A refinement of hRFC upstream gene structure
hRFC is regulated by multiple promoters, involving both ubiquitously expressed and tissue-specific transcription factors and cis
]. We originally identified 7 non-coding regions (designated A1, A2, A, B, C, D, and E) for hRFC by 5’RACE (5’ Rapid Amplification of cDNA Ends) assay in RNAs from normal human tissues [2
]. As many as 18 potential 5’UTRs were identified that could have arisen by variable splicing of the alternate non-coding exons. We later established that the A1 and A2 sequences are derived from a single ~1021 bp A1/A2 non-coding region that expands the first coding exon to ~1258 bp [4
]. Moreover, the original transcript forms AVI, CII, DIV, DV and E, if genuine, were certainly rare, having been identified in only one or two out of over 500 5’RACE clones [2
]. Accordingly, we have revised our model of hRFC gene structure to include the 5 major
non-coding exons that are spliced to generate as many as 14 different 5’UTRs fused to a common splice acceptor site at position −49, and to a common 1776 bp hRFC ORF (). The upstream non-coding sequences for hRFC were deposited in GenBank®1
In vitro translation analyses
Based on the scanning model for translation, efficient translation typically involves 5’UTRs that are moderately short (<210 bp), have a low percentage of guanines and cytosines (GC content <65%), are unstructured, and do not contain uAUGs [19
]. We used the Mfold algorithm (http://www.bioinfo.rpi.edu/applications/mfold
) to model RNA sequences and identify complex secondary structures (e.g., stem-loop structures) for the hRFC 5’UTRs [23
]. Using the predicted Gibbs free energy values (ΔG) calculated by Mfold, overall stabilities were determined by dividing the ΔG values by the individual 5’UTR lengths since -ΔG increases with 5’UTR length proportional to the extent of secondary structure along the length. Many of the 14 major hRFC 5’UTR sequences have features likely to adversely impact translation, most notably extensive secondary structure (as reflected in the high GC contents and free energy values), and the presence or absence of uAUGs ().
Characteristics of the hRFC 5’ UTRs.
To assess effects of the 5’UTRs on hRFC translation efficiency, we prepared constructs in which each of the 14 major 5’UTRs was fused (at position -49) to the hRFC ORF with a carboxyl HA epitope (). Equal amounts of the 5’UTR-hRFC constructs were added to in vitro
translation reactions. On Western blots (), levels of translation products for most of the 5’UTR hRFC constructs were low to undetectable. Only 3 constructs (AV, DII, and DIII), including some with the lowest percent GC and ΔG values and without uAUGs resulting in ORFs (see below), were efficiently translated in vitro
. Interestingly, BI, the major transcript form detected in a wide variety of cultured cell lines and primary tissues [2
], gave no detectable translation product in this assay at standard detection sensitivities (however, BI could be detected when the sensitivity of detection was increased; see below). This likely reflects the high GC content and high negative ΔG of the BI 5’UTR ().
In vitro transcription and translation of 5’ UTR hRFC expression constructs
For several of the A1/A2- and A-hRFC transcript forms, uAUGs occur in-frame with the hRFC ORF (without upstream termination codons). For the A1/A2 5’UTRs, translation can initiate at the uAUG 192 bp upstream of the normal translation start site [4
]. By in vitro
translation, the 72 kDa hRFC translated from the A1/A2I and III forms could be detected when the detection sensitivity was increased ( shows results for the A1/A2III form). Similarly, the AI, AIV, and AV transcripts (but not AII and AIII) included an in-frame uAUG 66 bp upstream from the normal translational start (). Initiation at this uAUG would result in a hRFC protein with a 22 amino acid leader and a molecular mass of 67,219 Da, or 2,222 Da larger than wild type hRFC. By extended electrophoresis, it was possible to detect highly reproducible differences (~2 kDa) between migrations of the in vitro
translation products for forms AI and AV from those derived from AII and BI (). When the Kozak consensus sequence for the downstream initiation site (GCCA/G
, where the purine at −3 and the G at +4 have the strongest effects) [24
] was mutated (to C
G) to further reduce translation initiation, the construct (designated km
AV-hRFC) encoded the ~67 kDa product, identical to that for the non-mutated AV construct ().
Effects of 5’UTRs on steady state hRFC mRNA and protein
We transfected the hRFC-null HeLa (R5) subline with the 14 5’UTR-hRFC constructs used for the in vitro translations. Steady state hRFC transcripts were measured on Northern blots and hRFC proteins were assayed on Western blots. In contrast with the in vitro translations, moderate to high levels of hRFC transcripts and proteins were detected for most of the 5’UTR-hRFC constructs, and the highest hRFC levels were measured for several of the constructs that translated most poorly in vitro (). Levels of steady state hRFC proteins generally paralleled levels of hRFC transcripts (). The slightly disproportionate levels of hRFC proteins for the AV- and DIII-hRFC transfectants are likely due to the efficient translation for these forms (). The transport activities for the hRFCs expressed in R5 HeLa cells were assayed with 3H-MTX. MTX uptakes paralleled levels of expressed hRFC protein (not shown).
Effects of the hRFC 5’ UTRs on the steady state hRFC transcripts and proteins in transfected HeLa cells
Effects of 5’UTRs on hRFC transcript stabilities
(e.g., AU) elements located in the 3’UTR of mRNAs are typically associated with transcript stabilities, 5’UTRs can also influence mRNA turnover [22
]. All 5’UTR-hRFC constructs had an identical 3’ non-coding sequence. To assess the possibility that differences in steady state levels of hRFC transcripts were a function of effects of their 5’UTRs on transcript stabilities, we transfected R5 cells with 5’UTR-hRFC constructs after which the cells were treated with Actinomycin D and total RNAs were isolated at intervals over 8 h. RNAs were analyzed on Northern blots with 32
P-labeled hRFC cDNA. hRFC transcript levels were normalized to RNA loading (assessed with ethidium bromide-stained 28S RNA) for calculation of half lives of first order transcript turnover. Results for the 14 5’UTR-hRFC constructs are shown in and the half lives are shown in parentheses. Interestingly, transcript half lives reproducibly fell into 3 distinct groups corresponding to fast (<3 h; A1/A2I, A1/A2II, AIV, DI), intermediate (3 to <6 h; A1/A2III, AI, AIII, AV, BII, C, DII, DIII), and slow (≥6 h; AII, BI) rates. Thus, turnover rates for the various hRFC mRNAs appear to be significantly determined by their 5’UTRs and are in general inversely proportional to steady state levels of hRFC transcripts in our R5 transfection model.
Effects of hRFC 5’ UTRs on the hRFC transcript stabilities
Further characterization of an alternate hRFC isoform translated from a uAUG within non-coding exon A
By in vitro translation (), a modified ~67 kDa hRFC polypeptide was identified that could be distinguished from the 65 kDa deglycosylated wild type hRFC initiated from the AUG at position +1. To assess the functional significance of the alternate ~67 kDa hRFC-A protein, hRFC-null K562 (K500E) cells were stably transfected with the AV- and kmAV-hRFC constructs. A parallel transfection was performed with the 11-hRFC wild type construct that includes 11 bp of upstream sequence and only the downstream AUG codon at position +1 for translation initiation. Transfectants were screened for similar levels of hRFC transcripts on Northerns (not shown) and for hRFC proteins on Westerns (). On Westerns, we identified clones expressing slight differences in the levels of hRFC proteins (~79% and 72% of wild type hRFC protein for the hRFC-A proteins from AV-hRFC and kmAV-hRFC, respectively). The glycosylated hRFC-A proteins migrated with masses slightly larger than for wild type hRFC (~85 kDa), a difference (~2kDa) that was more apparent following deglycosylation (, lower panel). Notably, there were no significant differences between the wild type and hRFC-A proteins in terms of subcellular localizations, as prominent signals at the cell surface were detected (although there was also some staining of intracellular structures; ). When the stabilities of the A-hRFC and wild type hRFC protein isoforms were compared on Westerns over 24 h following treatment with cycloheximide, there were no significant differences in rates of turnover ().
Expression of a hRFC-A protein isoform from an uAUG within non-coding exon
Initial rates of 3H-MTX uptake were assayed over a range of concentrations for the 11-hRFC and kmAV-hRFC transfectants to compare kinetic constants for the wild type hRFC and the hRFC-A protein isoform initiated from the uAUG in exon A. As summarized in , there were no significant differences in the absolute Vmax or Kt values for MTX uptake. However, when normalized for the slight difference in expression on Westerns (see above), there is a modest increase (~35%) in Vmax for the hRFC-A protein. By Dixon analyses, Kis were calculated for a number of folate and antifolate substrates including raltitrexed, pemetrexed, GW1843U89, leucovorin, and folic acid. There were no significant differences between the 11-hRFC and kmAV-hRFC transfected cells in the calculated Kis ().
Kinetic constants for 11-hRFC and kmAV-hRFC clones.
hRFC transcript levels are subject to tissue-specific controls at the level of transcription and involve use of alternate promoters [14
], reflecting levels and posttranslational modifications of transcription factors and/or effects of chromatin remodeling. We now show that posttranscriptional mechanisms can also be important, via use of alternate, variably spliced non-coding regions that result in transcript forms with differences in stabilities and propensities for 5’ cap-dependent translation, and/or that encode modified hRFC proteins. Although the effects of 5’UTR usage on steady state hRFC transcripts and proteins for several 5’UTRs appeared to be subtle and could be overshadowed by differences in promoter activities, for the 5’UTRs derived from several of the non-coding regions including A1/A2, A, and D, there were profound decreases in these parameters compared to other 5’UTRs. The biological significance of the N-terminally modified hRFC proteins translated from the A1/A2 [4
] and A 5’UTRs is unclear, however, these may be especially important for tissues that express high levels of these transcript and protein forms.
An important goal of future studies involves identification of putative physiologic effectors of hRFC transcriptional and posttranscriptional regulation, including dietary folates and/or other tissue-specific stimuli. Defects involving these processes may significantly impact net hRFC levels and activity in vivo
, including specialized tissue functions involved with reduced folate cofactors, thus contributing to the pathophysiology of folate deficiency (e.g., fetal abnormalities, cardiovascular disease, and cancer). Our recent results suggest that levels of hRFC-B transcripts are prognostic in children diagnosed with ALL whereas hRFC-A1/A2 transcripts are not [26
], in support of the notion that differences in posttranscriptional factors may represent important determinants of clinical responses to MTX chemotherapy, as well.