The B12/23 Restriction Is Critically Dependent on Recombination Signal Nonamer and Spacer Sequences¹

The cells of the adaptive immune response can recognize a near limitless number of foreign Ags. The basis for this diversity lies in the manner in which genes encoding Ag receptor chain variable regions are generated during lymphocyte development. Individual developing lymphocytes assemble Ag receptor variable region genes from component variable (V), joining (J), and, in some cases, diversity (D) gene segments by an enzymatic complex collectively referred to as the V(D)J recombinase. This process, termed V(D)J recombination, is regulated in several important contexts during lymphocyte development by constraints imposed on the V(D)J recombination reaction and through alterations in accessibility of gene segments to the V(D)J recombinase (1, 2).

V, D, and J gene segments are flanked by recombination signal (RS)⁴ sequences composed of conserved heptamers and nonamers flanking relatively nonconserved 12- or 23-bp spacers (hereafter referred to as 12RSs and 23RSs, respectively) (3). Recombination occurs only between gene segments flanked by RSs of dissimilar spacer length, a restriction known as the 12/23 rule (3). The V(D)J recombination reaction can be generally divided into DNA cleavage and joining steps. The DNA cleavage step is initiated through the formation of a synaptic complex that includes an appropriate RS pair and the recombinase-activating gene (RAG) 1 and 2 proteins (4, 5). In the context of a synaptic complex, the RAG-1/2 proteins introduce DNA double-strand breaks at the RS coding segment border, resulting in the formation of blunt phosphorylated signal ends and hairpin-sealed coding ends. Processing and joining of these DNA ends is mediated by proteins of the nonhomologous end-joining pathway of DNA double-strand break repair (1).

RSs can impose significant constraints on variable region gene assembly beyond enforcing the 12/23 rule. This restriction, termed B12/23, has been defined in the TCRβ locus (6, 7). TCRβ variable region genes are assembled from Vβ, Dβ, and Jβ gene segments. The murine TCRβ locus is composed of ∼35 Vβ gene segments and two Dβ-Jβ gene segment clusters, each with a single Dβ gene segment (Dβ1 and Dβ2) and six functional Jβ gene segments (6). Vβ and Jβ gene segments are flanked by 23RSs and 12RSs, respectively. Dβ gene segments are flanked 5′ by 12RSs and 3′ by 23RSs. Although direct Vβ to Jβ rearrangement is permitted by the 12/23 rule, Dβ gene segments are used in the assembly of essentially all TCRβ variable region genes (8). This is due to a requirement for the 5′Dβ 12RS to efficiently target rearrangement of Vβ 23RSs beyond simply enforcing the 12/23 rule (6, 7). At a minimum, the B12/23 restriction ensures Dβ gene segment utilization which is important for generating a diverse repertoire of functional TCRβ chains (8).

The B12/23 restriction is imposed at or before the DNA cleavage step of the V(D)J recombination reaction and does not exhibit an absolute requirement for lymphoid-specific factors other than the RAG-1/2 proteins (9, 10, 11). Furthermore, the Dβ1 12RS rearranges efficiently with Vβ and Jκ 23RSs, demonstrating that Dβ 12RSs enforce the B12/23 restriction in a manner that does not require specific Vβ/Dβ RS synapsis (10). The 5′ Dβ 12RSs have heptamers and nonamers that approximate consensus heptamer and nonamer sequences. Notably, a cross-species comparison of Dβ 12RSs spacer sequences reveals a relatively high level of sequence conservation (see below). Finally, the mouse 5′ Dβ1 12RS contains a consensus TATA box used by the PDβ promoter for the initiation of germline (GL) Dβ1-Jβ1 gene segment cluster transcripts (12, 13, 14). Although this consensus TATA box is not conserved in other Dβ 12RSs, it is not known whether these RSs possess other transcript initiation sequences that could contribute to the B12/23 restriction.

In this study, we assess the requirements for different components of the 5′ Dβ RS in enforcing the B12/23 restriction. For this purpose, we generate mutant 5′ Dβ1 RSs and analyze their function in a TCRβ minilocus that undergoes Dβ to Jβ and Vβ to DJβ rearrangement in developing thymocytes and in extrachromosomal substrates that can recombine in nonlymphoid cell lines. Our findings have important implications for the manner in which rearrangement is regulated within the context of the B12/23 rule and for the assembly of RS synaptic complexes in vivo.

Isolation of genomic DNA and Southern blot analysis were conducted as previously described using a 600-bp AccI fragment spanning the Vβ14 gene segment (probe P) (7, 15). Band intensities were determined using a PhosphorImager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA). Band intensities were corrected for background and the ratio of VDJ rearrangement as a function of total transgene rearrangement was calculated as VDJ band intensity/(DJ band intensity + VDJ band intensity).

The M12 and ES cell lines were maintained and transfected with the TCRβ^SPF minilocus as previously described (7, 12).

mRNA was isolated using TRIzol (Invitrogen, Carlsbad, CA). For the generation of S1 nuclease protection probes, 406-bp genomic DNA fragments spanning from immediately upstream of the 5′ Dβ1 RSs to the BglII site 3′ of the Dβ1 gene segment were subcloned into pBSSK. The 471-bp S1 nuclease probes were then generated by PCR using the T7 primer and the ClaIS1 primer: 5′-AGATCGATCTTTTAAAACAAAAC-3′. This results in S1 probes with ∼65 bp of nonhomologous polylinker DNA followed by 406 bp that are homologous to the 5′ Dβ1 RSs and downstream region. Initial hybridizations were conducted with 40 μg whole cell RNA and excess end-labeled probes in 15 μl of 80% Formamide (Fluka, Buchs, Switzerland), 40 mM PIPES (pH 6.4), 400 mM NaCl, and 1 mM EDTA at 51°C for 12 h after incubation for 5 min at 100°C. S1 nuclease digestion was conducted in 300 μl of 37°C for 30 min with 300–400 U of S1 nuclease (Promega, Madison, WI). Samples were size fractionated on 7 M urea/8% polyacrylamide gels.

Transient recombination assays using the pC substrate were performed and analyzed as previously described (10).

The TCRβ^PF minilocus is composed of a single Vβ gene segment (Vβ14), Dβ gene segment (Dβ1), and two Jβ gene segments (Jβ1.1 and Jβ1.2) linked to the IgH intronic enhancer (Eμ) and constant region gene (Cμ; Fig. 1 A) (7). Efficient DJβ and VDJβ rearrangement of this minilocus occurs during thymocyte development in chimeric mice generated by RAG-2-deficient blastocyst complementation (RDBC) using embryonic stem (ES) cells with integrated copies of the TCRβ^PF minilocus (7). Furthermore, as is the case with the endogenous TCRβ locus, the 5′ Dβ1 12RS is required to efficiently target Vβ rearrangement in the minilocus during thymocyte development due to constraints beyond simply enforcing the 12/23 rule (7).

The TCRβ^SPF:Dβ1 minilocus was generated from the TCRβ^PF minilocus by introducing restriction sites (HpaI and NheI) that flank the 5′ Dβ 12-RS, allowing for easy replacement of this RS with mutant RSs (Fig. 1,A). Thus, the TCRβ^PF and TCRβ^SPF:Dβ1 miniloci are identical except for the introduction of these restriction sites. Chimeric mice were generated by RDBC using four independently derived ES cell lines (TCRβ^SPF:Dβ1–163, 169, 184, and 189) containing two to four stably integrated copies of the TCRβ^SPF:Dβ1 minilocus (Fig. 1,B) (7, 16). Rearrangement of the TCRβ^SPF:Dβ1 minilocus was assayed by Southern blotting of BamHI- and BglII-digested thymocyte genomic DNA isolated from chimeric mice with >5 × 10⁷ thymocytes and normal fractions of thymocyte subsets as assessed by flow cytometry (Fig. 1,B and data not shown). Robust levels of Dβ to Jβ and Vβ to DJβ rearrangement of the TCRβ^SPF:Dβ1 minilocus were observed in thymocytes from chimeric mice generated from all four TCRβ^SPF:Dβ1 minilocus containing ES cells (Fig. 1 B).

Bands generated by probe hybridization to miniloci in the unrearranged (GL) DJβ and VDJβ configurations were quantitated by densitometry and the percentage of rearranged miniloci in the VDJβ configuration was calculated as described in Materials and Methods (Table I). Since local integration effects would impact Dβ to Jβ and Vβ to DJβ rearrangement, the fraction of the minilocus in the VDJβ configuration (VDJβ/VDJβ + DJβ) was used to permit quantitative comparisons of the level of Vβ to DJβ rearrangement between thymocytes from chimeric mice generated from ES cell lines with distinct minilocus integrants. The fraction of rearranged TCRβ^SPF:Dβ1 miniloci in the VDJβ configuration was similar (37–54%) when comparing seven chimeric mice derived from four independent-derived ES cells (Table I). Together these data demonstrate that during thymocyte development, the TCRβ^SPF:Dβ1 minilocus undergoes consistently efficient Vβ to DJβ rearrangement.

In the endogenous TCRβ locus, replacing the 5′ Dβ1 12RS with the Jβ1.2 12RS results in a block in Vβ to Dβ rearrangement (6). The TCRβ^SPF:JβH/N minilocus was generated by replacing 5′ Dβ1 12RS of the TCRβ^SPF:Dβ1 minilocus with a chimeric RS (JβH/N) composed of the Jβ1.2 heptamer/nonamer sequences and the 5′ Dβ1 12RS spacer sequence (Figs. 2 and 3,A). Thus, the TCRβ^SPF:JβH/N and TCRβ^SPF:Dβ1 miniloci are identical except for the altered 5′ Dβ1 12RS heptamer/nonamer sequences. Chimeric mice were generated by RDBC from two independently derived ES lines harboring the TCRβ^SPF:JβH/N minilocus (TCRβ^SPF:JβH/N-26 and -33; Fig. 3,A). Southern blot analysis of thymocytes isolated from these mice revealed that the TCRβ^SPF:JβH/N minilocus undergoes efficient Dβ to Jβ rearrangement but has a severe block in Vβ to DJβ rearrangement (Fig. 3,A and Table I).

To investigate the individual contributions of the heptamer and nonamer, chimeric RSs composed of the Jβ1.2 heptamer and 5′ Dβ1 spacer and nonamer (JβH) or the Jβ1.2 nonamer with the 5′ Dβ1 heptamer and spacer (JβN) were generated (Figs. 2 and 3, B and C). Whereas the TCRβ^SPF:JβH minilocus exhibits a modest reduction (∼2-fold) in VDJβ rearrangement, the TCRβ^SPF:JβN minilocus exhibits profound reduction in VDJβ rearrangement (Fig. 3,C and Table I). Together these findings demonstrate that the 5′ Dβ1 12RS nonamer sequence is critical for enforcing the B12/23 restriction.

The JβSP 12RS is composed of the 5′ Dβ 12RS heptamer and nonamer flanking the Jβ1.2 spacer (Fig. 2). Strikingly, the TCRβ^SPF:JβSP minilocus exhibits a reduction in Vβ to DJβ rearrangement of similar magnitude to that observed for the TCRβ^SPF:JβN minilocus (Fig. 3,A and Table I). Unlike the Jβ 12RS spacers, the 5′ Dβ 12RS spacers exhibit a high degree of sequence homology across different mammalian species (Fig. 4). Five of the 12 nt are absolutely conserved and 2 additional nucleotides are conserved in all but single RSs (Fig. 4). To test whether these conserved nucleotides are important for enforcing the B12/23 restriction, the CSM 12RS was generated by replacing the seven conserved nucleotides in the 5′ Dβ 12-RS with the corresponding bases in the Jβ1.2 spacer resulting in 4 bp changes (Fig. 2). Similar to what was observed for the TCRβ^SPF:JβSP minilocus, a severe reduction in Vβ to DJβ rearrangement was observed for the TCRβ^SPF:CSM minilocus, highlighting the importance of these conserved spacer nucleotides in enforcing the B12/23 restriction (Table I).

As demonstrated above, targeting of Vβ rearrangement by the 5′ Dβ1 12 RS relies on features of the nonamer and spacer sequences. A consensus TATA box, utilized by the PDβ1 promoter, spans these sequences and is not present in the mutant 12RSs (JβSP, CSM, JβH/N, and JβN) that fail to efficiently target Vβ rearrangement (Fig. 2). Unlike the 5′ Dβ1 12RS, the 5′ Dβ2 12RS does not contain a consensus TATA box sequence (Fig. 2). To determine whether this RS is capable of efficiently targeting the Vβ rearrangement, the TCRβ^SPF:Dβ2 minilocus was generated in which the 5′ Dβ1 12RS was replaced with the 5′ Dβ2 12RS (Fig. 2). This minilocus undergoes efficient Vβ to DJβ rearrangement, demonstrating that the 5′ Dβ2 12RS is capable of efficiently targeting Vβ rearrangement (Fig. 5 and Table I).

Although lacking a consensus TATA box, the 5′ Dβ2 12RS may possess cryptic transcript initiation sequences. To investigate this possibility, the TCRβ^SPF:Dβ1 and TCRβ^SPF:Dβ2 miniloci were stably introduced into the M12 cell line and transcript initiation from the PDβ1 promoter was assayed by S1 nuclease protection (Fig. 6). These analyses revealed a predominant transcript initiating from the TCRβ^SPF:Dβ1 minilocus ∼20 bp downstream of the 5′ Dβ1 RS TATA (+20, Fig. 6). These transcripts are similar to PDβ1 promoter-specific transcripts previously observed in the endogenous locus during thymocyte development and in the TCRβ minilocus in the M12 cell line (12, 14). Transcripts that initiate at this position or others downstream of the 5′ Dβ2 RS were notably absent in M12 cells harboring the TCRβ^SPF:Dβ2 minilocus (Fig. 6). Together, these findings demonstrate that, unlike the 5′ Dβ1 12RS, the 5′ Dβ2 12RS contains neither a consensus TATA box nor sequences that mediate transcript initiation immediately downstream of the RS, yet the 5′ Dβ2 12RS is capable of efficiently targeting the Vβ rearrangement.

The TGT 12RS was generated by converting the TATAAA TATA box in the 5′ Dβ1 12RS to TGTAAA (Fig. 2). Analysis of transcripts initiating from the TCRβ^SPF:TGT minilocus in M12 cells revealed a complete loss of the TATA-specific +20 transcripts observed in the TCRβ^SPF:Dβ1 minilocus (Fig. 6). However, the TCRβ^SPF:TGT minilocus exhibited only a mild reduction (∼2-fold) in Vβ to DJβ rearrangement in developing thymocytes (Table I). Together, these findings demonstrate that a functional TATA box in, or transcript initiation from, the 5′ Dβ 12RSs is not required for enforcement of the B12/23 restriction.

The pC competitive extrachromosomal recombination substrate has three positions (P1, P2, and P3) for RSs cloning (Fig. 7) (10). The pC substrate containing appropriate RS combinations can undergo rearrangement in nonlymphoid cell lines that transiently express the RAG-1/2 proteins (10, 17). Rearrangement of the RSs cloned at P1 to RSs cloned at P2 or P3 deletes the transcriptional terminator, permitting expression of the chloramphenicol acetyltranferase (CAT) gene which allows quantitation of rearrangement efficiency after bacterial transformation as previously described (10, 17). Rearrangement of the RS at P1 to the RS at P2 or P3 is assayed by PCR.

The pC:V¹⁴DD substrate has the Vβ14 23RS at P1 and the 5′ Dβ1 12RSS at P2 and P3 (10). As previously demonstrated, an intrinsic bias for rearrangement of the Vβ14 23RS to the 5′ Dβ1 12RSs at P2 or P3 is not observed in Chinese hamster ovary cells expressing the RAG-1/2 proteins (Table II) (10). The pC:V¹⁴DJ^H/N substrate was generated from the pC:V¹⁴DD substrate by replacing the 5′ Dβ1 12RS at P3 with the Jβ H/N 12RSS. This substrate exhibits a profound bias for rearrangement of the Vβ14 23RS to the 5′ Dβ1 12RS at P2 over the Jβ H/N 12RS at P3 similar in magnitude to that observed for the pC:V¹⁴DJ^1.2 substrate which has the Jβ1.2 12RS at P3 (Table II) (10). Analysis of extrachromosomal recombination substrates with the Jβ H (pC:V¹⁴DJ^H) and Jβ N (pC:V¹⁴DJ^N) RSs reveals that replacing the 5′ Dβ RS nonamer (Jβ N) has a more profound effect on Vβ RS targeting than replacing the heptamer (Jβ H) which had only a modest effect (Table II). Finally, the pC:V¹⁴DD^C substrate exhibits severely diminished levels of rearrangement of the Vβ 23RS to the CSM 12RS, as compared with the 5′ Dβ 12RS (Table II). Together these findings demonstrate that, similar to what was observed for rearrangement of the minilocus in vivo during thymocyte development, enforcing the B12/23 restriction on extrachromosomal substrates in nonlymphoid cells appears to be critically dependent on the 5′ Dβ1 12RS nonamer and spacer.

Assembly of TCRβ variable region genes during lymphocyte development is ordered with Dβ to Jβ rearrangement preceding Vβ to DJβ rearrangement. Jβ 12RSs readily rearrange with 3′ Dβ 23RSs yet fail to rearrange efficiently with Vβ 23RSs that rearrange efficiently with 5′ Dβ 12RSs due to B12/23 constraints imposed on Vβ rearrangement by the 5′ Dβ 12RSs (6, 7). In this study, we show that the B12/23 restriction is critically dependent on the nonamer and spacer sequences of the 5′ Dβ 12RS. This nonamer and spacer sequence requirement was observed for rearrangement both on chromosomal substrates in the setting of thymocyte development and on extrachromosomal substrates in nonlymphoid cell lines.

The CAC trinucleotide of the consensus heptamer sequence (CACAGTG) is essential for RS function (18, 19). As expected, this CAC trinucleotide is absolutely conserved in the 5′ Dβ and Jβ 12RS heptamer sequences. With respect to the four remaining nucleotides, the Dβ 12RS heptamers (CACAATG) differ from consensus at one position whereas the Jβ 12RSs, except for Jβ1.1, differ from consensus at two or three of these positions (20). Deviation of these four nucleotides from consensus can affect the efficiency of recombination of extrachromosomal substrates (18, 19). However, the heptamer sequence differences between the 5′ Dβ and Jβ 12RS do not appear to be the critical factor for enforcing the B12/23 rule as evidenced by the modest reduction (∼2-fold) in Vβ 23RSS rearrangement to the JβH 12RSS on chromosomal and extrachromosomal substrates. Thus, features of the 5′ Dβ 12RSS heptamer, other than the essential CAC trinucleotide, do not significantly impact the B12/23 restriction imposed on Vβ to Dβ rearrangement.

In striking contrast to the heptamer, replacing the nonamer of the 5′ Dβ 12RS with the Jβ1.2 nonamer (JβN RS) results in a profound block in Vβ to Dβ rearrangement of the TCRβ minilocus and on extrachromosomal substrates. Nonamer sequences vary widely with some deviating considerably from consensus (ACAAAAACC). Although there are no strict nonamer sequence requirements, alterations in one or both bases of the nonamer AA dinucleotide (ACAAAAACC) have led to reduced recombination efficiency on extrachromosomal substrates (18). This AA dinucleotide is intact in the 5′ Dβ, but not the Jβ1.2, 12RS nonamer. Although this difference may contribute to the B12/23 restriction, it is not likely to be the sole determinant of the nonamer requirement as many other Jβ 12RS nonamers have the AA dinucleotide (20).

The strict requirement for RS spacer nucleotide length is well established (18, 19). Single nucleotide deviations in spacer nucleotide length result in a reduction in rearrangement efficiency, and a gain or loss of two or more nucleotides essentially abolishes RS function (18, 19). Although the importance of spacer nucleotide sequence is less well established, a low level of spacer sequence conservation has been noted upon analysis of a large number of RSs (20). Furthermore, spacer sequence variations have been implicated in affecting the efficiency of RS utilization (21, 22, 23, 24). Perhaps most notably, Vκ RS spacer sequences affect rearrangement efficiency on extrachromosomal substrates with these effects likely impacting differential Vκ gene segment utilization in vivo (22). Analysis of the 5′ Dβ 12RS spacer sequences across species reveals a remarkable level of conservation with 5 of the 12 nt absolutely conserved and 7 conserved in all but 2 of the Dβ 12RSs. This high level of conservation is not observed for Jβ 12RS spacer sequences. Strikingly, replacing four of the conserved 5′ Dβ 12RS spacer nucleotides with the corresponding Jβ1.2 12RS spacer nucleotides (CSM 12RS) results in a profound reduction in Vβ to Dβ rearrangement both in the minilocus and on extrachromosomal substrates. Thus, the 5′ Dβ 12RS spacer nucleotides are critical for enforcing the B12/23 restriction.

It is conceivable that independent mechanistic constraints rely on features of the nonamer and spacer sequences to enforce the B12/23 restriction. It is unlikely, though, that this reflects a requirement for the binding of lymphoid-specific factors to nonamer and/or spacer sequences since these sequences are required to enforce the B12/23 restriction on extrachromosomal substrates in a nonlymphoid cell line. Alternatively, and perhaps more plausibly, a common B12/23 mechanistic constraint relies on features of both the nonamer and spacer sequences. Notably, this is not due to a requirement for transcription initiation sequences spanning the nonamer and spacer such as the TATA box present in the 5′ Dβ1 RS. The 5′ Dβ2 12RS efficiently targets Vβ rearrangement yet lacks a TATA box or other sequences that serve to initiate transcription. Furthermore, mutation of the TATA box in the 5′ Dβ1 RS has only a modest effect on targeting the Vβ rearrangement.

The B12/23 restriction is enforced at or before the DNA cleavage step of the V(D)J recombination reaction and does not absolutely require lymphoid-specific factors (9, 10, 11). Furthermore, specific synapsis between the 5′ Dβ 12RS and a Vβ 23RS is also not required (9). Within these contexts how might the 5′ Dβ 12RS nonamer and spacer sequences enforce the B12/23 restriction? Synaptic complex formation is initiated by RAG-1/2 binding to an RS. Purified RAG-1 can bind a single RSS in vitro (25, 26, 27). Although RAG-1 makes contacts with RS nonamer and spacer nucleotides, binding appears to rely primarily on the contacts made with the nonamer (25, 26, 28, 29, 30). In the case of 12RSs, the addition of RAG-2 to this complex promotes more extensive contacts with the spacer nucleotides, including those proximal to the heptamer that are highly conserved in the 5′ Dβ 12RS spacer (28, 29, 30, 31). The notion that these contacts may be important for RAG-1/2 binding to 12RSs is supported by the partial disruption of this binding upon chemical modification of the heptamer proximal spacer nucleotides (29). Thus, it is possible that the nonamer and spacer sequence requirements for the B12/23 restriction reflect, in part, a requirement for efficient binding of RAG-1/2 to the 5′ Dβ 12RS before functional synaptic complex formation with a Vβ 23RS.

Functional synaptic complex formation in vitro occurs most efficiently between a RS bound by RAG-1/2 and an unbound RSS (32, 33). Thus, the 5′ Dβ 12RS may serve to “nucleate” the formation of a functional Vβ 23RS/Dβ 12RS synaptic complex in vivo by efficiently binding RAG-1/2 followed by complex formation with an unbound Vβ 23RS. In this regard, the Jβ 12RSs may bind RAG-1/2 much less efficiently and as such be unable to readily nucleate functional synapsis with Vβ 23RSs. The notion that the Jβ 12RSs are generally less efficient at mediating recombination than the Dβ 12RSs is consistent with recent analyses of the functional effect of RS sequence variations (34, 35). However, rearrangement between the 3′ Dβ 23RS and Jβ 12RSs occurs efficiently in the endogenous TCRβ locus and on extrachromosomal recombination substrates (11). In this regard, it is conceivable that the 3′ Dβ 23RS functions to nucleate functional synapsis for Dβ to Jβ rearrangement. The notion that the 3′ Dβ 23RSs are generally more efficient at mediating recombination than the Vβ 23RSs is consistent with the observation that replacing the Vβ14 23RS with the 3′ Dβ 23RS in the endogenous TCRβ locus results in a dramatic increase in Vβ14 gene segment utilization (36). Thus, it is plausible that the B12/23 restriction may reflect a requirement that, at a minimum, one RS of a recombining pair has the capacity to nucleate functional synaptic complex formation through efficient and independent binding of the RAG-1/2 proteins. In this regard, specific nonamer and spacer sequence combinations may serve as optimal substrates for RAG-1/2 binding. Importantly, this would not preclude additional roles for these sequences in promoting synaptic complex stability once the complex forms (37). The general assembly of functional synaptic complexes in this manner in vivo has potentially important evolutionary implications because it would require that the sequence features permitting efficient nucleation of RAG-1/2 binding be strictly conserved at a single RS of a functional RS pair.

Note added in proof. Since the acceptance of this manuscript, three papers (38, 39, 40) with findings relevant to spacer and nonamer sequence function have been published.