (fig. (fig.1D1D and and1E).1E). In addition, it has been shown that miR-140 is expressed in the cortex of rat and monkey as well as in cultured primary cortical neurons from rat.^10,11 Thus, we suggest that the identified sequence variants will affect the amount of translated REEP1 in patients with HSP.

Three of the four coding changes led to alternative stop codons. It is likely that the resulting mRNA will be targeted for nonsense-mediated decay, resulting in haploinsufficiency of the mutant allele. Interestingly, both miRNA target-site mutations disrupted G:U base pairing and are therefore likely to lead to less translated protein. We suggest that loss of function and haploinsufficiency are the mechanisms of action in REEP1-related spastic paraplegia.

As shown by Saito et al.⁷ and extended in the present study, REEP1 is expressed in various nonneuronal and neuronal tissues, including spinal cord (fig. 3K). This follows the now-common finding of almost ubiquitous tissue expression for a number of genes that cause distinct neurodegenerative phenotypes. Saito et al. showed that REEP1 has a weak promoting effect on the expression of G protein–coupled odorant-receptor proteins at the cell surface.⁷ Thus, REEP1 might have a role in trafficking of odorant receptors and other molecules through cellular compartments such as the endoplasmic reticulum and Golgi apparatus. The yeast homologue of REEP1, Yop1P, is involved in Rab-mediated vesicle transport, a pathway that has recently been implicated in axonal neuropathy type CMT2B, which is caused by mutations in RAB7.^12,13 In silico analysis of REEP1 predicted two transmembrane domains and the conserved protein domain TB2/DP1/HVA22 (fig. 1B). As shown by Chen et al., the plant homologues of HVA22 are stress-induced genes—a characteristic known from heat-shock proteins.¹⁴ Heat-shock proteins play a fundamental role as chaperones to promote and maintain correct protein folding, especially in cell compartments rich in reactive oxygen species, such as mitochondria. Mutations in the mitochondrial heat-shock protein 60 have been shown to cause spastic paraplegia type SPG13.⁴ It has also been suggested that the spastic paraplegia gene paraplegin (SPG7 [MIM 602781]) fulfills chaperonelike activities in mitochondria.¹⁵

We designed two specific polyclonal antibodies that targeted the C terminal of REEP1 (fig. 4). Staining of COS7 and MN-1 cells with those antibodies revealed that endogenous REEP1 colocalized with mitochondria (fig. (fig.3A3A–3F). REEP1 was also present in the mitochondrial but not the cytosolic cellular fraction on an immunoblot (fig. 3I). Given the predicted transmembrane domains (figs. (figs.1B1B and and5),5), we suggest that REEP1 is a novel mitochondrial membrane protein.

Although the results of Saito et al. suggested localization of REEP1 to the secretory pathway,⁷ we did not detect colocalization of REEP1 with the Golgi (fig. 1H). It is conceivable that REEP1 has different cellular functions and that a fraction of the protein, which was not detectable with the available antibodies, indeed fulfills functions in the endoplasmic reticulum and Golgi. In support of this hypothesis, the ENSEMBL human genome database lists two alternative isoforms of REEP1 that might represent such differential functionality. However, we were not able to unequivocally reproduce the existence of this second isoform.

Although the specific function of REEP1 in mitochondria has not been elucidated, this finding contributes further to the evidence that mitochondrial integrity takes center stage in HSP and related neurodegenerative diseases.⁶