Hypoxia induces expression of a GPI-anchorless splice variant of the prion protein Yutaka Kikuchi1, Tomoshi Kakeya1, Osamu Nakajima2, Ayako Sakai1, Kikuko Ikeda3, Naoto Yamaguchi3, Takeshi Yamazaki4, Ken-ichi Tanamoto4, Haruo Matsuda5, Jun-ichi Sawada2 and Kosuke Takatori1 Division of Microbiology, National Institute of Health Sciences, Tokyo, Japan Division of Biochemistry and Immunochemistry, National Institute of Health Sciences, Tokyo, Japan Department of Molecular Cell Biology, Graduate School of Pharmaceutical Science, Chiba University, Japan Division of Food Additives, National Institute of Health Sciences, Tokyo, Japan Laboratory of Immunobiology, Graduate School of Biosphere Science, Hiroshima University, Japan Keywords alternative splicing; Creudzfeldt–Jakob disease; GPI anchor; hypoxia; prion Correspondence Y Kikuchi, Division of Microbiology, National Institute of Health Sciences, 1-18-1 Kamiyoga, Setagaya-ku, Tokyo 158-8501, Japan Fax: +81 3707 6950 Tel: +81 3700 1141 E-mail: kikuchi@nihs.go.jp (Received 30 January 2008, revised 24 March 2008, accepted April 2008) The human prion protein (PrP) is a glycoprotein with a glycosylphosphatidylinositol (GPI) anchor at its C-terminus Here we report alternative splicing within exon of the PrP gene (PRNP) in the human glioblastoma cell line T98G The open reading frame of the alternatively spliced mRNA lacked the GPI anchor signal sequence and encoded a 230 amino acid polypeptide Its product, GPI-anchorless PrP (GPI) PrPSV), was unglycosylated and soluble in non-ionic detergent, and was found in the cytosolic fraction We also detected low levels of alternatively spliced mRNA in human brain and non-neuronal tissues When long-term passaged T98G cells were placed in a low-oxygen environment, alternatively spliced mRNA expression increased and expression of normally spliced PrP mRNA decreased These findings imply that oxygen tension regulates GPI) PrPSV expression in T98G cells doi:10.1111/j.1742-4658.2008.06452.x Fatal human prion diseases, including sporadic, iatrogenic and variant Creudzfeldt–Jakob disease (CJD), inherited prion diseases and kuru, are transmissible spongiform encephalopathies characterized by the formation and accumulation of an abnormal isoform of prion protein (PrP) in the brain [1] Cellular prion protein (PrPC) is a glycoprotein that is anchored to the cell surface by a glycosylphosphatidylinositol (GPI) moiety [1] CJD is associated with the conversion of PrPC into a protease-resistant isoform (PrPres), either on the cell surface or within its compartments [1] Sporadic CJD is classified on the basis of the molecular mass of the unglycosylated fragment of PrPres as type (21 kDa) or type (19 kDa), and on the basis of the genotype at the methionine ⁄ valine polymorphic codon 129, i.e MM, MV or VV [2] In a previous study, we showed that the prion protein gene (PRNP) in human glioblastoma cell line T98G, which is of the MV genotype and produces a form of PrP that is partially resistant to proteinase K (PK) following long-term culture and high passage number, has no mutation in the coding region [3] The PrPres fragment described here, which differed from corresponding fragments in typical sporadic CJD, had a mass of 18 kDa after deglycosylation and was detergent-soluble [3] However, in one report, brain homogenates from dead patients with type PrP and MV (n = 5) or VV (n = 6) genotypes also contained PrP fragments that migrated at 18 kDa after deglycosy- Abbreviations AD, Alzheimer’s disease; CJD, Creudzfeldt–Jakob disease; GPI, glycosylphosphatidylinositol; GPI), GPI-anchorless; PK, proteinase K; PNGase F, peptide N-glycosidase F; PRNP, prion protein gene; PrP, prion protein; PrPC, cellular PrP; PrPres, protease-resistant isoform of PrP; PrPSV, splice valiant of PrP FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS 2965 Hypoxia induces expression of GPI-anchorless PrP Y Kikuchi et al lation and were detergent-soluble [4] These findings support the possibility that our findings in T98G cells may be relevant to PrPres in sporadic CJD brain Alzheimer’s disease (AD) and prion disease share a common feature – aggregation and deposition of abnormal proteins [5] Intracerebral injection of post mortem brain extracts from AD patients induced deposition of amyloid b–peptides in the hippocampus of b-amyloid preursor protein transgenic mice [6] Some cohort studies have indicated that cerebral ischemia and stroke significantly increase AD risk [7,8], and hypoxia seems to be an important contributor to the onset and progression of AD [9] Recent magnetic resonance imaging studies have also suggested that changes in areas of the brain with the highest oxygen requirement are associated with sporadic CJD [10,11] A retrospective study detected these changes in 39.1% of sporadic CJD patients (n = 1036) [10] A study of human cerebral ischemia and perinatal hypoxic ⁄ ischemic injury confirmed the presence of PrPC immunoreactivity within axons in the penumbra of white matter damage and within neuronal soma of gray matter damage [12] We therefore speculated that oxidative stress is a causative factor in prion disease To test this hypothesis, we investigated the effects of hypoxia on PrP expression using T98G cells as our model system Results Detection of the splice variant form of PrP mRNA in T98G cells First, we analyzed PRNP mRNA by RT-PCR We used total RNA isolated from T98G cells to generate RT-PCR products from PRNP exon We found that cells grown under normoxic conditions produced a 528 bp product (supplementary Fig S1) when the cells were cultured for 24 days after two passages (P2D24) and for 24 days after 18 passages (P18D24) (Fig 1A) In contrast, total RNA from P18D24 cells exposed to hypoxic conditions (5% O2) for the last day expressed a shorter product, i.e 296 bp, but the genomic DNA from P18D24 cells and total RNA from P2D24 cells did not express this product (Fig 1A) Because addition of cobalt ion can mimic hypoxic conditions [13], we next studied its effects on RT-PCR products Total RNA from P2D39 cells cultured with or 300 lm CoCl2 for the last day yielded the 528 bp RT-PCR product (Fig 1B) Total RNA from P13D24 cells cultured the same way, however, expressed the 528 bp product and the shorter RT-PCR product (Fig 1B), just as P18D24 cells exposed to hypoxic conditions (Fig 1A) To amplify the shorter RT-PCR product, 2966 we decreased the elongation time to 36 s to avoid saturation by the longer PCR product [14] With an elongation time of 60 s, the shorter RT-PCR product was not amplified and only the 528 bp product was produced (data not shown) When we performed direct sequencing of the shorter product, we found that a 232 bp sequence was missing from the 528 bp sequence (supplementary Fig S1) We identified an intronic sequence with the canonical dinucleotides for splicing (GT at the 5¢ end and AG at the 3¢ end) and a pyrimidine tract (16 pyrimidines ⁄ 20 bases) 20 nucleotides upstream of the 3¢ splice site [15,16] Thus, our data indicate that alternative splicing occurred within PRNP exon To determine the cryptic splice sites, we designed exon–exon junction primers that annealed with the donor and acceptor sites (E2SV3, E2SV4 and E2SV5; supplementary Table S1), and used total RNA isolated from P18D24 cells to generate RT-PCR products As expected, we detected a 1433 bp product from cells grown under hypoxic conditions In addition, we detected two shorter RT-PCR products – a 676 bp product when we used E2SV3 and a 553 bp product when we used E2SV4 (Fig 1C) Surprisingly, we also detected these products in total RNA from cells grown under normoxic condition (Fig 1C) However, when using genomic DNA, we detected only full-length PCR products (Fig 1C) Thus, the exon–exon junction primers were able to detect mRNA for the splice variant of PrP (PrPSV) Direct sequence analysis revealed that the only mutation in the 1433 bp RT-PCR product was an adenine to guanine substitution in the first position of codon 129, i.e the common M129V polymorphism (supplementary Fig S1; T98G PrP, accession numbers AB300823 and AB300824); the alternatively spliced 1201 bp product also contained the polymorphism (supplementary Fig S1; T98G PrPSV, accession numbers AB300825 and AB300826) PRNP encodes a 253 amino acid polypeptide, including an N-terminal signal sequence (residues 1–22) and a GPI anchor signal sequence (residues 231–253) (Fig 1D, upper part) Alternative splicing resulted in use of exons 2a and 2b with a cryptic donor site and a cryptic acceptor site (Fig 1D, lower part), with an open reading frame encoding a 230 amino acid polypeptide comprising the N-terminal portion (residues 1–217) of PrP from exon 2a and the C-terminal peptide (residues 218–230) from exon 2b (lower panel) Expression of the GPI-anchorless splice variant of PrP in T98G cells We next investigated the prion protein expressed by the alternatively spliced mRNA To detect PrPSV, we raised FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS Y Kikuchi et al Hypoxia induces expression of GPI-anchorless PrP Fig Expression of a splice variant of PrP mRNA in T98G cells (A) Exposure of T98G cells to hypoxia produces a splice variant of PrP mRNA P2D24 and P18D24 cells were exposed to hypoxia (5% O2) or normoxia for the last day in culture First-strand cDNA from total RNA and genomic DNA were used for PCR using PrP (E2U0 ⁄ E2L0, P) and b-actin (ACTBU1 ⁄ ACTBL1, A) primer sets with KOD Plus DNA polymerase The PCR products were separated on a 2% agarose gel and visualized with ethidium bromide (B) Cobalt chloride induced expression of a PrP mRNA splice variant in T98G cells P2D39 and P13D24 cells were cultured with or without 300 lM CoCl2 for the last day in culture We used first-strand cDNA for PCR using human PrP primer sets (E2U0 ⁄ E2L0, P) or b-actin (ACTBU1 ⁄ ACTBL1, A) primer sets with KOD Plus DNA polymerase The PCR products were analyzed as described above (C) Detection of a splice variant of PrP mRNA using exon–exon junction primers P18D24 cells were exposed to hypoxia or normoxia for the last day in culture Firststrand cDNA and genomic DNA were used for PCR using PrP (E2U1 ⁄ E2L4, P) and exon–exon junction (E2U1 ⁄ E2SV3, J1; E2SV4 ⁄ E2L4, J2) primer sets with Ex Taq DNA polymerase The PCR products were analyzed as described above (D) Schematic representation of alternative splicing of PRNP We confirmed the sequences of normally (upper panel) and alternatively (lower panel) spliced PRNP Cryptic donor and acceptor sites are designated as exons 2a and 2b, respectively The untranslated regions (white bars), open reading frames (blue bars), retained intron (turquoise line), additional open reading frame (magenta bar) and deduced amino acid sequences of normally (blue) and alternatively (magenta) spliced PRNP are indicated The arrow indicates a GPI anchoring site Asn-Xaa- (Ser/Thr) sequous for N-linked glycosylation an mAb, HPSV178, against the C-terminal portion of PrPSV (residues 214–230), and found that it reacted with recombinant PrPSV but not with recombinant PrP (supplementary Fig S2) We determined PrPSV expression by immunoblotting homogenates of cells with various passage numbers The mAb 6H4 against human PrP recognized di-, mono- and unglycosylated PrP (Fig 2A, upper panel) No PrPSV was detected in the homogenates of P40D40 T98G cells (Fig 2A, lower panel, lane 1), but HPSV178 did recognize a 25 kDa band in the homogenates of P52D40 and P77D40 T98G cells (Fig 2A, lower panel, lanes and 3) As we routinely sub-cultivated the cells once a week, these data indicated that detectable PrPSV first appeared after at least year (i.e after 52 passages) To assess the glycosylation status of PrPSV, we treated P77D40 cell homogenates with peptide N-glycosidase F (PNGase F), which yields a full-length (25 kDa) and an N-terminally truncated (18 kDa) form of PrPC [3] As shown in Fig 2B, PNGase F treatment reduced the 35 and 31 kDa glycosylated bands to the deglycosylated full-length 25 kDa band and N-terminally truncated 18 kDa band, respectively (upper panel) The mobility of PrPSV was unaltered (lower panel), showing that PrPSV was mainly unglycosylated To determine the subcellular location of PrPSV, we prepared membrane and cytosolic fractions from the homogenates As shown in Fig 2C, PrP was detected only in the membrane fraction, and PrPSV was detected only in the cytosolic fraction (lower panel), indicating that PrPSV lacks a GPI anchor To test the detergent solubility of PrPSV, we centrifuged the homogenates in the presence of non-ionic detergents A large proportion of immunoreactive PrP and PrPSV was found only in the soluble fraction (Fig 2D), indicating that PrPSV, like PrP, is soluble in non-ionic detergents FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS 2967 Hypoxia induces expression of GPI-anchorless PrP Y Kikuchi et al Fig Characterization of the GPI-anchorless splice variant of PrP in T98G cells (A) Immunoblot analysis using mAb against GPI) PrPSV in T98G cells Homogenates (50 lg protein) of P40D40, P52D40 and P77D40 cells were subjected to immunoblotting with 6H4 (upper panel) and HPSV178 (lower panel) monoclonal antibodies as described in Experimental procedures (B) Analysis of deglycosylated forms of GPI) PrPSV Homogenates (50 lg protein) of P77D40 cells were incubated with (+) or without ()) PNGase F and the products were subjected to immunoblotting as described above (C) Subcellular localization of GPI) PrPSV Homogenates (50 lg protein) of P77D40 cells were separated into membrane and methanol-precipitated cytosolic fractions, resuspended in the same volume of NaCl ⁄ Pi containing 2.5 mM EDTA, and subjected to immunoblotting as described above (D) Detergent solubility of GPI) PrPSV Homogenates (50 lg protein) of P77D40 cells were dissolved in nine volumes of 0.5% NP-40 ⁄ 0.5% deoxycholate ⁄ NaCl ⁄ Pi and centrifuged The pellet fraction (insoluble fr.) and the methanol-precipitated supernatant fraction (soluble fr.) were resuspended in the same volume of NaCl ⁄ Pi containing 2.5 mM EDTA and subjected to immunoblotting as described above Epitope sites located within PrP and PrPSV are shown on the left as residue numbers Thus, GPI-anchorless (GPI)) PrPSV is localized in the cytosol in unglycosylated form Induction of production of the GPI-anchorless splice variant of PrP by hypoxia We then examined whether PrPres is propagated under hypoxic conditions For P40D40 cells, PrP mRNA levels were significantly lower when they were cultured for the last 1–4 days in 300 lm CoCl2 rather than under normoxic conditions (Fig 3A, upper panel), and immunoblotting with 6H4 mAb showed that PrP protein levels decreased (Fig 3C, upper panel) After treatment with 300 lm CoCl2 for the last days, no PrPres was detected, and PrPSV mRNA levels and their ratio to PrP mRNA levels increased (Fig 3A, middle and lower panels) However, GPI) PrPSV was not observed (Fig 3C, middle panel) For P90D40 cells, PrP mRNA levels decreased significantly under hypoxic conditions (100 lm CoCl2 or 2% O2) for the last days (Fig 3B, upper panel), but PrPres was still detected by 6H4 mAb (Fig 3D, upper panel) PrPSV mRNA levels and their ratios to PrP mRNA levels were significantly higher (Fig 3B, middle and lower panels), and GPI) PrPSV was detected in the cell 2968 homogenates (Fig 3D, middle panel), and is equivalent to the band shown in Fig 2A–D (lower panels) No GPI) PrPSV band was detected following PK treatment of the P90D40 cell homogenates (Fig 3D, middle panel) Quantitative densitometric analysis of the 25 kDa bands (Fig 3D, middle panel) revealed that the amounts of 100 lm CoCl2- and 2% O2-induced GPI) PrPSV increased to almost 4.3 and 4.8 times their original levels, respectively (Fig 3E, lower panel) Detection of the splice variant of PrP mRNA in human brain To investigate whether PrPSV mRNA is expressed in human brain, we subjected total RNA of adult and fetal brains to RT-PCR analysis using the exon–exon junction primers As shown in Fig 4A (lanes and 5), 1433 bp products were detected in both samples using primers E2U1 ⁄ E2L4 (primer set P) Surprisingly, two shorter RT-PCR products were detected when exon–exon junction primer sets were used – a 676 bp product with primers E2U1 ⁄ E2SV3 (primer set J1) and a 553 bp product with primers E2SV4 ⁄ E2L4 (primer set J2) These results indicate that very low levels of PrPSV mRNA are constitu- FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS Y Kikuchi et al Hypoxia induces expression of GPI-anchorless PrP Fig Hypoxia-induced expression of the GPI-anchorless splice variant of PrP in T98G cells (A,B) Quantification of PrPSV mRNA in T98G cells P40D40 cells were exposed to CoCl2 (300 lM) for the last 0, or days (A) P90D40 cells were exposed to hypoxia (2% O2), CoCl2 (100 lM) or normoxia for the last days (B) The mRNA levels were analyzed by quantitative RT-PCR using PrP and exon–exon junction primer sets as described in Experimental procedures The expression values of PrP and PrPSV and their ratios were normalized to those of b-actin Values are the mean ± standard error of three independent cell samples *P < 0.05 compared with day (A) or normoxia (B) (Dunnett test) (C,D) Proteinase K sensitivity of GPI) PrPSV Methanol-precipitated homogenates (50 lg protein) of P40D40 (A) and P90D40 (B) cells were treated with or 10 lgỈmL)1 PK at 37 °C for 30 The products were subjected to immunoblotting with 6H4 (upper panel), HPSV178 (middle panel) or AC-15 (lower panel) antibodies as described in Experimental procedures (E) Densitometric quantification of GPI) PrPSV Quantitative analysis of GPI) PrPSV and b-actin shown in (D) at 25 kDa (middle panel, odd-numbered lanes) and at 42 kDa (lower panel, odd-numbered lanes), respectively, were performed by computer-assisted densitometry The density of GPI) PrPSV (%) was normalized to that of b-actin (lower panel) Values are the means ± SD of triplicate measurements within one experiment *P < 0.05 compared with normoxia (Dunnett test) FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS 2969 Hypoxia induces expression of GPI-anchorless PrP Y Kikuchi et al Fig Expression of the splice variant of PrP mRNA in human tissues (A) Detection of the splice variant of PrP mRNA in total RNA from human brains We prepared first-strand cDNA from total RNA from adult (43-year-old Caucasian male) and fetal human brains (5 lg each), and subjected it to PCR using PRNP exon (E2U1 ⁄ E2L4, P), exon–exon junction (E2U1 ⁄ E2SV3, J1; E2SV4 ⁄ E2L4, J2) and b-actin (ACTBU1 ⁄ ACTBL1, A) primer sets using Ex Taq DNA polymerase PCR products were separated on a 2% agarose gel and visualized with ethidium bromide (B) Quantification of the splice variant of PrP mRNA in total RNA from human brain We analyzed the human brain total RNA shown in (A) by quantitative RT-PCR using PrP primers and exon–exon junction primer sets as described in Experimental procedures The expression values of PrP and PrPSV and their ratios were normalized to those of b-actin Values are the means ± SD of triplicate measurements within one experiment *P < 0.05 compared with adult (Student’s t test) (C) Expression of the splice variant of PrP mRNA in human tissues RT-PCR showing the splice variant of PrP mRNA using exon–exon junction primers in various human tissues We prepared total RNA from adult human tissues (5 lg each) and subjected it to PCR using exon–exon junction (E2SV4 ⁄ E2L4, upper panel) and b-actin (ACTBU1 ⁄ ACTBL1, lower panel) primer sets with Ex Taq DNA polymerase Amplification of b-actin is shown as a quality check for total RNA (lower panel) The PCR products were analyzed as described above tively expressed in fetal and adult human brains Relative quantitative RT-PCR analysis, which measures PrP mRNA in total RNA (Fig 4B, upper panel), showed that the PrP mRNA level in adult brain was higher than that in fetal brain PrPSV mRNA levels were not significantly different between the two samples (Fig 4B, middle panel) We then used RT-PCR to determine PrPSV mRNA distribution in total RNA from a panel of adult human tissues (Fig 4C) We found expression in all the tested samples, including non-neuronal samples, indicating that PrPSV mRNA is constitutively expressed in all major organs and tissues Discussion At least two homologs have been mapped to the PRNP locus (20pter-p12): PRND ⁄ Doppel and PRNT 2970 [17] Expression of the human PrP-like glycoprotein PrPLP ⁄ Dpl by these homologs has been confirmed in human adult testes [18] and astrocytomas [19] The 176 amino acids of PrPLP ⁄ Dpl include an N-terminal signal sequence (residues 1–23) and a GPI anchor signal sequence (residues 153–176) [17] PrPLP ⁄ Dpl (residues 24–152) has 25% sequence homology with PrP and has a truncated N-terminus Despite their low homology, PrP and PrPLP ⁄ Dpl show similar structural features The PrPLP ⁄ Dpl protein is expressed by intergenic splicing between PRNP and PRND/Doppel GPI) PrPSV protein, on the other hand, is expressed by intragenic splicing within PRNP exon and has 94% sequence homology with normally spliced PrP (residues 23–230) PRNP mRNA is ubiquitously expressed in brain and various tissues, including nonneuronal tissues [17], and, as discussed above, we demonstrated constitutive expression of PrPSV mRNA FEBS Journal 275 (2008) 2965–2976 ª 2008 The Authors Journal compilation ª 2008 FEBS Y Kikuchi et al in all major organs and tissues, including brain Although expression levels were low, the ratio of PrPSV mRNA to PrP mRNA was 0.00026% in adult brain (Fig 4B) We also detected PrPSV mRNA in human astrocytoma U373MG cells and human myelocytic leukemia HL-60 cells (supplementary Table S2) We detected GPI) PrPSV production in T98G cells after 52 passages The protein was mainly unglycosylated and appeared in the cytosolic fraction This agrees with findings that mouse recombinant PrP lacking a GPI anchor signal sequence is unglycosylated in human neuroblastoma cell line SH-SY5Y [20] Furthermore, GPI) PrP in the transgenic mouse brain is unglycosylated and is found in the cytosolic fraction [21,22] Two N-glycosylation sites – Asn (181) and Asn (197) – are 73 and 57 residues from the C-terminus in PrP, and 50 and 34 residues from the C-terminus in PrPSV, respectively (Fig 1D) N-glycosylation of mouse PrP was abolished when the first Asn-Xaa(Ser/Thr) sequon for N-linked glycosylation was