Novel Action of Retinoic Acid STABILIZATION OF NEWLY SYNTHESIZED ALKALINE PHOSPHATASE TRANSCRIPTS*

Several observations led us to investigate the possibil- ity that retinoic acid achieved its marked induction of alkaline phosphatase gene expression through a post- transcriptional effect in the nuclei of clonal rat pre-os-teoblastic UMR 201 cells. The steady-state level of alka- line phosphatase mRNA was significantly stimulated by retinoic acid. Although nuclear run-on analysis showed that M retinoic acid caused an increase in alkaline phosphatase gene transcription, this was transient compared with the rise in alkaline phosphatase mRNAwhich continued to accumulate for many hours after retinoic acid stimulation of gene transcription had ceased. More- over, the modest increase in transcriptional rate (-2-fold) was not sufficient to account for the magnitude of the rise in mRNAleve1. In order, therefore, to examine the influence of retinoic acid on nuclear processing events, a cellular subfractionation method was applied. fraction of

Several observations led us to investigate the possibility that retinoic acid achieved its marked induction of alkaline phosphatase gene expression through a posttranscriptional effect in the nuclei of clonal rat pre-osteoblastic U M R 201 cells. The steady-state level of alkaline phosphatase mRNA was significantly stimulated by retinoic acid. Although nuclear run-on analysis showed that M retinoic acid caused an increase in alkaline phosphatase gene transcription, this was transient compared with the rise in alkaline phosphatase mRNAwhich continued to accumulate for many hours after retinoic acid stimulation of gene transcription had ceased. Moreover, the modest increase in transcriptional rate (-2fold) was not sufficient to account for the magnitude of the rise in mRNAleve1. In order, therefore, to examine the influence of retinoic acid on nuclear processing events, a cellular subfractionation method was applied.
By nuclease protection analysis, and also by using reverse transcription-polymerase chain reaction, sequences corresponding to intron 2 and intron 4, respectively, were demonstrated specifically in the nuclear matrix fraction of both control and retinoic acid-treated cells. Mature (spliced) alkaline phosphatase mRNA accumulated in the non-matrix (DNase h a l t eluate, nuclear membrane) and cytoplasmic fractions of retinoic acid-treated cells at more than 100-fold greater levels than in control cells. This implies that nuclear processing of the primary RNA transcript occurred only in cells treated with retinoic acid. The post-transcriptional action of retinoic acid was inhibited by cotreatment with 10 pg/ml cycloheximide.
Transforming growth factor p (TGFP) (1 ng/ml) did not influence whole cell alkaline phosphatase levels in UMR 201 cells. Nevertheless, TGFp increased the transcriptional rate of the alkaline phosphatase gene. Although precursor mRNAwas detected in the nuclear matrix fraction of TGFP-treated cells, there was no evidence of further mRNA nuclear processing. The data are consistent with stabilization of nascent alkaline phosphatase mRNA chains by retinoic acid treatment and suggests that regulation of mRNA processing can be independent of gene transcription. This study demonstrates a novel post-transcriptional action of retinoic acid which plays an important, if not a dominant role, in determining the steady-state level of alkaline phosphatase mRNA.
tional Health and Medical Research Council (Australia). Portions of this * This work was supported by a Program Grant 923214 from the Nawork were presented at the 15th annual meeting ofthe American Society of Bone and Mineral Research, September [18][19][20][21][22]1993, Tampa, FL. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18  Alkaline phosphatase is a marker of the mature, differentiated osteoblast (1). Functionally, alkaline phosphatase is likely to be involved in bone mineralization. This is exemplified by the clinical entity of hypophosphatasia where a single point mutation in the alkaline phosphatase gene in affected subjects results in low levels of alkaline phosphatase activity associated with poor bone mineralization (2,3). In bone, only mature osteoblasts express high levels of alkaline phosphatase activity and the expression of alkaline phosphatase mRNA precedes mineralization of osteoblasts in long term culture (4).
There are three isoforms of alkaline phosphatase. Placental and intestinal isoforms are tissue-specific while the liverhone/ kidney alkaline phosphatase isoform is found in almost all tissues, albeit a t a much higher level in osteoblasts (5). Kiledjian and Kadesch (6) carried out a detailed comparison of human liverhonekidney alkaline phosphatase gene expression between osteoblast-like osteosarcoma Saos-2 cells, which express high levels of liverhonekidney alkaline phosphatase mRNA, and HepG2 hepatoblastoma cells, whose levels of the same mRNA are approximately 1000-fold lower. Their results showed that not only were the same promoter sequences utilized to initiate transcription of the gene, but there was also no difference in the transcriptional rate or in the stability of the cytoplasmic mRNA in both cell types. It was concluded that differential expression of alkaline phosphatase mRNA was controlled at a very early step post-transcriptionally, possibly involving the specific destabilization of the nascent liverhone/ kidney alkaline phosphatase mRNA in the non-osteoblastic HepG2 cells.
The results above raise an important question about the possible levels at which alkaline phosphatase gene expression can be controlled in committed preosteoblasts when they are induced to differentiate from a state of low alkaline phosphatase expression to one more typical of mature osteoblasts. In this study, we examined the regulation of gene expression of alkaline phosphatase in a neonatal rat calvarial-derived clonal cell line, UMR 201, which possesses the phenotypic characteristics of preosteoblasts. These cells express a very low basal alkaline phosphatase activity which is massively increased by retinoic acid treatment in parallel with a significant induction of alkaline phosphatase mRNA expression (7,8). Retinoic acid has also been shown to up-regulate alkaline phosphatase gene expression in another preosteoblastic cell line (RCT-1) (9). Retinoic acid treatment of UMR 201 cells also stimulated the expression of other osteoblast-related mRNA species such as osteonectin, osteopontin, matrix gla-protein, and pro-al(I) collagen (10,111, consistent with a differentiating action of this agent. In the embryo, retinoic acid has been implicated as the natural morphogen responsible for pattern formation in developing chick limb buds (12-141, acting via receptors recognized as members of the steroid and thyroid hormone receptor superfamily that have been thought to act primarily through the direct modulation of gene transcription (15). However, there is 22433 recent evidence that steroids are capable of influencing steadystate mRNA levels through additional post-transcriptional mechanisms. For instance, glucocorticoids can stabilize human growth hormone mRNA (16) or destabilize interleukin 1p mRNA (17). Estrogens have also been shown to stabilize the mRNA corresponding to the genes for the very low density apolipoprotein I1 and vitellogenin (18,19).
Alkaline phosphatase mRNA cannot be detected by Northern blot analysis in untreated UMR 201 or RCT-1 cells, but is greatly induced in both cell types by retinoic acid treatment (7-9). In contrast, transcriptional assays clearly showed that the alkaline phosphatase gene is constitutively expressed in RCT-1 cells, but retinoic acid treatment resulted in no more than a 2.5-fold increase in its transcriptional rate (20). Such a transcriptional response is insufficient to account for the magnitude of the increase in steady-state levels of alkaline phosphatase mRNA, and a similar behavior of UMR 201 cells is shown in the present work. The same discrepancy between a small increase in transcriptional activity of the alkaline phosphatase gene and subsequent mRNA accumulation in response to treatment with retinoic acid has been reported in F9 teratocarcinoma cells (21). These observations draw attention to the possible regulatory contribution of post-transcriptional events in the nucleus. Apart from the initiation of transcription, potential control points include the stabilization and processing of nascent RNA, the release of mature mRNA from the nuclear matrix, and its translocation from the nucleus to the cytoplasm through the nuclear pore complex, before associating with the cytoskeleton where translation occurs (22)(23)(24). Until recently, most studies of the regulation of eukaryotic mRNA levels have focused on the control of gene transcription or the regulation of cytoplasmic mRNA stability. The role of hormones, growth factors, or cytokines in the control of nuclear RNA processing and nucleocytoplasmic transport of mRNA remains largely unexplored.
We have explored the post-transcriptional action of retinoic acid in alkaline phosphatase gene expression by fractionating the nuclear compartment to isolate RNA from the nuclear matrix, DNase Vhigh salt eluate, and nuclear membrane in order to determine the influence of retinoic acid on the nuclear processing of alkaline phosphatase RNA through these various fractions. Our results show, for the first time, that retinoic acid plays a significant role in stabilization of the nascent transcript to enable nuclear processing of alkaline phosphatase mRNA, in addition to its action on the transcription of the gene.

Stabilized by Retinoic Acid
Cell Cultures-UMR 201 cells were routinely grown in a-MEM containing 10% FBS. Incubation was carried out at 37 "C in a humidified atmosphere equilibrated with 5% CO, in air.
Nuclear Danscriptional Run-on Assays-Transcriptional rates of alkaline phosphatase and actin genes were determined by nuclear run-on analysis. UMR 201 cells were grown in 500 cm2 trays in a-MEM containing 10% FBS until 80% confluent, before treatment with 10" M retinoic acid in a-MEM containing 2% vitamin A-deficient FBS for the indicated incubation times. Approximately 1 to 2 x IO7 nuclei were isolated by gentle homogenization of cells in a glass homogenizer, and nascent RNA transcripts were allowed to elongate in the presence of 100 pCi of [a-32PlUTP (-800 Cilmmol; Amersham Corp.) using a previously described method (26) with minor modifications as follows. RNA was extracted with an equal volume of phenol/chloroform/isoamyl alcohol (24:l) and precipitated with 2 volumes of absolute ethanol overnight at -20 "C. The RNA pellet was washed with 80% ethanol, dissolved in guanidinium thiocyanate, and precipitated with isopropyl alcohol. The cDNAs for alkaline phosphatase (10 pg) and actin (5 pg) were denatured and immobilized on nitrocellulose (Hybond-C, Amersham) using a slotblot apparatus. Prehybridization was carried out at 42 "C for 18 h in hybridization buffer containing 50% formamide, 5 x SSC (1 x SSC = 0.15 M NaCl, 0.015 M sodium citrate, pH 7),0.1% SDS, 10 mM EDTA, 100 mM Tris-HC1 (pH 7.51, 4 x Denhardt's solution, 50 pg/ml transfer RNA, 10 pg/ml polyadenylic acid (Sigma), and 50 pg/ml denatured salmon sperm DNA. 32P-Labeled RNA was then added and allowed to hybridize to target sequences for 72 h at 42 "C in the same hybridization buffer. Filters were washed in 2 x SSC a t 65 "C for 30 min, then treated a t 37 "C for 15 min with 2 x SSC containing 10 p g h l RNase A, followed by a final wash in 2 x SSC at 37 "C. Filters were then autoradiographed a t -70 "C with intensifying screens. Quantitation of appropriate exposures was carried out by densitometry (Molecular Dynamics, model 300A) and normalized by reference to the actin signal on the same filter.
Northern Blot Analysis-Cytoplasmic RNA was isolated from cells during preparation of nuclei for nuclear run-on analysis. After the cells were homogenized and the nuclei pelleted by centrifugation, the supernatant containing cytoplasmic RNA was added to 4 ml of protein denaturing buffer containing 7 M urea, 350 mM NaCl, 10 mM EDTA, 1% SDS, and 10 mM Tris-HC1, pH 7.4 (27). RNA was extracted by adding 8 ml of phenol/chlorofodisoamylalcohol (50:50:1), and precipitated with 8 ml of isopropyl alcohol. The RNA pellet was washed once with 70% ethanol and resuspended in TE buffer (10 mM Tris-HC1, pH 8.0,l nm EDTA, pH 8.0) at a concentration of 2 pglpl. The samples were used fresh or else stored at -70 "C. Total RNA was separated in a 1.5% agarose-formaldehyde gel and transferred to nylon filters (28). Filters (Hybond-N, Amersham) were hybridized overnight in SSPE buffer as described previously (11). cDNA probes were nick-translated with [a-32PldCTP to a specific activity of 1 x IO9 disintegrationdmidpg DNA (Boehringer Mannheim). The filters were washed sequentially in 2 x SSPE (1 X SSPE = 150 mM NaCl, 10 mM NaH,PO,H,O, 1 mM sodium EDTA, pH 7.4) with 0.1% SDS at 42 "C for 15 min, 1 x SSPE with 0.1% SDS at 65 "C for 30 min and finally 0.1 x SSPE with 0.1% SDS at room temperature for 15 min. Specifically bound probe was visualized by autoradiography and quantified by densitometry. Relative mRNAlevels were normalized for loading variability by comparison with actin mRNA levels in the same filters. Subcellular Fractionation-UMR 201 cells were grown in 500-cm2 trays in a-MEM containing 10% FBS until 80% confluent before treatment with M retinoic acid or M retinoic acid together with 10 pgfml cycloheximide in a-MEM containing 2% vitamin A-deficient FBS for the indicated incubation times. We used the method described by Leppard and Shenk (29) to isolate the subcellular fractions. All procedures were carried out on ice. Briefly, monolayers were washed thrice phosphate-buffered saline, scraped, and pelleted by centrifugation. Cells were resuspended in isotonic buffer (50 mM NaCl, 10 mM Tris, pH 7.5, 1.5 mM MgCI,, 175 pg/ml phenylmethylsulfonyl fluoride) and lysed by the addition of 10% Nonidet P-40. The nuclei were separated by centrifugation and the supernatant reserved as the cytoplasmic fraction. The nuclear membrane fraction was isolated by suspending the nuclei in isotonic buffer containing 10% Nonidet P-40 and 10% sodium deoxycholate. After centrifugation, the supernatant containing the nuclear membrane fraction was decanted (30). Pelleted nuclei were resuspended in RSB buffer (10 nm NaCl, 10 mM Tris, pH 7.5,3 mM M&I,, 175 pVml phenylmethylsulfonyl fluoride), digested with RNase-free DNase I, and centrifuged to separate the DNase I eluate in the supernatant.
Nuclei were then depleted of digested chromatin by resuspending in 5 M NaCl (31, 32). The chromatin-depleted nuclei were pelleted by centrifugation and the salt eluate was pooled with the Supernatant from the DNase I treatment to give the DNase Usalt eluate which is enriched in mature mRNA. The remaining pellet, defined as the nuclear matrix (33,34), and containing predominantly precursor mRNA (35), was solubilized in isotonic buffer by adding SDS to a final concentration of 0.2% (w/v) and EDTA to a final concentration of 10 mM.
Total RNA from the various fractions was prepared for Northern blot analysis using the method described above. Equal loading was confirmed by staining the gels with ethidium bromide and also by probing with an oligonucleotide specific for 18 S rRNA. SI Nuclease Protection Assay-SI nuclease protection assay was performed as described previously (36,37). A 27 nucleotide oligomer (5'-CAATATAGCTGCCACATGCCTGCTCAC-3') complementary to the start of the second intron of the rat liver alkaline phosphatase gene ( Fig.  3 4 Ref. 38) was labeled with [y3'P1ATP at the 5' end using T4 polynucleotide kinase. 50,000 countdmin of the labeled oligomer was hybridized to 20 pg of total RNA from each of the subcellular fractions obtained as described above, in 30 p1 of buffer containing 400 mM NaCl, 40 mM PIPES, 1 mM EDTA, 1% sodium dodecyl sulfate, pH 6.4, and covered with 50 pl of parafin oil. RNA was denatured by heating for 10 min at 85 "C before transferring to a 60 "C waterbath to incubate overnight. Digestion with SI nuclease (75 units in 300 pl of 300 mM NaC1, 30 mM sodium acetate, 3 mM zinc acetate, pH 4.74, containing 10 pg of salmon sperm DNA) was performed a t 30 "C for 90 min. After addition of 50 pl of termination buffer (4 M ammonium acetate, 100 mM EDTA), the mixture was extracted with phenol-chloroform, ethanol precipitated, electrophoresed on a 20% acrylamide gel containing 8 M urea, and the dried gel subjected to autoradiography.

Polymerase Chain Reaction Amplification
of Reverse nunscribed mRNA-First strand cDNA was synthesized from 2 pg of total RNA by incubating for 1 h a t 42 "C with 15 units of avian myeloblastosis virus reverse transcriptase (Promega, Madison, WI) following oligo(dT) priming. 2 pl of this reaction mixture was submitted to PCR to amplify the region of the rat alkaline phosphatase gene transcript across exons 4 and 5 to include intron 4 (see Fig.  3A). Primers were: a 5' primer (5'-AGAAAGAGAAAGACCCCAGTT-3') representing nucleotides 1-21 of exon 4 and a 3' primer (5'-Cl"GGAGAGAGCCACAAAGG-3') representing nucleotides 97-116 of exon 5 (38). Preliminary experiments indicated that for intronless (corresponding to spliced mRNA) PCR products, 20 cycles of amplification was appropriate to yield products during exponential amplification. However, in order to demonstrate product corresponding to intron-containing (unspliced) mRNA, we needed to use 40 cycles of amplification. Amplification using Taq DNA polymerase (Boehringer Mannheim) with an annealing temperature of 55 "C was employed in a Perkin-Elmer Cetus 480 thermal cycler. 20 pl of each PCR reaction mixture was run on a 2% agarose gel, transferred to nylon, and products authenticated by probing with a mixture of three digoxigenin-labeled oligonucleotides corresponding to sequences within exons 4 and 5 as well as in intron 4 (see Fig. 3A ). Oligonucleotides were labeled with digoxigenin-dUTP using a 3"tailing kit (Boehringer Mannheim). Hybridization was carried out with 2 pmollml labeled oligonucleotide in a buffer containing 5 x SSC, 0.02% SDS, 0.1% sarcosine, and 100 pg/ml poly(A), at 55 "C for 14 h. Detection was by chemiluminescence using Lumigen PPD (Boehringer Mannheim), according to the manufacturer's instructions. Products of the expected sizes were obtained, uiz., 636 bp for unspliced and 236 bp for spliced species.

RESULTS
Effect of Retinoic Acid on the Danscriptional Rate ofAlkaline Phosphatase Gene-Nuclear run-on assays were performed on nuclei isolated from UMR 201 cells after they had been treated for 2, 4, 6, 8, and 14 h with lo4 M RA. The results (Fig. 1, top   panel) show that the alkaline phosphatase gene was constitutively transcribed by UMR 201 cells though this was relatively weak compared with the transcriptional rate of actin. Retinoic acid treatment transiently increased the transcriptional rate of the alkaline phosphatase gene by approximately 2-fold compared with control, when normalized by reference to the actin ... signal (of appropriate exposure) on the same filter. On the other hand, cytoplasmic RNA from the same preparations showed a very marked retinoic acid-stimulated increase in the steadystate level of mRNA for alkaline phosphatase (Fig. 1, lower  panel). This was time dependent, evident by 4 h and peaked at 24 h (data not shown), consistent with earlier observations (7,8). Importantly, no cytoplasmic mRNAfor alkaline phosphatase was detected in control UMR 201 cells despite the constitutive transcription of this gene.
It seemed unlikely that the progressive and substantial increase in the steady-state level of alkaline phosphatase mRNA in response to retinoic acid could be explained by the mild and transient increase in transcriptional rate of the alkaline phosphatase gene. This discrepancy indicated the need for study of post-transcriptional events in the nucleus. Subcellular Localization of Precursor mRNA for Alkaline Phosphatase-In order to examine the effects of retinoic acid on nuclear processing of mRNA for alkaline phosphatase, UMR 201 cells were subfractionated after treatment with retinoic acid (1O"j M) for various times. The subcellular localization of exonic and intronic sequences was compared in control and retinoic acid-treated cells. Very sensitive methods were necessary to detect intronic material corresponding to precursor (prespliced) alkaline phosphatase mRNA. First, we used a S1 nuclease protection assay in which detection was by hybridization to a 27-mer oligonucleotide, complementary to a sequence in intron 2 of the alkaline phosphatase gene (see Fig. 3-41. The results show that in both control cells and cells treated with retinoic acid for 4 h, precursor mRNA for alkaline phosphatase was present almost exclusively in the nuclear matrix fraction (Fig. 2). A similar picture was obtained with cells treated with retinoic acid for 8 and 24 h (data not shown).
In order to verify these results and at the same time to seek evidence of elongation of the nascent transcript, we employed RT-PCR to amplify a sequence corresponding to intron 4 of the alkaline phosphatase gene, using primers complementary to exons 4 and 5 (see Fig. 3A). An amplified product corresponding to RNA containing intron 4 was present only in the nuclear matrix fractions of both control and retinoic acid-treated cells (Fig. 3B). DNase treatment during the nuclear fractionation procedure ensured that this product did not result from contaminating genomic DNA. Neither of the two above methods as used is strictly quantitative. However, together they provide strong confirmatory evidence that (i) the nuclear matrix fraction specifically contains primary alkaline phosphatase mRNA transcripts, and (ii) this material is present in both retinoic acid-treated and untreated cells. Subcellular Localization of Mature mRNA for Alkaline Phosphatase-To determine whether this precursor mRNA was capable of being processed into the mature, spliced product required for protein translation in the cytoplasm, UMR 201 cells were subfractionated after treatment with retinoic acid M) for 4 and 8 h. When these samples were subjected to RT-PCR analysis, using primers complementary to exons 4 and 5, a product corresponding to spliced mRNA was seen. Furthermore, amplification for an appropriate number of cycles (20 cycles, see "Experimental Procedures") confirmed a much greater abundance of alkaline phosphatase mRNA, relative to GAF'DH mRNA, in fractions from retinoic acid-treated cells, than from control cells (Fig. 3C). There was clearly more mature alkaline phosphatase mRNA after 8 h than after 4 h of retinoic acid treatment. In addition, spliced product in the nuclear matrix fraction was only evident in the cells treated for 8 h with RA. The data indicate that spliced mRNA may progress quickly to other nuclear compartments. Given sufticient amplification cycles, product corresponding to spliced mRNA could also be demonstrated in all fractions from control cells, albeit at levels very much less than in retinoic acid-treated cells (data not shown).
After cells were treated for 4, 8, 12, and 24 h, equal quantities of RNA from each cellular fraction were electrophoresed and transferred to filters. The filters were probed with a 2.4-kb full-length cDNA encoding rat alkaline phosphatase. No alkaline phosphatase mRNA was detected in any subcellular fraction of control UMR 201 cells (Fig. 4, top row). When treated with retinoic acid, mature alkaline phosphatase mRNA was detected in the subcellular fractions, and there was a distinct pattern in the order of its appearance. Mature alkaline phosphatase mRNA was consistently detected in the nuclear membrane fraction a few hours prior to its appearance in the DNase h a l t eluate and cytoplasmic fraction. The amount of cytoplas-Stabilized by Retinoic Acid mic alkaline phosphatase mRNA progressively increased, peaking a t 24 h. Nuclear matrix RNA hybridized only weakly with the alkaline phosphatase cDNA probe and only after treatment with retinoic acid for 8 h, which was consistent with the observations made with RT-PCR analysis (Fig. 3 0 . To determine whether the post-transcriptional effects of retinoic acid were dependent on de novo protein synthesis, the above experiment was performed with cells treated with retinoic acid together with 10 pg/ml of the protein synthesis inhibitor, cycloheximide. This concentration of cycloheximide was chosen because it reproducibly inhibited about 95% of protein synthesis in the UMR 201 cells (data not shown). Combined treatment resulted in a reduction in the amount of mature alkaline phosphatase mRNA detected in most samples, together with a slightly different distribution of this species at 4 and 8 h (Fig. 4). Cycloheximide treatment alone did not induce alkaline phosphatase mRNA in any of the fractions a t any time (data not shown). The net result of protein synthesis inhibition was a reduction in the amount of alkaline phosphatase mRNA reaching the cytoplasm.
Effect of TGFP on Banscription and Nuclear Processing of Alkaline Phosphatase mRNA-It was necessary to consider the possibility that enhanced nuclear processing in retinoic acidtreated cells may simply be a consequence of an increase in gene transcription. To test this, UMR 201 cells were treated with M retinoic acid or 1 ng/ml TGFP for 8 and 14 h. Both TGFP and retinoic acid treatment increased the transcriptional rate of the alkaline phosphatase gene by approximately 2-fold at 8 h (Fig. 5A). By 14 h, the retinoic acid-induced transcription of alkaline phosphatase had returned to the basal level but nuclei from TGFP-treated cells were still transcribing alkaline phosphatase at three times the control rate. Importantly, despite the increase in the rate of alkaline phosphatase gene transcription, no cytoplasmic mRNA for alkaline phosphatase was detected in TGFP-treated cells at either time point (Fig. 5B).
We therefore tested the effect of TGFP treatment on nuclear processing of alkaline phosphatase mRNA. RNA prepared from the various subcellular fractions was probed for the presence of precursor or mature alkaline phosphatase mRNA. Fig. 6A shows that precursor mRNA was present in the nuclear matrix fraction of control cells and of cells treated with TGFP for 8 h. A similar picture was obtained in cells treated with TGFP for 14 h (data not shown). In sharp contrast, a t 14 h, when nuclei from TGFP-treated cells were still actively transcribing the gene at a rate above that of control or retinoic acid-treated cells, mature mRNA was detected in the non-matrix and cytoplasmic fractions of retinoic acid-treated but not in TGFP-treated cells (Fig. 6B). DISCUSSION These studies identify a novel mechanism of retinoic acid regulation of gene expression, in which the immediate posttranscriptional degradation of nascent mRNA is impeded by retinoic acid treatment. The effect was demonstrated on the alkaline phosphatase gene whose expression is enhanced greatly by retinoic acid in preosteoblasts, despite only modest effects on gene transcription. The nuclear subfractionation method used is one devised by Leppard and Shenk (29) to define the site of action of the E1B gene product in facilitating the accumulation and movement of newly synthesized late viral mRNAs through various nuclear compartments. Their kinetic analysis was consistent with the notion that emigration of mRNA is linked spatially to processing. This is depicted diagrammatically in Fig. 7.
Nuclear processing of newly synthesized RNA includes 5' capping, splicing, and 3' polyadenylation. DNA replication,  (38). The locations of the oligonucleotides used for RT-PCR primers are indicated by the arrows; oligonucleotide probes to authenticate PCR products are indicated by solid bars and labeled 1 3 . Oligonucleotide 4 was used for S1 nuclease analysis. Intron 4 is 400 bp. Splicing out intron 4 results in a PCR product of 236 bp; unspliced mRNAyields a RT-PCR product of 636 bp. B , subcellular fractions were prepared from control cells or cells treated for 4 or 8 h with M retinoic acid (RA). Fractions are: cyto = cytoplasmic; memb = nuclear membrane; DNase I = DNase Vhigh salt eluate; and NM = nuclear matrix. 40 cycles of amplification, of reverse-transcribed mRNAfrom each fraction, gave rise to an amplified fragment of approximately 636 bp only in the nuclear matrix fraction. The fragment hybridized to specific digoxigenin-labeled oligonucleotides ( 1 3 ) complementary to exons 4 and 5 and intron 4. Chemiluminescence was detected by exposure of filters to x-ray films a t room temperature. C, the PCR reaction described inB was carried out for 20 cycles of amplification only, which produced a 236-bp product corresponding to exon 4 spliced to exon 5. D, the reversetranscribed mRNA used in A and B was amplified for 20 cycles using rat GAF'DH primers to produce the expected 414-bp fragment which hybridized to a specific digoxigenin-labeled internal oligonucleotide (see "Experimental Procedures"). There is increasing evidence that the nuclear matrix provides more than just a structural framework. Polymerases involved in DNA or RNA synthesis are bound to the matrix (48-501, suggesting an active role for the matrix in determining the location of polymerizing complexes. Nucleocytoplasmic transport involves a series of events between nuclear processing and translation of cytoplasmic mRNA, but comparatively little is known about the regulation of these post-transcriptional processes prior to the appearance of mature mRNA in the cytoplasm. Since only mature mRNA is transported out of the nuclei (511, splicing reactions, for those transcripts which contain introns, must be completed before translocation to the cytoplasm can occur. The process of translocation of poly(A)+ mRNA through the nuclear envelope is energy dependent and signal-mediated (24, 52-55). There is scant information about the role of hormones in nucleocytoplasmic transport. It has been shown, for instance, that insulin and epidermal growth factor differentially modulate the eftlux of mRNA from isolated rat liver nuclei at the level of translocation through the nuclear pores (56).
A study to examine the subnuclear location of RNA precursors of ovalbumin and ovomucoid mRNA found that high molecular weight precursor RNAs were exclusively associated with the nuclear matrix (35). Similarly, our results show that precursor mRNA for alkaline phosphatase is detectable only in the nuclear matrix fraction of control, retinoic acid, and TGFPtreated cells. Two very sensitive methods were used to identify precursor mRNA, targeting two different intronic sequences of the alkaline phosphatase gene separated by over 14 kb. This result is also consistent with recent results obtained by in situ hybridization which show that intron-containing RNAis highly restricted in the nucleus, proximal to the gene of transcription (57). On the other hand, when the subcellular fractions were probed with the full-length cDNA for the presence of mature, spliced alkaline phosphatase mRNA, we found this to be demonstrable predominantly in the non-matrix and cytoplasmic fractions of retinoic acid-treated cells. Again, these results were confirmed using a RT-PCR strategy. These data and our evidence for transcript elongation suggest that newly transcribed alkaline phosphatase mRNA in control UMR 201 cells is probably degraded at an early stage following completion of transcription, a fate apparently analogous to the product of the constitutively expressed alkaline phosphatase gene in HepG2 cells (6). This degradation step is likely to take place while the nascent mRNA is attached to the nuclear matrix. On the other hand, retinoic acid treatment of UMR 201 cells resulted in a time-dependent increase in the amount of mature alkaline phosphatase mRNA present in the DNase h a l t eluate, nuclear membrane, and cytoplasm. This is consistent with a role for retinoic acid in the post-transcriptional regulation of alkaline phosphatase gene expression, acting perhaps to stabilize the nascent alkaline phosphatase mRNA, thereby facilitating its processing into mature mRNA and translocation into the cytoplasm. Consistent with this is the detection of spliced mRNA in nuclear matrix fraction only when cells had been treated for 8 ] 12h ] 24h FM:. 4. Subcellular localization of mature mRNA for alkaline phosphatase in UMR 201 cells using a cDNA (exonic) probe. Crlls w r r r -own in 500-rm' t r a y s r i t h r r untreated (control), trratrd with 10"' M retinoic acid tlin ) or 10"' >I retinoic acid and 10 pdml cycloheximide ( R A + CIfX) for 4, 8. 12. and 24 h. The isolated suhcellular fractions, obtained as described under "Experimental Procedures," were cytoplasm, nuclear mrmhrane, DNasr h a l t e l u a t r , a n d n u c l r a r m a t r i x . Total RNA was prepared from each fraction and 20 pg loaded into each lane. The filters were probed with a plasmid containing a 2.4 kh, fulllength cDNA for rat alkalinr phosphatase. This experiment was prrformrd three times, and a representative result is shown. h with retinoic acid. It is possible also that retinoic acid could increase the efficiency of nuclear processing of alkaline phosphatase mRNA by facilitating its movement through the various nuclear fractions. These different mechanisms of actions are not necessarily mutually exclusive.
An action o f retinoic acid in post-transcriptional regulation of alkaline phosphatase gene expression would explain why cytoplasmic alkaline phosphatase was detectable in retinoic acid-treated cells before the increase in alkaline phosphatase transcriptional rate was evident. In particular, it would explain the steady accumulation of cytoplasmic alkaline phosphatase mRNA long after the small and transient rise in alkaline phosphatase gene transcription has returned to a basal level. In F9 teratocarcinoma cells, it has been shown that retinoic acid does not increase the stability of cytoplasmic alkaline phosphatase mRNA (21 ).
The possibility that enhanced nuclear processing of alkaline phosphatase mRNA by retinoic acid may simply be a consequence of an increase in the rate of transcription of the alkaline phosphatase gene was considered. However, an increased alkaline phosphatase transcriptional rate as a result ofTGFP treatment did not result in increased accumulation of mature alkaline phosphatase mRNA levels in the nuclear fractions or cytoplasm even though precursor mRNA was readily detected in the nuclear matrix fraction. This illustrates that while an increased transcriptional rate may be a major determinant of enhanced gene expression, it does not automatically result in enhanced nuclear processing and elevated mRNA levels in the cytoplasm. It supports the concept that regulation of post-tran- T h r isolated suhcvllular fractions. ohtaincd ns rIr*scrlhrd undrr "Exprrimental I'rocrdurrs," werr cytoplasm. nuclrar mrmhr:rnr, 1)Snsr I w l t eluate, and nuclear matrix. 20-pg aliquots of total HSA w w r usrd for S 1 nuclease analysis to drtrct r;cqurncca corrrsponding to intron 2 . as drscrihrd in the lrgrnd to Fig. 2  scriptional processes can he indrprndent of t h r control of grnr transcription. There is also the possibility that TGFB may h a w a n additional action that rrsults in t h r rapid drgradation of nascent alkaline phosphatase RNA.
When the cells were treated with rrtinoic acid and cyclohrximide together, there was a significant rrduction in thc amount of alkaline phosphatase mRNA locatrd in most hut not a11 of t h r various subcellular fractions. This may he rvidrncr for A rolr of CIP n o w protein synthesis in thr regulation of nuclrar procrssing and nucleocytoplasmic transport hv rrtinoic acid. I'ossihlr protein candidates for action would include thr h~trropcnrous nuclear (hn) as well a s small nuclrar l s n ) prntrins which arc Newly transcribed, precursor mRNA is closely associated with nuclear matrix where processing reactions such as capping, splicing, and addition of polyfA)' tail take place. Processed mRNA is released into the DNase Ysalt eluate before translocation across the nuclear membrane into the cytopfasm. Regulation of gene expression can occur at the level of gene transcription, nuclear stabilization, and/or processing or cytoplasmic stability of mRNA. rapidly bound to newly synthesized RNA to form various ribonucleoprotein complexes (hnRNPs and snRNPs) (58). The functions of these nuclear proteins are not clearly understood but there is good evidence that snRNPs play an important role in mRNA splicing (59). In the influenza virus, export of newly assembled viral ribonucleoproteins depends on a viral membrane protein, M1, which is imported from the cytoplasm and is required to be associated with RNPs in the host nuclei before they can be exported to the cytoplasm (60). A protein, regulated by retinoic acid and performing a similar role, may be important in eukaryotic nucleocytoplasmic transport, acting to facilitate the movement of mRNA through the various nuclear fractions as it is being processed, and at the same time, protecting it from degradation.
In conclusion, we have identified a novel action of retinoic acid in regulating gene expression of alkaline phosphatase at the post-transcriptional as well as the transcriptional level. Early post-transcriptional degradation of nascent alkaline phosphatase mRNA may be an important control point in the regulation of alkaline phosphatase gene expression and one which is overcome by retinoic acid but not by an alternative enhancer of transcription such as TGFP. Our results support the notion that gene transcription is not the sole, or indeed necessarily the dominant, factor responsible for establishing the level of nuclear mRNA. Changes in stability and/or efficiency of processing can also be regulated to alter the relative abundance of specific mRNAs between the nucleus and cytoplasm.