Gene Codon Composition Determines Differentiation-Dependent Expression of a Viral Capsid Gene in Keratinocytes In Vitro and In Vivo

Abstract

By establishing mouse primary keratinocytes (KCs) in culture, we were able, for the first time, to express papillomavirus major capsid (L1) proteins by transient transfection of authentic or codon-modified L1 gene expression plasmids. We demonstrate in vitro and in vivo that gene codon composition is in part responsible for differentiation-dependent expression of L1 protein in KCs. L1 mRNA was present in similar amounts in differentiated and undifferentiated KCs transfected with authentic or codon-modified L1 genes and had a similar half-life, demonstrating that L1 protein production is posttranscriptionally regulated. We demonstrate further that KCs substantially change their tRNA profiles upon differentiation. Aminoacyl-tRNAs from differentiated KCs but not undifferentiated KCs enhanced the translation of authentic L1 mRNA, suggesting that differentiation-associated change to tRNA profiles enhances L1 expression in differentiated KCs. Thus, our data reveal a novel mechanism for regulation of gene expression utilized by a virus to direct viral capsid protein expression to the site of virion assembly in mature KCs. Analysis of two structural proteins of KCs, involucrin and keratin 14, suggests that translation of their mRNAs is also regulated, in association with KC differentiation in vitro, by a similar mechanism.

Direction of gene expression to undifferentiated or differentiated cells is classically determined by altered promoter methylation or by production of specific transcription factors or posttranscriptionally by interaction of regulatory mRNA sequences with translational regulators (38). Papillomaviruses (PVs) are a family of double-stranded DNA viruses which replicate exclusively in epithelium, promote cell growth, and affect cellular differentiation, giving rise to benign tumors with, for some virus types, potential for malignancy. mRNA encoding the PV major capsid L1 protein can be transcribed from L1 gene expression plasmids in many types of mammalian cells (20). However, translation of the transcribed mRNA to L1 protein is limited in vivo to differentiated keratinocytes (KCs) (5, 53) and to yeast cells (50, 67). Although inhibitory mechanisms have been proposed to explain the blockage of PV L1 gene translation in undifferentiated cells (10, 11), no inhibitory factors have been identified as specific for epithelial cells in vitro or in vivo. Thus, the mechanism for the tight differentiation-specific translation of the PV L1 gene in KCs remains to be determined.

PVs, like many mammalian DNA viruses, use relatively few “mammalian consensus” codons to encode their capsid genes, manifesting a high A+T genome content due to third-nucleotide bias to A+T (68). In humans, codon-mediated translational controls may play an important role in the differentiation and regulation of tissue-specific gene products (47). Blockage to translation of PV L1 mRNAs has been overcome by codon modification utilizing mammalian preferred codons without changing the protein sequence (36, 41, 42, 69), but it remains unclear whether codon modification assists L1 protein synthesis by removal of sequences inhibitory to mRNA translation or destabilizing of mRNA or by some other mechanism. Mechanisms postulated to determine instability of L1 mRNA in undifferentiated cells (32, 54, 62) have not been demonstrated to distinguish between undifferentiated and differentiated KCs, the host cells of PV infection.

Terminally differentiated KCs flatten and develop a cornified envelope, which provides the barrier function of epithelia (1). The proliferation and differentiation capacity of cultured epidermal cells makes KCs ideal candidates for gene targeting and drug therapy (35). It is therefore desirable to develop models for the study of regulation of gene expression in KCs as they develop. In this study, we examined expression of the PV major capsid (L1) proteins from authentic or codon-modified (Mod) L1 gene expression plasmids in KC culture in vitro and in mouse skin in vivo. We demonstrate that gene codon composition determines the timing of PV L1 capsid protein expression in KC culture in vitro upon differentiation and the differential expression of the L1 gene between the superficial and basal epithelium of the skin in vivo. Substantial differences were demonstrated in the tRNA pools of differentiated and undifferentiated KCs. A change in the aminoacyl-tRNA (aa-tRNA) pool upon KC differentiation enhances translation of authentic but not Mod PV L1 mRNA, likely reflecting a better match in differentiated KCs between available aa-tRNAs and the codons present in the PV L1 gene but rarely used in most mammalian genes.

MATERIALS AND METHODS

Construction of native (Nat) and Mod PV L1 genes.

The plasmids used in the experiments described here (pCDNA₃HPV6b Nat L1, pCDNA₃HPV6b Mod L1, pCDNA₃BPV1 Nat L1, and pCDNA₃BPV1 Mod L1) have previously been described (69). Briefly, both the bovine PV 1 (BPV1) and human PV 6b (HPV6b) wild-type L1 open reading frames (ORFs) are about 1.5 kb in length, encoding ∼500 amino acids. The PV wild-type L1 genes show a strong codon usage bias, among degenerately encoded amino acids, toward 18 codons mainly with T at the third position that are rarely used by mammalian genes (68, 69). We artificially modified BPV1 and HPV6b L1 genes in which the L1 ORFs are substituted with codons preferentially used in the mammalian genome. We made about 250 base substitutions in 250 codons rarely used in mammalian cells to produce unmodified L1 proteins encoded from the L1 ORFs with consensus codon usage (69). All the Nat and Mod PV L1 sequences were sequenced and found to be error free; they were then cloned into the mammalian expression vector pCDNA3 containing simian virus 40 ori (Invitrogen), giving four expression plasmids, pCDNA₃HPV6b Nat L1, pCDNA₃HPV6b Mod L1, pCDNA₃BPV1 Nat L1, and pCDNA₃BPV1 Mod L1.

Cell culture and DNA transfection.

KCs were isolated from newborn mouse skin as previously described (51). Isolated KCs were grown as adherent cultures in a freshly prepared medium (365 ml DMEM medium, 2 mM glutamine, 100 U/ml penicillin, 100 U/ml streptomycin, 125 ml Ham's F12 medium, 50 ml fetal bovine serum, 2.5 mg transferrin, 2.5 mg insulin, 4.2 mg cholera toxin, 0.12 mg hydrocortisone, 17 mg adenine, 10 mg gentamicin) for 1 day and then cultured in KC-SFM medium with low calcium (GIBCO) for 7 days to induce cell differentiation. KCs cultured for 1 or 7 days were transfected with PV L1 gene expression constructs (pCDNA3 Nat HPV6b L1, pCDNA3 Mod HPV6b L1, pCDNA3 Nat BPV1 L1, and pCDNA3 Mod BPV1 L1) using Lipofectamine (Invitrogen) according to the manufacturer's protocol. After transfection, DNA-transfected KCs continued to grow in KC-SFM medium for 42 h before collection for RNA and protein preparation.

RNA Northern blot analysis.

Total RNA was extracted from L1 DNA-transfected KCs using a NucleoSpin RNAII Kit (Mackery-Nagel). For cytoplasmic RNA purification, buffer RLN (50 mM Tris [pH 8.0], 140 mM NaCl, 1.5 mM MgCl₂, 0.5% Nonidet P-40) was directly added to monolayer cells, and cells were lysed at 4°C for 5 min. After the nuclei were removed by centrifugation, cytoplasmic RNAs were purified by a QIAGEN kit (QIAGEN). Following DNase I treatment, 10- or 15-μg RNA samples were electrophoresed in 1.2% denatured agarose gels and blotted onto a Nylon N+ membrane (Amersham). The Northern blots were probed with an equal mixture of ³²P-labeled Nat and Mod PV L1 gene probes. To visualize internal controls, the Northern blots were stripped and reprobed with a ³²P-labeled glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene probe.

Reverse transcriptase PCR (RT-PCR) and quantitative RT-PCR.

A 0.5-μg sample of RNA purified from cultured KCs transfected with different PV L1 gene expression constructs was converted to cDNA using random primers and PowerScript RT (Clontech) according to the manufacturer's protocol. We used 20 ng of cDNA from each RNA sample in a 20-μl RT-PCR mixture using the FastStart DNA Master SYBR Green I kit (Roche) supplemented with 3 mM MgCl₂ and Platinum Taq polymerase (Invitrogen). Quantitative RT-PCR was undertaken using the TaqMan system (Applied Biosystems). The efficiency of amplification for each pair of primers was determined using a standard curve that was generated using serially diluted plasmid DNA. Transcription of each investigated L1 gene, mouse K14, and involucrin was compared to γ-actin (46).

Western blot analysis.

DNA-transfected KCs were collected for protein preparation 42 h posttransfection. Cell pellets were resuspended in phosphate-buffered 0.15 M sodium chloride (PBS), pH 7.4, containing 2 mM phenylmethylsulfonyl fluoride and sonicated for 40 s. Fifty-microgram total protein samples were separated by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis and blotted onto polyvinylidene difluoride membrane. The blots were first probed by monoclonal antibodies against PV L1 protein and β-tubulin. Blots were then probed with horseradish peroxidase-conjugated goat anti-mouse immunoglobulin G (IgG; Sigma) and visualized using a chemiluminescence system (Amersham).

Use of the gene gun.

Flank and belly skin of BALB/c mice 6 weeks old was shot by particle bombardment with DNA-coated gold beads coated with Nat and Mod human and bovine PV L1 gene plasmids (2 μg DNA per dose) using the helium-powered Helios gene gun delivery system at a pressure setting of 480 lb/in² based on the expression of a Mod gfp gene (Bio-Rad Laboratories, Richmond, CA). Eight mice were used for the gene gun delivery of each L1 plasmid DNA. Skin was collected 42 h after particle bombardment, fixed using 10% neutral buffered formalin, and embedded in paraffin for sectioning.

Immunofluorescence labeling of L1, K14, and involucrin protein in vitro and in vivo.

KCs were grown on eight-well chamber slides, transfected with the different plasmids, fixed, and permeabilized with 85% ethanol 42 h posttransfection. Fixed KCs were blocked with 5% skim milk-PBS and probed with monoclonal antibody against PV1 L1 protein (67), followed by Cy3-conjugated anti-mouse IgG (Sigma). L1-labeled KCs were further blocked with 5% skim milk-PBS and probed with fluorescein isothiocyanate-conjugated monoclonal antibody against keratin 14 (K14; Covance). Fixed KCs were probed with antibody against involucrin protein (Covance), followed by fluorescein isothiocyanate-conjugated secondary antibody. Nuclei were counterstained by 4′,6′-diamidino-2-phenylindole (DAPI). KCs were examined by immunofluorescence microscopy. Fixed skin section samples were similarly stained for PV L1 protein.

Isolation of differentiated and undifferentiated KCs from mouse and cow skin.

Differentiated and undifferentiated KCs were isolated from mouse and cow skin as previously described (48).

Preparation of aa-tRNAs.

Total tRNAs were extracted and purified from undifferentiated and differentiated KCs using a QIAGEN kit (QIAGEN). aa-tRNAs were produced as previously described (69).

High-pressure liquid chromatography (HPLC) analysis of tRNAs.

HPLC analysis was carried out on a Waters liquid chromatography system equipped with an LC-100 column oven, a spectrophotometric detector with a 254-nm filter, and a Waters Chromatopac (Waters). tRNA (40 to 60 μg) treated with 20% trifluoroacetic acid (TFA) and formic acid for 1 h was injected into a Luna 5-μm C₁₈ column prefitted with a 7-mm guard column (Phenomene). Elution was achieved using a 0% to 30% linear gradient of 1 M ammonium acetate-0.1% TFA, pH 3.2, and 50% acetonitrile-0.025 M potassium orthophosphate-0.1% TFA, pH 5.0, over 120 min. The chromatographic run was carried out at 37°C at a flow rate of 0.4 ml/min.

tRNA dot blot hybridization.

One hundred nanograms of tRNA, after denaturation in 1 M deionized glyoxal-20 mM NaPO₄ (pH 7.0) at 50°C for 1 h, was applied to a Nytran blot using a 24-well slot blot apparatus. The blot, after incubation in 20 mM Tris-HCl (pH 8.0) at 100°C for about 10 min, was then air dried and cross linked by 254-nm irradiation. The cross-linked blot was prehybridized with hybridization buffer containing 6× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 10× Denhardt solution, 0.2% SDS, and 1 mM EDTA at 37°C for at least 4 h. The blot was then hybridized with DNA oligonucleotide probes complementary to tRNA^Met(initiator), tRNA^Ala(CGA), tRNA^Arg(CGA), tRNA^Asp(GAC), tRNA^Asn(AAC), and tRNA^Asn(AAG) in 4× SET buffer (1× SET is 0.15 M NaCl, 0.03 M Tris HCl, and 2 mM Na₂EDTA, pH 8.0) at 37°C overnight. The DNA oligonucleotide probe complementary to mammalian tRNA^Met(initiator) is 5′-TAGCAGAGGATGGTTTC-3′, and that complementary to tRNA^Asn(AAT) is 5′-CGTCCCTGGGTGGGCTC-3′. DNA oligonucleotide probes complementary to mouse and bovine tRNA^Ala(GCA) are 5′-TAAGGACTGTAAGACTT-3′ (mouse) and 5′-TAAGGATTGCAAGACTA-3′ (bovine), those complementary to mouse and bovine tRNA^Arg(CGA) are 5′-CGAGCCAGCCAGGAGTC-3′ (mouse) and 5′-TTGGTAATTATGAATTA-3′ (bovine), those complementary to mouse and bovine tRNA^Asp(GAC) are 5′-TAAGATATATAGATTAT-3′ (mouse) and 5′-TGAGGTGTACAGGACTT-3′ (bovine), and those complementary to mouse and bovine tRNA^Asn(AAC) are 5′-CTAGATTGGCAGGAATT-3′ (mouse) and 5′-CTAGACTGGTGGGCTCC-3′ (bovine). The DNA oligomers were labeled with T4 polynucleotide kinase (Amersham) and [γ-³²P]ATP (3,000 Ci/mmol; Amersham) at the first 5′ end. Specific activities of 10⁸ to 10⁹ cpm/μg were generally reached. Approximately 10⁷ cpm of oligomers was used per blot in hybridization reactions. Blots were washed with 1× SET buffer at 37°C and autoradiographed.

Cell-free in vitro translation assay.

For in vitro translation, L1 plasmid (1 μg) was added to 20 μCi of [³⁵S]methionine (Amersham), and 40 μl of T7 DNA polymerase-coupled rabbit reticulocyte lysates (Promega), with or without additional aa-tRNAs as indicated. Translation was performed at 30°C and stopped by adding SDS loading buffer. The L1 proteins were separated by SDS-polyacrylamide gel electrophoresis on a 10% gel and blotted onto polyvinylidene difluoride membrane. The blots were imaged by phosphor screen and quantified by densitometric analysis using the ImageQuant program (Molecular Dynamics).

Codon usage analysis of mouse K14, involucrin, and HPV6 L1 genes.

DNA-coding sequences for the K14, involucrin, and HPV6 L1 genes were downloaded from the website of the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/mapview/). The codon usage pattern of the three genes was analyzed by a computer program [CodonFrequency(GCG)] at the website of the ANGIS Bioinformatic Forum (http://www.angis.org.au/html/index.html). The frequency of use of 64 codons for each gene was tabulated (available on request).

RESULTS

Codon usage determines differentiation-dependent expression of PV L1 genes in KC cultures in vitro.

We prepared primary KC cultures from newborn mouse skin and induced differentiation in vitro. Progressive differentiation of the KCs cultured from day 1 to day 7 was confirmed by microscopic appearance, increased Papanicolaou staining for cytokeratins (43), progressive loss of expression of the basal cell keratin K14 (4, 19, 44), terminally differentiation-dependent expression of involucrin (13, 40), and progressive nuclear condensation (Fig. 1A to F). KCs cultured for 1 or 7 days were transiently transfected with the Nat L1 gene of two PVs (HPV6b, BPV1). After 42 h, L1 mRNA was detected in undifferentiated and differentiated KCs, whereas L1 protein was detected only in cultures of differentiated KCs, as determined by immunofluorescence in situ (Fig. 2A and B) or by immunoblotting of cell extracts (Fig. 2C and D). In further experiments, KCs were transfected after a period of 0 to 7 days in culture, during which time progressive KC differentiation was observed. Expression of L1 protein over the 24 h following transfection increased progressively, when cells were allowed to differentiate more prior to transfection (data not shown). These results confirm that for adherent primary KC cultures, as for suspension cultures (49), expression of PV L1 proteins is differentiation dependent.

FIG. 1.

Cultured primary KCs differentiate between day 1 (D1) and day 7 (D7) in vitro. Primary in vitro murine KC cultures were exposed to differentiation medium for 7 days. At days 1 and 7, cells were compared for morphology (A) and for cellular differentiation using Papanicolaou stain (B). Expression of K14, an early differentiation marker detected by indirect immunofluorescence, is shown in green (C), and nuclei from the same field were counterstained blue with DAPI (D). Expression of involucrin, a late differentiation marker detected by indirect immunofluorescence, is shown in green (E), and nuclei from the same field were counterstained blue with DAPI (F).

FIG. 2.

Codon usage determines the timing of expression of the PV L1 capsid gene as KCs differentiate in vitro. KCs cultured for 1 day (D1) or 7 days (D7) were transfected with an expression construct for a Nat or Mod PV L1 capsid gene as indicated. (A and B) At 42 h after transfection, cells were stained by indirect immunofluorescence (A, 1 day; B, 7 days) for human PV L1 protein (red) and for K14 (green) and nuclei counterstained with DAPI (blue). Overlaid panels combine all three images. Magnification of all images, ×400. (C and D) For 1- and 7-day cultured KCs similarly transfected with Nat or Mod human (C) and bovine (D) PV L1 genes, L1 mRNA was assessed by RT-PCR using cellular actin (Act) as an internal standard and by quantitative RT-PCR. Results are shown as the mean (± the standard error of the mean) of four individual assays from two independent experiments. L1 protein levels were assessed by immunoblotting using cellular tubulin (Tub) as an internal standard.

PV L1 and L2 genes demonstrate a strong bias toward the usage of 18 codons with T at the third position that mammalian genes rarely include (68, 69). To test the hypothesis that codon usage bias might determine the differentiation-dependent translation of the L1 genes in cultured KCs, we examined the expression of two heavily codon-substituted L1 genes in differentiating KCs, as these genes, encoding unmodified L1 proteins, have been shown to be expressed well in a wide range of cells in vitro (24, 41, 69). The Mod L1 genes were each transcribed in both undifferentiated and differentiated KCs (Fig. 2C and D); L1 protein, however, was expressed in D1-transfected (undifferentiated) KCs but not in in vitro D7-transfected (differentiated) KCs (Fig. 2A and B). Thus, these data suggest that the synonymous codon use of the Nat L1 genes is adapted to expression of their protein products in KCs according to the differentiation stage of the cells.

Codon usage determines how L1 protein expression is targeted to different sites of the mouse skin in vivo.

To confirm our in vitro findings in vivo, expression of the Nat and Mod L1 genes in mouse skin was investigated following gene delivery by DNA particle bombardment. A gfp gene encoding green fluorescent protein modified for expression in mammalian cells, and known to be expressed well in all layers of skin, was used as a control for DNA particle bombardment (Fig. 3B). Using the same helium pressure setting, both Nat and Mod L1 genes were delivered to mouse skin. RNA in situ hybridization revealed that the L1 genes, whether Nat or Mod, were each transcribed in both basal and superficial epithelial cells (data not shown). However, in keeping with the in vitro findings, L1 protein was expressed from the Nat sequence L1 genes only in the most superficial epithelial cells (Fig. 3C and D), including superficial differentiated KCs in deeper hair follicles (Fig. 3C and D). In contrast, the Mod L1 genes were expressed more extensively throughout the epidermis and dermis, except for the most superficial KC layers (Fig. 3E and F). Thus, the results indicate that the codon usage of the PV L1 gene determines the site of expression of L1 protein within the mouse epidermis in vivo.

FIG. 3.

Codon usage determines the spatial expression of the PV L1 capsid gene in skin in vivo. Expression constructs for Nat (C and D) or Mod (E and F) HPV6 and BPV1 L1 genes or for humanized (hm) green fluorescent protein (B) were delivered on gold beads (2 μg/dose) to the flank skin of BALB/c mice using a biolistic delivery system (Helios; Bio-Rad Laboratories, Richmond, CA). Control skin (A) was left untreated. After 42 h, skin was fixed in 10% neutral buffered formalin, sectioned, and stained as indicated for L1 protein by indirect immunofluorescence (red) or for nuclei using DAPI (blue). Epidermis (E), dermis (D), and hair follicles (H) of the skin section are indicated in panel A. Magnification of all sections, ×400.

Codon modification does not affect the stability of PV L1 mRNA in undifferentiated and differentiated KCs.

The above results indicate that the site of expression of L1 protein from either Nat or Mod PV L1 genes is determined by a differentiation-linked process apparently acting at a posttranscriptional level. Regulation of mRNA stability is an important element in the posttranscriptional control of gene expression (37). Accordingly, we examined the stability of the Nat and Mod PV L1 mRNAs in undifferentiated and differentiated KCs. KCs cultured for 1 or 7 days were transiently transfected with the Nat or Mod L1 genes of two PVs (HPV6b, BPV1). At 18 h posttransfection, the KCs, treated with actinomycin D (Act D) to prevent further gene transcription for 0, 2, 5, or 8 h, were harvested for total RNA preparation. L1 mRNA was detected by Northern blot analysis and compared with GAPDH mRNA (Fig. 4A and B, insets). As shown by quantitative RT-PCR (Fig. 2C and D), the steady-state levels of L1 mRNA transcribed from the Mod PV L1 genes were several times higher than the levels expressed from Nat PV L1 genes (Fig. 4A and B, insets). The half-lives of both Nat and Mod L1 mRNAs were then calculated as the time period necessary to reduce the amount of L1 mRNA in Act D-treated KCs to 50% of the original L1 mRNA abundance at time point zero (0 h) (Fig. 4A and B). In D1-transfected undifferentiated KCs, the half-lives of the Nat L1 mRNAs were 5.3 ± 0.8 h for HPV6b and 5.2 ± 0.7 h for BPV1, whereas those of the Mod L1 mRNAs were nonsignificantly shorter at 4.7 ± 0.8 h for HPV6b and 4.8 ± 0.8 h for BPV1 (Fig. 4A and B). In D7-transfected differentiated KCs, the half-lives of the Nat L1 mRNAs were 4.3 ± 0.4 h for HPV6b and 4.4 ± 0.6 h for BPV1 and the half-lives of the Mod L1 mRNAs were nonsignificantly longer at 5.0 ± 0.7 h for HPV6b and 4.6 ± 0.4 h for BPV1 (Fig. 4A and B). Thus, codon modification does not alter the stability of PV L1 mRNAs in cultured KCs sufficiently to explain the observed differences in steady-state L1 protein levels.

FIG. 4.

Half-lives of L1 mRNAs transcribed from the Nat or Mod PV L1 capsid gene in cultured KCs. KCs cultured for 1 day (D1) or 7 days (D7) were transfected with an expression construct for a Nat or Mod PV L1 capsid gene as indicated. At 18 h after transfection, cells were treated with Act D (500 ng/ml) and total RNAs were harvested at 0, 2, 5, and 8 h for Northern blot hybridization and mRNA stability analysis. Fifteen micrograms of total RNA from Nat PV L1-transfected KCs (10 μg of total RNA from Mod PV L1-transfected KCs) treated with DNase I was electrophoresed on a 1.2% denatured agarose gel and blotted onto a nylon membrane. The Northern blots were probed with an equal mixture of ³²P-labeled Nat and Mod PV L1 gene probes. To visualize the internal controls, the Northern blots were stripped and reprobed with a ³²P-labeled GAPDH gene probe. (A, B) Densitometric quantification of Nat and Mod HPV6b and BPV1 L1 mRNA levels following normalization to GAPDH. The half-lives of Nat and Mod HPV6b and BPV1 L1 mRNAs, calculated as the times required for a given transcript to decrease to 50% of its initial abundance (horizontal dotted lines), are shown in parentheses. The data shown are the means (± the standard error of the mean) of four individual blotting assays from two independent experiments. Insets are representative Northern blot analyses (n = 4) of Nat and Mod HPV6b and BPV1 L1 and GAPDH RNA expression in day 1 and day 7 L1-transfected KCs. Statistical analysis of the results from the experiments was conducted using one-way analysis of variance. No significant difference between Nat and Mod L1 half-lives was obtained (P > 0.05).

As the transport of mRNAs to the cytoplasm is delayed until RNA splicing is completed (30, 56), we repeated, using cytoplasmic mRNA, the L1 mRNA half-life studies in differentiated KCs and quantitated mRNA by Northern blotting hybridization (Fig. 5A and B) and quantitative RT-PCR (data not shown). L1 mRNAs, whether transcribed from Nat or Mod PV L1 genes, were efficiently exported from the nucleus to the cytoplasm of KCs (Fig. 5A and B). The half-lives of the cytoplasmic Nat L1 mRNAs were 3.9 ± 0.7 h for HPV6b and 5.2 ± 0.7 h for BPV1, whereas those of the Mod L1 mRNAs were nonsignificantly different from the Nat mRNAs at 4.3 ± 0.5 h for HPV6b and 5.5 ± 0.8 h for BPV1 (Fig. 5A and B). Thus, differences in the cytoplasmic pools of mRNA are unlikely to account for the observed differences in L1 production from Nat and Mod L1 genes in differentiated cells.

FIG. 5.

Stability of cytoplasmic L1 mRNAs transcribed from the Nat or Mod PV L1 capsid genes in differentiated KCs. KCs cultured for 7 days (D7) were transfected with an expression construct for a Nat or Mod PV L1 capsid gene as indicated. At 18 h after transfection, KCs were treated with Act D (500 ng/ml) and cytoplasmic RNAs were prepared at 0, 2, 5, and 8 h for Northern blot hybridization and mRNA stability analysis. Ten micrograms of cytoplasmic RNA treated with DNase I was electrophoresed on a 1.2% denatured agarose gel and blotted onto a nylon membrane. The Northern blots were probed with an equal mixture of ³²P-labeled Nat and Mod PV L1 gene probes. As internal controls, the Northern blots were stripped and reprobed with a ³²P-labeled GAPDH gene probe. (A and B, left) Representative Northern blot analysis (n = 2) of Nat and Mod HPV6b and BPV1 cytoplasmic L1 and GAPDH RNA expression in day 7 L1-transfected KCs. (Right) Half-lives of Nat and Mod HPV6b and BPV1 cytoplasmic L1 mRNAs assessed by densitometric quantification following normalization to GAPDH. Data shown in parentheses are the means (± the standard error of the mean) of duplicate blot assays.

tRNA profiles differ between undifferentiated and differentiated KCs.

To explain the observation that codon modification of the L1 gene could alter the expression of L1 protein in differentiating KCs without altering mRNA levels or stability, we hypothesized that observed translational differences likely reflected different availability of aa-tRNAs between differentiated and undifferentiated KCs. We therefore extracted tRNAs from murine and bovine epidermal cells, sorted by size and buoyant density into smaller undifferentiated cells and larger differentiated cells, and confirmed the differentiation state of the cells by K14 and involucrin staining and morphology (Fig. 6). tRNA species were then profiled for each cell population by HPLC separation (Fig. 6). tRNA profiles of murine and bovine epithelial cells were similar (Fig. 6A, B, C, and D), and the profiles of differentiated cells were in each case distinct from undifferentiated cells (Fig. 6), showing, as for studies with tumor cells and cells of parent tissues (34), that cell differentiation can alter the aa-tRNA pool. As no validated technology is available for identification and quantitation of individual tRNA species in cells by HPLC or otherwise, we arbitrarily chose several tRNA oligonucleotide probes complementary to specific regions of six tRNAs, tRNA^Met(Initiator), tRNA^Ala(CGA), tRNA^Arg(CGA), tRNA^Asp(GAC), tRNA^Asn(AAC), and tRNA^Asn(AAT), to examine whether reactivity with these probes differed between undifferentiated and differentiated KCs. Reactivity with the tRNA^Met(Initiator) probe was similar between undifferentiated and differentiated KCs (Fig. 6E). Reactivity with the probes corresponding to four tRNAs, tRNA^Ala(GCA), tRNA^Arg(CGA), tRNA^Asp(GAC), and tRNA^Asn(AAC), was consistently stronger in undifferentiated KCs than in differentiated KCs (Fig. 6E), and reactivity with tRNA^Asn(AAT) was consistently weaker (Fig. 6E). Thus, HPLC tRNA profile differences between differentiated and undifferentiated KCs are reflected in differences in hybridization-determined tRNA reactivity.

FIG. 6.

tRNA profiles and dot blot-examined tRNA species differ between differentiated and undifferentiated KCs. (A to D) Total tRNA was extracted, using a kit (QIAGEN), from single-cell suspensions derived by collagenase and trypsin digestion from murine flank skin (A and B) and bovine ear skin (C and D) and separated into basal (undifferentiated) cells (A and C, insets) and squamous (differentiated) cells (B and D, insets) by density flotation on a discontinuous 30% to 70% Percoll gradient. The differentiation state of the isolated undifferentiated and differentiated cells was confirmed by K14 and involucrin (Invol.) staining (A and B). Nucl., nucleus. Extracted tRNAs were dissolved in water, and HPLC retention profiles were determined using a Luna 5-μm C₁₈ column (Phenomene) and a linear elution gradient of acetonitrile up to 30% over 2 h at 37°C. (E) tRNA dot blot hybridization of tRNA samples from undifferentiated (Undi.) and differentiated (Diff.) KCs. The details are described in Materials and Methods.

Translation of Nat PV L1 genes is preferentially enhanced by aa-tRNAs from differentiated KCs.

We have previously shown that altering the available pool of aa-tRNAs alters the expression of a Nat L1 gene in cell-free in vitro translational studies (69). To show directly that the aa-tRNA populations of differentiated KCs favor translation of Nat L1, we firstly confirmed that translation of L1 protein from Nat L1 genes was much slower than that from the Mod L1 genes (Fig. 7A) using a rabbit reticulocyte lysate cell-free in vitro translation system that has fixed pools of aa-tRNAs and amino acids. We then examined whether addition of aa-tRNAs from undifferentiated or differentiated murine KCs could enhance the translation of the Nat and Mod L1 genes (Fig. 7B). Introduction of exogenous tRNAs from differentiated KCs but not undifferentiated KCs enhanced the translation of Nat L1 genes in a dose-dependent manner, with optimum efficiency at 10⁻⁷ M (Fig. 7B). aa-tRNA from differentiated KCs at 10⁻⁷ M significantly enhanced the translation of L1 3.1-fold ± 0.7-fold (standard error of the mean) for HPV6b and 2.5-fold ± 0.9-fold for BPV1 at 12 min, whereas 10⁻⁷ M aa-tRNA from undifferentiated KCs gave less-significant enhancement (1.7-fold ± 0.5-fold increase for HPV6b and 1.4-fold ± 0.3-fold for BPV) compared to control reaction mixtures without added aa-tRNAs (Fig. 7C). Supplementation of the aa-tRNAs from undifferentiated or differentiated murine KCs had no significant effect on the translation of the Mod L1 genes (data not shown). We thus conclude that aa-tRNA availability changes in KCs with cellular differentiation, and the aa-tRNA pool in differentiated KCs enables more-efficient translation of the “rarely used” codons in the Nat L1 gene than the aa-tRNA pool in undifferentiated KCs.

FIG. 7.

Translation of PV L1 capsid genes is preferentially enhanced by aa-tRNA from differentiated KCs. L1 genes from HPV6 and BPV1 were translated in a cell-free system using rabbit reticulocyte lysate (in the presence of [³⁵S] methionine. Production of ³⁵S-labeled L1 was assessed by autoradiography of polyacrylamide gel electrophoresis-separated proteins. (A). Production over time of L1 protein from Nat L1 genes (HPV6b and BPV1) is less rapid than from Mod L1 genes. (B). Supplementation of the in vitro translation reaction mixture with aa-tRNA from differentiated (Diff.) or undifferentiated (Undi.) KCs enhances L1 production from Nat L1 genes in a dose-dependent manner from 10⁻⁸ to 10⁻⁶ M; inhibition of translation is evident at the highest concentrations of aa-tRNA. (C) Addition of aa-tRNAs prepared from differentiated KCs (10⁻⁷ M) increased the rate of translation of Nat L1 genes more effectively than aa-tRNAs from undifferentiated KCs over a time course of 8 to 16 min. (D) Means (± the standard error of the mean) of L1 production from three independent experiments quantitated by scanning densitometry of autoradiographs. Statistical analysis of the results from individual time points from these experiments was conducted using analysis of variance (P < 0.05 and P < 0.01 compared with no addition of aa-tRNAs as shown).

Codon usage may also regulate the expression of some mammalian proteins associated with KC differentiation.

To investigate whether differentiation-dependent expression of mammalian proteins might in part be determined by their codon composition and by tRNA abundance differences associated with cellular differentiation, we examined the expression of the K14 and involucrin genes in day 1 and day 7 cultured KCs by RT-PCR and immunoblotting (Fig. 8A and B). As shown by RT-PCR, the mRNA transcripts of both K14 and involucrin were at a steady-state level in day 1 and day 7 cultured KCs (Fig. 8A and B). However, immunoblotting of day 7 cultured KCs revealed that expression of K14 protein was down-regulated, in contrast to the significantly up-regulated expression of involucrin (Fig. 8A and B), consistent with the observations of immunofluorescence in situ (Fig. 1C and E). We examined the codon usage of mouse K14 and involucrin genes in comparison with the HPV6 Nat L1 gene by tabulating the percentage of codons with G/C or A/T at the third position for each gene (Table 1). The K14 gene has a significantly greater usage of third-position G/C residues than the involucrin gene (Table 1), while the HPV6 L1 Nat gene has the least use of third-position G/C residues (Table 1). The involucrin protein includes several glutamine residues, 102 of which are encoded by a Glu(GAG) codon. If these glutamine residues are excluded, the involucrin gene has as high a percentage of AT-ending codons (56/100) as the HPV6 L1 Nat gene. In view of this finding and tRNA dot blot hybridization data showing that undifferentiated KCs are apparently rich in tRNA-Asp(GAC) and tRNA-Asn(AAC) (Fig. 6E), we compared the codon usage of the three genes for the amino acids Asp and Asn (Table 2). The K14 gene, in contrast to the involucrin and HPV6 Nat L1 genes, has a high percentage of GAC codons for amino acid Asp, supporting our finding that tRNA-Asp(GAC) abundance correlates with high-level expression of K14 protein in undifferentiated KCs (Fig. 6E and 8A). Thus, at least two mammalian genes differentially expressed between differentiated and undifferentiated KCs may use a subset of favored codons matching the abundance of the isoacceptor tRNAs to regulate expression of the corresponding proteins during KC differentiation.

FIG. 8.

mRNA translation efficiency of KC structural genes differs with KC differentiation. KCs cultured for 1 day (D1) or 7 days (D7) were harvested for RNA and protein preparation. K14 and involucrin (Invol) mRNAs were assessed by RT-PCR and proteins by immunoblotting using actin as an internal standard. The results are representative of two independent experiments (n = 4) in day 1 and day 7 cultured KCs.

TABLE 1.

Comparison of codon usage between K14, involucrin, and HPV6 L1^a

Codon usage	n (%), P value
	K14	Involucrin	HPV6 L1
Total	511 (100)	467 (100)	500 (100)
GC ending	337 (66)	268 (57)	174 (35), 0.0000 (K14-L1)
AT ending	174 (34)	199 (43), 0.0045 (K14-Inv)	326 (65), 0.0000 (Inv-L1)

^aCodon usage for the whole gene sequence: the number of codons, third-position GC codons (GC ending), and AT codons (AT ending) (n), and their frequencies (%) are shown. P values reflect whether or not two genes differ in their expected usage of GC-ending codons and AT-ending codons (Fisher exact test).

TABLE 2.

Usage of codons corresponding to abundant aa-tRNAs in KCs^a

Amino acid	Codon	n (%), P value
		K14	Involucrin	HPV6 L1
Asp	GAT	7 (33)	10 (77),0.0323 (K14-Inv)	20 (77), 0.0036 (K14-L1)
	GAC	14 (67)	3 (23)	6 (23), 1.0000 (Inv-L1)
Asn	AAT	8 (53)	0 (0), 0.4706 (K14-Inv)	14 (54), 0.9999 (K14-L1)
	AAC	7 (47)	2 (100)	12 (46), 0.4815 (Inv-L1)

^aFor each codon, we report the total number in each gene (n) and the relative frequency (%). The P values reflect whether the two compared genes differ in their expected encoding of the amino acids (Fisher exact test).

DISCUSSION

We demonstrate here that KCs differentiated in vitro and displaying KC differentiation markers produce PV L1 protein more efficiently from the L1 gene than undifferentiated KCs. Further, mRNA quantitation demonstrates that the difference in L1 protein expression between differentiated and undifferentiated KCs is largely determined posttranscriptionally. Generalized substitution, within the L1 gene, of isoencoding codons with a G or a C at redundant positions within the codon triplet enhances expression of L1 protein in undifferentiated KCs and impairs expression in differentiated KCs. In vivo observations in epidermis transduced with Nat or Mod L1 are consistent with our in vitro data and with the specific expression of PV L1 protein in differentiated KCs observed in the course of natural PV infection.

Several posttranscriptional mechanisms for regulation of expression of broad subsets of genes on cell differentiation are recognized, including changes in the assembly of the eukaryotic translation-initiation factor complex and expression of specific RNA binding proteins targeting cis-regulatory sequences associated with genes expressed on differentiation (reviewed by Calkhoven et al. [6]) (47). However, we show here that provision of aa-tRNAs from differentiated KCs but not undifferentiated KCs enhances the translational efficiency of the Nat L1 gene in a cell-free system and has no effect on translation using the same expression vector of a Mod L1 gene which, as a result of codon modification, is well expressed in undifferentiated KCs. This observation is difficult to explain as a consequence of the generally active posttranslational mechanisms for regulation of gene expression discussed by Calkhoven et al. as associated with cell differentiation and more strongly supports a hypothesis that aa-tRNA changes associated with cell differentiation may, in association with selective codon usage, regulate the translation of some genes within a cell lineage.

Replacement of less-preferred codons within a prokaryotic gene with synonyms more commonly observed in mammalian genes can greatly increase gene expression in eukaryotic cells, with improvement in translational efficiency attributed to correction of a mismatch between eukaryotic cell tRNA pools and preferred prokaryotic gene codon bias (26, 39, 61, 70). Plotkin et al., using in silico analysis of codon usage in eukaryotic genes differentially expressed in uterus and testis tissues and in brain and liver tissues (47), have recently hypothesized that posttranscriptional controls based on gene codon usage may also play an important role in the regulation of tissue-specific gene expression in mammals. Further, impaired viral protein expression in mammalian host cells, attributed to codon usage divergence of viral and human genes, is observed for the latent genes of Epstein-Barr virus (31), the env gene of human immunodeficiency virus type 1 (25), and the E7 (9), L1, and L2 genes (69) of PV. Thus, the codon composition of PV L1 genes might be expected to influence L1 expression according to the availability of aa-tRNA within the cell. The correlation observed in the present study between differences in aa-tRNA species and differential L1 gene expression for differentiated and undifferentiated KCs may have broad significance for selective gene expression within multicellular eukaryotic organisms. In Escherichia coli and in yeasts, highly expressed genes use a subset of codons corresponding to the highly expressed isoacceptor tRNAs (3, 23) and the synthesis of a number of colicins is linked to the difference in tRNA availability for the various codons used by the relevant genes (60). More generally, the pool of available aa-tRNAs is held to be rate limiting for accuracy and efficiency of gene translation (18, 27, 28, 59, 66). However, in multicellular eukaryotes, there are very few experimental data on tRNA abundance (15), though a relationship is reported between tRNA abundance and codon usage in Drosophila (45). Reliable methods for accurately measuring the range of specific tRNAs in eukaryotic cells and for distinguishing among free tRNA, free aa-tRNA, and ribosome-associated aa-tRNA have not yet been developed. Our tRNA dot blot hybridization data for tRNA abundance are limited to a small subset of tRNAs and cannot distinguish tRNA from aa-tRNA. These data show higher total tRNA^Asp(GAC) and tRNA^Asn(AAC) levels in undifferentiated KCs than in differentiated KCs. The ^Asp(GAT) and ^Asn(AAT) codons, which are abundant in natural L1 genes (68, 69), are substituted in the Mod L1 genes with ^Asp(GAC) and ^Asn(AAC). Better translation of the Mod L1 genes in undifferentiated KCs is thus consistent with the tRNA abundance data and suggests that the ^Asp(GAT) and ^Asn(AAT) aa-tRNAs may be rate limiting for expression of L1 in differentiated KCs. Further correlations of specific aa-tRNA abundance with gene translation efficacy await the development of better methods for measuring the complete set of aa-tRNAs.

Alternate methods for regulation of PV L1 expression in differentiating epithelium have been proposed. The PV late promoter, from which the L1 gene is naturally expressed, is more active in differentiated than undifferentiated epithelium (2, 33), though studies with raft cultures suggest that late gene translation in differentiated cells requires the viral genome to be extrachromosomal (21, 33), suggesting a more complex model of regulation involving additional posttranscriptional controls. Regulation of mRNA stability or decay is an important control point for gene expression and is mediated by nucleotide sequence elements, specific cellular protein factors, and endoribonucleases (63). Duan and Antezana (14) reported that synonymous codon substitution in the coding region of gene DRD2 can affect mRNA stability. Many studies have described determinants of mRNA stability in the coding region in a range of genes (7, 8, 12, 58, 65), including the HPV16 L2 gene (55). It therefore seems important to consider the extent to which mRNA stability contributes to differentiation-dependent differences in PV L1 gene expression observed in KCs. HPV1 mRNAs containing an AU-rich noncoding sequence which, in sense orientation, reduces their half-life: mRNA destabilization requires multiple RNA binding proteins (54). Described mRNA stability elements in PV L1 and L2 genes are mainly located in the PV-encoded 5′ and 3′ untranslated regions containing AU-rich sequences (32, 54, 62), and these would not be expected in our engineered L1 expression constructs. Further, the effects of stability sequences on mRNA stability, at least for BPV1 L1, appear to be smaller in magnitude (22) than our observed difference in L1 translation efficiency between differentiated and undifferentiated KCs, suggesting that even if our L1 constructs encode the RNA sequences that bind destabilizing RNA binding proteins, and these proteins are only present in undifferentiated KCs, they are still insufficient to explain our observed difference in L1 expression with KC differentiation. cis-acting regulatory sequences have been described in the HPV16 L1 gene 3′ coding sequence (11) but have not been shown to be effective in epithelium or, more significantly, to be selectively lost in KCs on differentiation, as would be required to explain specific expression of PV L1 protein in differentiated KCs. Rather, the present study shows that Nat HPV16 L1 mRNAs are of similar stability in differentiated and undifferentiated KCs, as are mRNAs from Mod L1. Thus, it is unlikely that L1 mRNA stability is a major determinant of the differentiation-determined selective posttranscriptional block to L1 capsid protein synthesis we observed for Nat and Mod PV L1 transcripts in cultured KCs. Further, the half-life of the less efficiently translated Nat and Mod PV L1 transcripts in KCs appears paradoxically to be somewhat longer than for the more efficiently translated L1 mRNAs. The stability of some mRNAs is affected by translation (52, 57, 64). Based on the present observations, the stability of PV L1 transcripts, rather than being a major determinant of efficient L1 translation, may be regulated in part by L1 mRNA translation efficiency, determined in KCs by a differentiation-determined match between L1 codon usage and the availability of aa-tRNAs.

In the present study, HPLC analysis has revealed that tRNA profiles of differentiated cells were in each case distinct from undifferentiated cells in both murine and bovine epithelium. What mechanism may drive the difference in tRNA profiles between undifferentiated and differentiated KCs? Epidermal KCs are highly specialized epithelial cells designed to perform a very specific function, separation of the organism from its environment. During the KC differentiation process, numerous genes are turned on and off at specific stages (16, 17). One possibility is therefore that the observed changes in the available aa-tRNA pool between differentiated KCs and undifferentiated KCs reflects regulated tRNA production or aminoacylation of tRNAs with differentiation. Alternatively, the pool of free aa-tRNAs might be determined by the extent to which particular aa-tRNAs are needed by the cell for protein production, as the majority of charged tRNAs are associated with nascent protein production in the ribosome in metabolically active cells (29). Further elucidation of the mechanism varying the pool of aa-tRNA with epithelial differentiation must await better methods of quantitating specific aa-tRNA species change in real time.