On BC1 RNA and the fragile X mental retardation protein

In Vitro Interactions of BC1 RNA with FMRP.

FMRP has been reported to bind to BC1 RNA under conditions of exceptionally high salt concentrations (750 mM NaCl plus 100 mM KCl) (17). We performed EMSAs to reassess binding of BC1 RNA to FMRP as a function of salt concentration in vitro. In the presence of 850 mM monovalent cations (buffer Z) (17) we were unable to detect any mobility shift of BC1 RNA over a wide range of FMRP concentrations (Fig. 1A). When salt concentrations in buffer Z were reduced to 100 mM KCl (modified buffer Z), a mobility shift of BC1 RNA became apparent at FMRP levels of 100 ng per assay and higher (Fig. 1A).

Notably, however, buffer Z does not use competitor RNA to reduce unspecific binding. EMSAs were therefore repeated under conditions of physiological salt concentrations (150 mM KCl, buffer B) (18) in the presence of 100 ng/μl tRNA (Fig. 1B). In this case, no mobility shift was observed except at an extreme level of FMRP (800 ng). N8 RNA, an FMR1 mRNA segment that does not specifically bind to FMRP (19), readily displaced BC1 RNA from FMRP in a manner analogous to tRNA (data not shown). In contrast, when we used N19 RNA, an FMR1 mRNA segment that contains a G-quartet-forming structure and is a genuine FMRP _target RNA (19), tRNA did not compete because a clear shift was observed in its presence over the entire FMRP concentration range (Fig. 1B).

For a quantitative assessment of BC1–FMRP interactions, we used a filter binding RNA competition assay (20) that allows appraisal of nonspecific binding contributions. As shown in Fig. 1C, addition of tRNA or BC1 RNA to the reaction had little effect on the binding of N19 RNA to FMRP, demonstrating that BC1 RNA is as ineffective as tRNA at displacing N19 RNA, a bona fide FMRP _target RNA. In contrast, both tRNA and N19 RNA displaced BC1 RNA from FMRP, as evidenced by a substantial reduction of binding in the presence of either competitor RNA (Fig. 1C). The data show that whereas N19 RNA, BC1 RNA, and tRNA appear to bind to the same site on FMRP, BC1 RNA but not N19 RNA is competed off this site by tRNA. In view of the these data, it is concluded that in vitro BC1–FMRP interactions are nonspecific and are of low affinity. It is possible that many RNAs bind to FMRP in such nonspecific, low-affinity manner in vitro (5).

Interactions of BC1 RNA with FMRP _target mRNAs.

We next turned to the question of whether BC1 RNA physically interacts with mRNAs that are _targets of FMRP regulation. Zalfa and colleagues (16, 17) proposed that the 5′ BC1 domain directly interacts, via base-pairing, with such FMRP _target mRNAs and that BC1 RNA thus bridges these mRNAs and FMRP. These authors probed BC1–mRNA interactions by annealing biotinylated BC1 RNA with total RNA from mouse brain and by subsequently amplifying RNAs that were recovered from streptavidin beads (17).

To reexamine this issue, we replicated the previously reported experiments under identical conditions and probed for FMRP _target and non_target RNAs (Fig. 2). Two approaches were used in parallel in these experiments. (i) Biotinylated BC1 RNA was incubated with total RNA isolated from mouse brain, as described (17). (ii) In addition, we used a biotinylated oligodeoxyribonucleotide complementary to the 3′ BC1 domain. This probe (called bio-unique) has recently been used for the specific capture, per affinity purification, of BC1 RNA from total mouse brain RNA (21). It will capture endogenous, native BC1 RNA together with any mRNA that may be bound to its 5′ domain, an advantage over the first approach, in which any mRNA bound to endogenous BC1 RNA has to be competed off and captured by externally added biotinylated BC1 RNA. For further controls, total mouse brain RNA was prepared from BC1^−/− [knockout (KO)] animals (22) in addition to WT animals.

As is shown in Fig. 2, we obtained RT-PCR products of microtubule-associated protein 1B (MAP1B), calcium/calmodulin-dependent kinase IIα (CaMKIIα), and Arc (Arg-3.1)mRNAs after capture with biotinylated BC1 RNA from WT and BC1 KO mouse brain total RNA (Fig. 2 A and B, lanes 1 and 4). Fig. 2C Right, lane 8, confirms that BC1 RNA is absent in total RNA from BC1 KO brains. The three FMRP _target mRNAs were also captured and amplified from WT mouse brain total RNA using the bio-unique oligonucleotide that is directed against the 3′ BC1 domain (lane 5 in Fig. 2 A and B). At the same time, however, both approaches also yielded amplification products of RNAs that have not been reported to be FMRP _targets, i.e., β-actin mRNA and U3 snoRNA (Fig. 2 B and C). Both RNAs are ubiquitous RNAs, the latter a small nucleolar RNA that participates in pre-rRNA processing. Neither RNA exhibits any significant sequence complementarity to BC1 RNA. The combined results let us conclude that the BC1 capture approaches used in this and a previous article (17) do not discriminate between FMRP _target and non_target mRNAs.

It appears that this lack of discrimination is the result of nonspecific interactions. Thus, we noted that, under the experimental conditions used and previously reported, MAP1B and CaMKIIα mRNAs also bind, albeit with lower apparent affinity, to the streptavidin matrix (Fig. 2 A and B, lane 6). Furthermore, when using the bio-unique oligonucleotide directed against the 3′ BC1 domain, RNAs were captured and amplified even from BC1 KO mouse brain total RNA (Fig. 2, lane 2). The results show that the chosen experimental routine is prone to yielding nonspecific amplification products.

The central domain of BC1 RNA contains a single-stranded homopolymeric A segment of 22 nucleotides. We therefore hypothesized that, in capture experiments in vitro, BC1 RNA may interact, by base-pairing of at least part of this A22 segment, with RNAs that contain U-rich segments. To test this hypothesis, we annealed total RNA from mouse brain to poly(A) RNA agarose (Fig. 3). Bound RNA was recovered, converted to cDNA, and amplified with gene-specific primers [supporting information (SI) Materials and Methods] using previously described conditions (17). From total mouse brain RNA we amplified Fmr1, eukaryotic elongation factor 1A (eEF1A), CaMKIIα, and MAP1B mRNAs (Fig. 3, lane 1). Each of these mRNAs was also present in the poly(A) annealed fraction (Fig. 3, lane 3). These four mRNAs are FMRP _targets that contain U-rich segments (23). However, we also recovered β-actin and dynamin A1 mRNAs (Fig. 3), mRNAs that feature U-rich segments but are not known to be FMRP _target mRNAs (23). The experimental approach therefore selects U-rich mRNAs regardless of whether they are FMRP _targets or not. The strength of the band representing the respective captured mRNA (Fig. 3, lane 3) corresponds to the length of U-element(s) in that mRNA (see SI Table 1). As a further control, we performed the annealing reaction in the presence of excess free poly(A) RNA to verify that amplification did not arise from contaminating unbound RNA (see SI Fig. 6).

Taken together, the above results lead to the conclusion that the chosen RNA capture and amplification procedure (17) is unable to substantiate any specific binding of BC1 RNA to FMRP _target mRNAs. Various RNAs are captured and amplified in this procedure because of (i) lack of specificity of the procedure per se and (ii) the selection of U-rich RNAs by an A-rich segment in BC1 RNA.

On the Interaction of BC1 RNA, FMRP, and FMRP _target mRNAs in Vivo.

The model proposed by Zalfa and colleagues (16, 17) specifically predicted that BC1 RNA is required as an adaptor to link FMRP with its _target mRNAs. If so, FMRP should be unable to interact with such _target mRNAs in the absence of BC1 RNA. We tested this hypothesis in vivo using BC1^−/− animals (22).

We performed immunoprecipitation (IP) with brain extracts prepared from WT and BC1^−/− (KO) animals, using FMRP-specific monoclonal antibody 7G1-1 (24). IP RNA was amplified by RT-PCR with mRNA-specific primers. We probed for CaMKIIα, Arc, and MAP1B mRNAs, i.e., those mRNAs that Zalfa et al. (17) proposed to be physically linked by BC1 RNA to FMRP. In addition, we probed for Fmr1 mRNA, another FMRP _target mRNA, and GAPDH mRNA, a FMRP non_target control mRNA (24).

As shown in Fig. 4, FMRP _target mRNAs, but not non_target GAPDH mRNA, were amplified from IP RNA. Notably, levels of none of the FMRP _target mRNAs that were amplified from IP RNA were found to differ discernibly between preparations from WT and BC1 KO brains (Fig. 4A). Quantitative and statistical analysis did not reveal any significant differences, for any of the probed FMRP _target mRNAs, between immunoprecipitated RNAs derived from WT and BC1 KO brains, respectively (Fig. 4B). The data show that FMRP associates with _target mRNAs in the absence of BC1 RNA, a result incompatible with a requirement for BC1 RNA to provide a physical link between FMRP and its _target mRNAs, as postulated by Zalfa and colleagues (16, 17).

Does BC1 RNA associate with FMRP in vivo? We performed IP experiments to address this question. FMRP (F-IP) and PABP (P-IP) were IP-_targeted in whole-brain homogenates prepared from fmr1^−/− (KO) and WT mice. PABP was chosen as a positive control because it has previously been shown to interact with BC1 RNA (4, 25, 26). F-IP from fmr1^−/− (KO F-IP) brains was performed to exclude the possibility of unspecific RNA precipitation. As an additional negative control, we performed IP with irrelevant IgGs (IgG-IP).

Extracted IP RNA was probed for BC1 RNA by RT-PCR and real-time PCR. The RT-PCR data show that BC1 RNA was identified in P-IP RNA but not in F-IP RNA (Fig. 5A). In contrast, FMRP _target PSD-95 mRNA (27) was readily detectable in both F-IP and P-IP (SI Fig. 7). These results were independently confirmed by real-time PCR performed with RNA from three independent tissue preparations. A high P-IP vs. F-IP ΔCT value of 14.01 ± 0.60 confirms that in WT brains BC1 RNA is detectable in PABP but not in FMRP immunoprecipitates. In addition, ΔCT values for KO F-IP vs. WT F-IP of 0.60 ± 0.62 and for WT IgG-IP vs. WT F-IP of 0.46 ± 0.29, i.e., close to 0 in either case, imply that BC1 levels in FMRP immunoprecipitates from WT brains are indistinguishable from background levels. Finally, when detecting PSD-95 mRNA, a ΔCT value for WT F-IP vs. KO F-IP of 3.01 ± 0.83 shows that the antibody efficiently immunoprecipitates PSD-95 mRNA associated with FMRP from WT but not KO brain homogenates. Thus, the lack of BC1 RNA in WT F-IP is not due to a general failure of the antibody to precipitate FMRP–RNA complexes, but to a genuine absence of BC1 RNA from such complexes.

The above experiments were performed with polyclonal anti-FMRP antibody H-120. Analogous work was performed, in separate experiments in a different laboratory, with monoclonal anti-FMRP antibody 7G1-1. Again, an anti-PABP antibody was used for positive control IP. In this case, F-IP RNA and P-IP RNA were probed for BC1 RNA by filter (Northern) hybridization (25). The results (Fig. 5B) demonstrate again that BC1 RNA is not detectable in FMRP immunoprecipitates although its presence in PABP immunoprecipitates is clearly evident. The combined data therefore show that BC1 RNA does not associate with FMRP in vivo.

EMSA and Filter Binding Assays.

His-tagged FMRP was expressed in Escherichia coli as described (36). BC1 and N19 RNAs were ³²P-labeled at 40,000 cpm per reaction (≈10 ng) and were incubated with increasing amounts of FMRP (0–800 ng per assay) in 20 μl of 10 mM Hepes (pH 7.9), 100 mM KCl, 750 mM NaCl, 3 mM MgCl₂, 7 mM 2-mercaptoethanol, 10 mM DTT, 5% glycerol, 1 μg/μl BSA, and 1.33 μg/μl heparin (buffer Z) (17) or in a buffer containing 50 mM Tris·HCl (pH 7.6), 150 mM KCl, 1 mM MgCl₂, 1 mM EDTA, 1 mM DTT, 0.2 units/μl RNase inhibitor (Amersham), 100 ng/μl total yeast tRNA, and 100 ng/μl BSA (buffer B) (18). If competitor RNAs were used, they were preincubated with FMRP for 10 min before addition of the labeled RNA.

Filter binding assays were performed as described (18). For EMSA experiments, RNA–protein complexes were resolved on 5% polyacrylamide gels. It should be noted that under the conditions used RNA–FMRP complexes are typically detected near the top of the gel (see also refs. 16 and 32). We presume that such bands represent homomeric FMRP complexes that bind RNA. However, in the interest of experimental conditions being consistent with previous work (16), we refrained from using dissociating agents such as urea.

In Vitro Annealing to Mouse Brain RNA.

Biotinylated BC1 RNA was bound to streptavidin-conjugated magnetic beads and incubated with total mouse brain RNA as described (17). In separate experiments, we used 100 pmol of a biotinylated oligonucleotide complementary to 3′ BC1 domain (bio-unique) (21). First-strand cDNA synthesis was performed by using Transcriptor reverse transcriptase (Roche Diagnostics) and random hexamer oligonucleotides, followed by 30 cycles of PCR amplification (17) with RNA-specific oligonucleotide primers.

Hybridization of mRNAs to poly(A) RNA.

Total RNA (1 μg) from mouse cortex was incubated with 20 μl of poly(A) RNA agarose resin (Sigma) at 80°C for 5 min in 20 μl of 10 mM Tris·HCl (pH 7.5), 2 mM MgCl₂, 400 mM NaCl, and 0.2% SDS (17) and allowed to cool at room temperature for 2 h. Nonbound material was removed with two washes (100 μl each) of the same buffer. Bound RNA was isolated with Tri-Reagent; after precipitation, half was converted to cDNA with the SuperScript III first-strand synthesis system (Invitrogen), and the other half was incubated in the absence of reverse transcriptase. Samples (1/20 aliquots) were amplified by PCR for the indicated mRNAs.

Immunoprecipitation.

IP with anti-FMRP was performed (i) to probe for FMRP _target mRNAs in the presence or absence of BC1 RNA and (ii) to probe for an association of BC1 RNA with FMRP. In the first approach we used a modified version of a previously published procedure (37). Dynabeads protein G slurry (Invitrogen) was washed with 100 mM sodium phosphate (pH 8.0) and incubated overnight at 4°C with 2.5 μg/μl monoclonal anti-FMRP 7G1-1 (Developmental Studies Hybridoma Bank) (24). Beads were incubated with brain extracts at 4°C for 2 h and washed extensively, and RNA was extracted by using TRIzol (Invitrogen).

In the second approach (variant S.K. laboratory), aliquots of brain extracts were incubated overnight with 15 μg of polyclonal anti-FMRP H-120 (Santa Cruz Biotechnology), affinity-purified polyclonal anti-PABP (38) (a gift of Evita Mohr, Institute of Anatomy I, Cellular Neurobiology, University Hospital Hamburg-Eppendorf, Hamburg, Germany), or rabbit IgG (Sigma–Aldrich). Samples were incubated for 40 min with 100 μl of preblocked protein A agarose. Beads were precipitated, washed, resuspended, and treated with proteinase K (Roche). RNA was extracted by using the RNeasy Mini Kit (Qiagen). In the second approach (variant H.T. laboratory), monoclonal anti-FMRP 7G1-1 was used. The anti-PABP used was described earlier (4); in addition, a newly generated affinity-purified polyclonal anti-PABP was used with identical results. IP methods used were those described for the first approach.

The fact that the laboratories contributing to this article used different materials and methods, as exemplified here by IP methods, is a consequence of their not being aware of each other's work. Regardless of the different methods used, however, the results obtained were congruent. Detailed IP methods are provided in SI Materials and Methods.