The Isolation and Characterization of HeLa Cell Messenger RNA-like Molecules Containing Uridylic Acid-rich Oligonucleotide Sequences*

HeLa cell mRNA, isolated from LURNPS released from polysomes by EDTA to minimize contamination from nuclear RNA, has been separated into four classes of molecules: 1) those containing poly(A) sequences Ipoly(A+)] as well as oligo(U) sequences [oligo(U+)] (4 to 13%); 2) poly(A+) oligo(U-) (-62%); 3) poly(A-) oligo(U+) (-2%); 4) poly(A-) oligo(U-) (-33%). The oligocu) segments are 89% UMP and range from 20 to 50 nucleotides in length. The poly(A+) oligo(U’) mRNAs appear to contain the oligo(U) in a region with secondary structure (possibly an intramolecular duplex with the 3’-poly(A)) since they did not bind to poly(A) Sepharose without prior HCHO modification. HCHO modi- fies the exocyclic amino groups of CMP, GMP, and AMP and prevents hydrogen bonding but reacts only slightly with UMP. After removal of the excess HCHO, the oUgo(LJ) of the mRNA was free to bind to the poly(A) Sepharose. Control experiments indicated that the binding of oligo(U’) mRNA to poly(A) Sepharose was specific and the HCHO did not cause cross-linking of the RNAs. “he poly(A+) oligo(U+) RNAs appeared to contain one oligo(U)/molecule and could re-bind to poly(A) Sepharose with

Previous experiments have described the isolation of uridylic acid-rich oligonucleotide sequences (80% UMP and 20 to 50 nucleotides (ntds)' long) from HeLa cell heterogeneous nuclear RNA (hnRNA) by the use of TI RNase digestion (specific for GMP residues) and poly(A) Sepharose (1). These studies also reported that the molar concentration of oligo(U) increased with the sue of the hnRNA. hnRNA -30 kilobases (kb) long contained, on the average, one oligo(U) while only 1 in 12 hnRNAs -5 kb long contained one oligo(U) and only 1 in a 100 cytoplasmic messenger RNAs (mRNAs) -2 kb long contained one oligo(U). This suggested that if the oligo(U) regions were present in mRNA precursors, they were destroyed in the nucleus as the mRNA sequences were processed and transported to the cytoplasm (1). However, in these experiments, when mRNA was assayed for oligo(U) care was not taken to disrupt any hybrids that might have formed between the 3'-poly (A) and oligo(U) and rendered the oligo(U) inaccessible to poly(A) Sepharose. Subsequent experiments which took this precaution (by heating the TI RNase digest to 65°C) indicated that poly(AC) hnRNA 20 kb long contained -2 oligo(U) regions located within -8 kb of the 5'-terminus but mRNA was not reassayed (2). Recent experiments by Edmonds and co-workers have reported that -20% of HeLa mRNA can be shown to contain oligo(U) sequences but only when caution is taken to add excess synthetic polyW) to the TI RNase digested mRNA to prevent hybrids from forming between the oligo(U) and the 3"poly(A) (3).
This paper reports that 10 to 20% of the poly(A+) mRNAs can be retained specifically on columns of poly(A) Sepharose but only when the mRNAs were treated with HCHO prior to chromatography. The retained mRNAs were shown to contain oligo(U) segments (89% UMP, 20 to 50 ntds long) by a procedure which employed TI RNase digestion followed by HCHO treatment and poly(A) Sepharose. HCHO reacts primarily with the exocyclic amino groups of AMP, GMP, and CMP (forming methylol derivatives which are stable up to 80°C) and prevents hydrogen bonding but reacts only slightly with UMP (4, 5 ) . Consequently, as these experiments with mRNA demonstrated, HCHO modified the 3'-poly(A) and prevented it from hydrogen bonding with the oligofu). After removal of excess HCHO, the oligo(U) region of the mRNA was free to bind the poly(A) Sepharose. This permitted the isolation of intact oligo(U)-containing RNAs as well as oligo(U) sequences produced by TI RNase digestion. The isolation of intact oligo(U') mRNA and hnRNA, which was not accomplished in previous studies (3, 6, 7), may permit further study of the structure and translational properties of the poly(A+) oligo(U+) mRNAs and their relationship to the oligo(U+) hnRNA.

MATERIALS AND METHODS
The miniprint supplement contains all experimental details.'

Isolation of mRNAs that Contain Oligo(U)
Sequences-It was necessary to employ a procedure that would yield mRNA free of contamination by hnRNA since hnRNA contains -75% of the cellular oligo(U) (1, 3). Therefore, polysomes were isolated (Fig. lA) and adjusted to 10 m~ EDTA to release the

HeLa Cell mRNA-like Molecules Containing
Oligo(U) Sequences messenger ribonucleoprotein particles (mRNPs) and the mRNPs between 30 S and 70 S were isolated on subsequent sucrose gradients (Fig. IB). Since the sedimentation value of hnRNPs i s not affected by EDTA, any hnRNPs which may have co-sedimented with the polysomes would resediment as RNPs 2100 S and should not contaminate the mRNP (8-10). Fig. 1B shows there were very few RNPs >70 S in size. Only RNA purified from the 30 S to 70 S mRNPs will be referred to as mRNA. Also, Korwek et al. (3) showed that -80% of the oligo(U) tracts in HeLa polysomal RNA were released from polysomes by EDTA. Table 1 shows it was necessary to react poly(A') mRNA with HCHO to get maximal binding to poly(A) Sepharose at 4°C. Seventeen to eighteen percent of the HCHO-treated mRNA bound to the Sepharose and was eluted with formamide elution buffer (FEB) (Samples 2, 6, and 11), whereas, only 2 to 4% of the untreated mRNA was bound (Samples 1, 5 , and IO). HCHO treatment of synthetic poly(U) did not affect its binding to poly(A) Sepharose. The poly(A-) mRNAs (molecules that did not bind to poly(U) Sepharose) were not as dependent on HCHO treatment since 5 to 8% of these molecules bound to poly(A) Sepharose with or without prior HCHO treatment. In some experiments, as much as 1670 of the (A-) mRNAs bound (Sample 7). Subsequent isolations of (A-) (U') mRNAs did not employ HCHO. HCHO treatment of the (A-) mRNA resulted in the formation of an insoluble residue which is thought to be due to the high concentrations of unlabeled rRNA present (see "Materials and Methods"). Binding to poIy(A) Sepharose is not solely a characteristic of poly(A') RNA that had entered polysomes since 22 to 24% of the poly(A') RNA isolated from cytoplasmic particles 280 S in size was retained (Samples 8 and 9).
Specificity of Binding to Poly(A) Sepharose-The observation that the binding of poly(A') mRNA to poly(A) Sepharose was enhanced 5-to 8-fold after HCHO treatment suggested that these mRNAs contained intramolecular oligo(U): poly(A) duplex regions and that the oligo(U) sequences were only free to bind to the sepharose after HCHO had modified the 3"terminal poly (A). To test this possibility, it was necessary to establish that (a) RNAs known to contain oligo(U) sequences hydrogen-bonded to poly(A) sequences did not bind to poly(A) Sepharose without prior HCHO treatment, and ( b ) HCHO treatment did not cause oligo(U-) RNAs to bind to poly(A) Sepharose.
Poliovirus replicative form (RF) RNA was employed because it is considered to be a completely double-stranded, intermolecular RNA duplex containing the 35 S positive strand RNA with a 3"terminal poly(A) sequence and the negative strand RNA with a 5"terminal poly(U) sequence (11)(12)(13). If the 5'-poly(U) is devoid of extensive unpaired regions, the RF should not bind to poly(A) Sepharose without HCHO treatment but -50% of the RNA (the negative strand) should bind after HCHO treatment, provided the two RNA strands contain equal amounts of label. Table I1 shows that only 2% of the RF bound without HCHO treatment but that 41% bound after HCHO treatment (Samples 1 and 2). T h i s result also strongly suggests that HCHO did not cause extensive cross-linking of the two RNA strands because this would have caused much more than 41% of the RNA to bind. When the bound RNA was analyzed on acrylamide-formamide gels (14), it migrated predominantly as a discrete RNA 35 S in size as expected of the negative strand RNA (data not shown). Also, Kwan et al. (15) reported that HCHO treatment of hemoglobin mRNA caused no detectable cross-linking of the molecules. The present results were substantiated through binding studies with poliovirus replicate intermediate (RI), and RNA complex consisting of the 35 S negative RNA strand and several associated positive RNA strands, some of which contain poly(A) (12,16). Without HCHO treatment, only 2% of the RI bound but after ECHO 27% of the RI was retained.
The data in Table I1 also indicate that RNAs lacking poly(U) sequences did not bind to poly(A) Sepharose, even after HCHO treatment. Without HCHO treatment only 1% of 45 S pre-rRNA, poly (A), and poly(C) was bound and only 3% of mitochondrial poly(A') mRNA was retained; HCHO treatment of these RNAs did not increase the binding (Samples 5 to 10). The binding of cytoplasmic poliovirus 35 S positive strand RNA and poly(G) was 1% without HCHO treatment and this increased to only 4% after HCHO (Samples 11 to 13). Characterization of the Oligo(U) Tracts in the mRNAs which Bound to Poly(A) Sepharose-Poly(A+) mRNAs were treated with HCHO and chromatographed on poly(A) Sepharose possibly to separate mRNAs with short oligo(U) tracts (-30 ntds), which may represent adjacent phenylalanine (Phe) codons, from mRNAs with longer oligo(U) tracts which may have a regulatory function. The mRNAs were applied to Sepharose at 4°C and the oligo(U-) mRNAs were removed by washing with 0.4 M NEPSark and EPSark buffer, The column was then equilibrated with NEPSark buffer (0.1 M NaCl), adjusted to 25"C, and washed with NEPSark. This removed one-half to two-thirds of the bound mRNAs, which were referred to as the "0. and contained 90% UMP and -3% purines (Fig. 2B). These results are consistent with the finding that synthetic poly(U) fragments which were -33 ntds long could not be removed from poly(A) Sepharose with NEPSark buffer but were eluted with FEB (data not shown). Fig. 2B also shows the migration of the VSV "leader" RNA (which is 48 ntds long) and illustrates the resolution on 15% acrylamide, 8 M urea geh and the heterogeneity of the oligo(U) tracts. Table IV  In these experiments, -65% of the total mRNA was found to be poly(A'). Based on the per cent of mRNAs which were oligo(U') (Tables I and III), these results have separated HeLa mRNAs into four classes of molecules: 1) (A') (U') (4 to 13%); 2) (A') (U-) (-52%); 3) (A-) (U') (-2%); and 4) (A-) (U-) (-33%). Consistent with these results, total mRNA contained 0.1 to 0.2% of its ntds in oligo(U) tracts which were isolated a t 25°C (Table IV, Samples 8 and 9). The conditions used to isolate the poly(A') mRNAs minimized the possibility that the oligo(U) segments were really present in poly(A-) mRNAs that had formed intermolecular duplexes with the poly(A) of other mRNAs. The mRNAs were heated to 65°C in low salt buffer prior to binding to poly(U) Sepharose. Any intermolecular duplexes which resisted this treatment should have been disrupted by the formamide gradient and eluted from the column prior to elution of poly(A') mRNA.
Size of the Four Classes of mRNA and the Ability of OLigo(V) RNAs to Rebind to Poly(A) Sepharose- Fig. 3 shows the size of the four mRNA classes as determined on HCHO-sucrose gradients. The oligo(U') mRNAs were selected to contain oligo(U) tracts >20 ntds (see Table 111). The (A') (U') mRNAs were heterogeneous in length and were enriched with larger mRNAs so that the average size was -3 kb (Fig. 3B). The other three classes of mRNA were also heterogeneous but with an average size of -2 kb.
The oligo(U') RNAs containing oligo(U) tracts >20 ntds were tested for their ability to rebind to poly(A) Sepharose in 0.1 M NaCl buffer at 25°C. The FEB oligo(U') mRNAs rebound with 31 to 46% efficiency (Table V, Samples 1 and 2). The oligo(U') RNAs from subpolysomal particles rebound with 55 to 75% efficiency (Sample 3). Therefore, the (U') mRNAs can be bound to poly(A) Sepharose several times to enhance purification. However, it was necessary to re-treat the RNAs with HCHO to obtain maximal rebinding. Only 2 to 8% of the 0.1 M (A') (U') mRNA rebound to poly(A) Sepharose (data not shown) which, undoubtedly, was due to their short oligo(U) tracts.
Correlation between the Ability of a n RNA to Bind to Poly(A) Sepharose and the Modification of its J-poly(A) by HCHO- Table VI indicates that an RNA molecule containing both a poly (A) and an oligo(U) sequence w i l l not bind to poly(A) Sepharose until HCHO modification has rendered the 3'-poly(A) incapable of hydrogen bonding. Poly(A') cytoplasmic RNA was f i t treated with HCHO for 10 min at either 37°C or 65°C and then tested for its ability to bind to poly(U) and poly(A) Sepharose. Poly(A') RNA treated with HCHO at 37°C re-bound to poly(U) Sepharose almost completely (Sa%, Sample 1) while only 2% bound to poly(A) Sepharose (Sample 6). The same RNA treated at 65°C completely lost it ability to rebind to poly(U) Sepharose (Sample 2) but 19% of the RNA was now bound to poly(A) Sepharose (Sample 7).
Table VI also shows that the retention or loss of the ability of synthetic poly (A) to participate in hydrogen bonding after treatment with HCHO for 10 min at 37°C or 65°C is similar to that observed with cytoplasmic poly(A') RNA (Samples 3 and 5). Exposure to HCHO for 30 min at 37°C reduced the binding of poly(A) to poly(U) Sepharose by 67%, suggesting that it may be possible to isolate the (A') (U') mRNAs after less extensive HCHO modification (Sample 4). Treatment of the poly(A') RNA with (CH&SO, a reversible denaturant, only resulted in the binding of 2% of the RNA to poly(A) Sepharose (Sample 8).
The amount of mRNA which was nonspeciflcally bound to poly (A) Sepharose that was treated with ethanolamine and stored as a powder (see "Materials and Methods") was usually -1% and less (Tables 111 and V). The higher amounts of  nonspecifically bound mRNA in Tables I, 11, and VI resulted from using poly(A) Sepharose that was not treated with ethanolamine and stored at 4°C for longer than 3 weeks.
Finally, while the oligo(U+) polysomal RNAs were released from polysomes by EDTA, as expected of mRNAs, we are investigating their translational capacity (17), metabolic stability (la), and buoyant density of their mRNPs (8,9) to verify that they are mRNAs.

DISCUSSION
HeLa cell mRNA has been separated into four classes of molecules by the use of poly(U) Sepharose, HCHO modification, and poly(A) Sepharose: 1) poly(A') oligo(U') (4 to 13%); 2) poly(A') oligo(U-) (-52%); 3) poly(A-) oligo(U') (-2%); 4) poly(A-) oligo(U-) (-33%) (Table I and 111). These percentages may not be completely defiitive and could be affected by experimental fragmentation of the mRNA. However, significant amounts of fragments of mRNA containing either poly (A) or oligo(U) are not apparent in these experiments (Fig. 3). Since the (A') (U') mRNAs did not bind to poly(A) Sepharose without prior HCHO modification, they appear to contain the oligo(U) in a region with secondary structure and possibly in an intramolecular duplex with the 3'-poly (A).
Analysis of the mRNAs on denaturing HCHO-sucrose gradients revealed that the (AC) (U') mRNAs were heterogeneous in length and enriched with larger mRNAs so that the average size was -3 kb (Fig. 3B). The other three classes of mRNA were also heterogeneous with an average size of -2 kb (Fig. 3). These results are in agreement with previous analyses of HeLa poly(A') and poly(A-) mRNA (10,19) and are consistent with the fact that HeLa cells contain -3 X lo4 different mRNAs (20). Our finding that -65% of total mRNA was poly(A') mRNA was slightly less than the amount (-70%) reported by Milcarek et al. (10). This was probably because we selected for mRNAs with poly(A) tracts -100 to 150 ntds in length while Milcarek et al. (lo), using oligo(dT)cellulose, would have isolated all mRNAs with poly(A) tracts 220 ntds. It is possible that some of our poly(A-) oligo(U+) mRNAs contained short poly(A) tracts (530 ntds) that were hydrogenbonded to the oligo(U) and, therefore, could not bind to poly(U) Sepharose.
The (A') (U') mRNAs containing short oligo(U) tracts ( t 2 0 ntds and 90% UMP) have been separated from mRNAs whose oligo(U) tracts are significantly longer (20 to 50 ntds with an average size of 27 ntds and a composition of CzU24G, Fig. 2). The short oligo(U) tracts may represent adjacent Phe codons. However, the longer oligo(U) sequences would appear to be too long to function as Phe codons, although this possibility cannot be excluded. Depending on the location of the CMP residues, the minimum number of consecutive Phe codons would be two, with each two being adjacent to a serine codon and two additional sets of two Phe codons (i.e. UJJCUUJJCUUJJUG). The maximum would be eight consecutive phenylalanine codons (CCUz4G); this number would be even greater for the longer oligo(U)s. This many consecutive Phe residues is rarely observed in proteins (21). Future experiments in which the oligo(U+) mRNA wiU be isolated to preserve its template activity for in vitro translation (which is modified by HCHO (22)) may clarify the function of oligo(U).
It is possible that the oligo(U) sequences may affect the rate of translation or stability of the oligo(U') mRNAs. Oligo(U) has been localized in the 20 kb long poly(A') hnRNA and is situated within -8 kb of the 5'-end (2). Some of the 20 kb poly(A+) oligo(U+) hnRNA may give rise to the poly(A+) oligo(U+) mRNAs by a pathway which "splices out" intervening sequences (23) and preserves the 3'-poly (A) and one of the oligo(U)s. These mRNAs may contain the oligo(U) close to the 5'-terminus and be arranged in a circular configuration possible stabilized by protein(@. Previous experiments with mouse sarcoma (24) and HeLa cells (25) reported that after polysomes were digested with RNase A (specific for pyrimidines), UMP-rich fragments of mRNA were protected via their hydrogen bonding to the 3'-poly (A). Although the fragments were not established to be oligo(U) sequences, it was suggested that a poly(A):oligo(U) duplex was maintained, possibly through the action of poly(A)-binding protein(@ (25). It has now been shown that oligo(U) and poly (A) are paired together in HeLa HnRNPs in uztro and that the poly(A)binding protein appears to stabiliie this duplex (26). Recent electron micrographs have shown that the genome RNAs of Uukuniemi (27) and Sindbis (28) virus and 1 to 2% of HeLa cell cytoplasmic RNA (29) are in a circular configuration. However, for Sindbis RNA the circular form is not thought to be maintained by a poly(A):oligo(U) duplex (28).
The rate of ribosome binding to the poly(A+) oligo(U+) mRNAs during initiation may depend upon whether these mRNAs are linear or maintained in a circle via the poly(A)binding protein(s). Kozak (30) has shown recently that covalently closed circular, synthetic mRNAs containing an AUG initiator will bind to Escherichia coli ribosomes but not to wheat germ or reticulocyte ribosomes. Linearization of these mRNAs restored their ability to bind to the eukaryotic ribosomes, suggesting that during initiation eukaryotic ribosomes must recognize a free 5'-terminus (capped or uncapped) of an mRNA. Ilan and Ilan (31) have described a protein associated with reticulocyte initiator factor 3 (eIF-3) that denatures synthetic poly(rA:rU) and globin mRNA. When AUG(U), was used as an mRNA, polyphenylalanine synthesis could be inhibited completely when the polymer (A&), was hybridized to the AUG(U),. The addition of eIF-3 restored 80% of translation (31). Such a protein could also disrupt the poly(A): oligo(U) duplexes in HeLa cell mRNA and facilitate ribosome binding.
It should be noted that the present experiments were not designed to detect any small UMP-rich RNAs which may have been intermolecularly hydrogen-bonded to mRNA. Such an RNA has been reported to be associated with myosin mRNA in a manner to possibly affect translation (32).