A Spontaneous RAG1 Nonsense Mutation Unveils Naturally Occurring N-Terminal Truncated RAG1 Isoforms

The RAG1/RAG2 hetero-tetramer has been extensively studied in its role as the enzyme responsible for rearranging Ag receptor genes in B and T cells in the process known as VDJ recombination (1, 2). RAG1 contains the catalytic domain necessary for DNA cleavage for this process, and thus, patients with loss-of-function mutations in either RAG1 or RAG2 suffer from SCID with complete loss of mature T and B cells. In contrast, RAG1/2 mutations that reduce but do not ablate recombination activity lead to Omenn syndrome (OS), an immunodeficiency of low B and T cell numbers, concurrent with autoimmunity (3, 4). Notably, Santagata et al. (5) described patients with homozygous frameshift mutations in the N-terminal region of RAG1 that result in early translation termination and should not yield functional RAG1 protein. Despite this, these patients had low but detectable circulating B and T cells. RAG1 is translated from a single exon; thus alternative splicing was not considered. It was hypothesized that N-truncated RAG1 products arise from internal translation initiation at downstream AUG start sites, leading to a hypomorphic protein. Correlative with this hypothesis, expression of N-truncated cDNA gives rise to RAG1 proteins of smaller size that catalyze VDJ reactions in vitro. However, neither the mechanism by which these truncated RAG1 proteins are made nor whether they occur normally were elucidated (5).

During a routine analysis, we identified a strain of mice in our colony with reduced T and B cells and increased activated T cell frequencies. Whole-exome sequencing revealed a novel homozygous nonsense mutation at aa 60 (Q60X), which we termed the N-terminal stop or RAG1^NX mouse. This mutation, like frameshift mutations present in OS patients, would be predicted to cause premature translation termination and absence of functional RAG1. However, the presence of T and B cells in these mice, although at lower numbers, necessitated the presence of residual RAG1 activity. In this study, we show that the Q60X mutation in RAG1^NX thymocytes results in expression of smaller RAG1 isoforms. Similarly sized smaller RAG1 isoforms are present in homozygous RAG1^WT thymocytes. Further, in vitro experiments suggest that these smaller RAG1 isoforms are generated using internal translation initiation sites (TIS). Our data demonstrate that an underappreciated mechanism of internal translation likely leads to multiple RAG1 isoforms in wild-type (WT) mice and allows for the escape from early truncating mutations in the gene that would otherwise have devastating immune consequences. There is significant evidence that the N-terminal regions of RAG1 have important roles in VDJ recombination (6–10), and RAG1 has been studied as a single isoform, containing the complete N-terminal region for over 30 y. The conceptual advance that multiple N-truncated RAG1 isoforms exist, even in WT cells, is important as these N-truncated proteins demonstrate altered function and have not previously been interrogated as part of RAG biology. This novel murine model of OS provides an important platform to determine how N-truncated RAG1 proteins function normally and in the absence of full-length RAG1 protein.

C57BL/6-RAG1^WT mice were purchased from The Jackson Laboratory and bred in our facility. C57BL/6-RAG1^NX and -RAG1^Het mice were bred in our facility. All animal studies were performed with the approval of The Children’s Hospital of Philadelphia Institutional Animal Care and Use Committee.

Genomic DNA was isolated from an RAG1^NX spleen using the DNeasy Blood and Tissue Kit (Qiagen) per manufacturer’s instructions. Exon capture for whole-exome sequencing was performed using the Agilent SureSelect XT Mouse All Exon kit. Sequencing was performed on the Illumina HiSeq 4000 to produce 150-bp paired-end reads with an average depth ×100. Sequence reads were aligned to the reference mouse genome (GRCm38/mm10) using Novoalign (V3.03.01; http://www.novocraft.com). Picard was used for marking duplicates, and then variants were called using GATK’s HaplotypeCaller. Single-nucleotide variants and insertions/deletions were functionally annotated with SnpEff (http://snpeff.sourceforge.net) and filtered to retain only moderate- and high-effect variants. After applying quality filters and excluding variants that have been reported in the Single Nucleotide Polymorphism Database, 299 variants remained. Genes were then annotated using gene ontologies to identify variants in genes involved in B and T cell development. A homozygous stop gain mutation was identified in Rag1 (NM_009019.2; p.Gln60*) that met all filtering and quality-control criteria.

Splenocytes, thymocytes, and bone marrow leukocytes were stained with LIVE/DEAD fixable viability dye (Life Technologies) and Abs against respective surface Ags (BD Pharmingen, eBioscience, and BioLegend). For intracellular HA, FLAG, GFP, and Foxp3 staining, cells were stained using the eBioscience Foxp3 kit according to manufacturer’s instructions. All samples were acquired on an MACSQuant flow cytometer (Miltenyi Biotec) or LSRII Fortessa (BD Biosciences) and analyzed using FlowJo software version 10.5.3 (Tree Star).

Splenocytes (10⁶) were cultured in the absence or presence of 50 ng/ml PMA (Sigma-Aldrich) and 1 μg/ml ionomycin (Cell Signaling Technology), with 2 μg/ml brefeldin A (Sigma-Aldrich) and 2 μM monensin (eBioscience) for 5 h at 37°C. After staining for LIVE/DEAD and for surface Ags as described earlier, cells were stained for the respective cytokines using the Cytofix/Cytoperm kit according to manufacturer’s instructions (BD Bioscience).

Relevant thymocyte populations were stained as described above, and sorted using a FACSAria Fusion. RNA was isolated using RNeasy Mini kit (Qiagen) and reverse transcribed to cDNA using SuperScript III Reverse Transcriptase (Invitrogen) with random hexamers. Quantitative PCR (qPCR) was performed using Power SYBR Green (Applied Biosystems), and HPRT or RAG1 primers (Qiagen).

Thymocytes were isolated, then stained with PE-labeled anti-CD4, CD8, CD11b, CD11c, NK1.1, Gr1, and Ter119. Nonlabeled cells were enriched by MACS depletion using anti-PE microbeads and LS columns (Miltenyi). Enriched cells were then stained with CD4, CD8, CD44, and CD25. DN3 cells were sorted using a FACSAria Fusion. Genomic DNA was extracted from the sorted DN3 cells using the DNeasy Blood and Tissue Kit (QIAGEN). A qPCR assay to measure Vβ-Dβ1-Jβ1 and Vβ-Dβ2-Jβ2 rearrangement frequencies was designed with a panel of primers specific for each functional Vβ paired with a probe (FAM, HEX) specific for either Jβ1.1 or Jβ2.1, respectively. Rearrangements were measured by TaqMan qPCR with PCR conditions according to the manufacturer’s instructions (PrimeTime; Integrated DNA Technologies). PCR analysis of TRDV1-JD1 and CD19 were used for normalization. Primers and probes have been described previously (11).

Hep2 anti-nuclear Ab (ANA) slides (MBL International) were incubated with mouse serum diluted 100-fold in PBS for 15 min in the dark. Slides were washed with PBS and stained with DAPI (Thermo Fisher Scientific) and anti-mouse Ig-AF488 (Jackson ImmunoResearch Laboratories). Images were acquired on a 20× plan apochromat and 1.4 numerical aperture objective on a spinning disc confocal system (UltraView ERS 6; PerkinElmer, Waltham, MA) equipped with an ORCA-ER camera (Hamamatsu Photonics, Bridgewater, NJ) and velocity software (v6.1.1; PerkinElmer). Instrument settings were fixed for all images, and researchers were blinded to the sample identification. Nuclear AF488 intensity was quantified using FIJI software, using DAPI to designate the nuclear region of interest. A minimum of 30 nuclei were analyzed per well, and the average AF488 intensity was reported.

FLAG-RAG1-HA was amplified by PCR using the murine RAG1-HA template from the Bassing laboratory and cloned into the NotI and XhoI sites of pZHK:CMV-IRES-GFP (12) to create the base pZHK:CMV-FLAG-RAG1(WT)-HA-IRES-GFP construct. PCR were performed using Takara ex-taq DNA polymerase (Takara).

The RAG1 cloning primers were as follows: forward: 5′-TGAGCGGGTACCCAATTGGCCAATTGGGATCCGCGG-3′ and reverse: 5′-TGAGCGGTCGACCAATTGGCCAATTGCTCGAGTCTA-3′.

For mutagenesis of methionines and Q60X, RAG1 was subcloned into pUC18, and mutagenesis was performed using Quikchange Lightning Kit (Agilent). Mutagenesis primers were designed using the online Quikchange Primer Design software (Agilent). Correctly mutated fragments were swapped back into the pZHK vector using NotI and XhoI sites for expression.

293T cells were grown in six-well plates to 70–80% confluency. 293T cells were transfected with 2 μg of the respective RAG1 constructs using TransIT-Lenti reagent (Mirus Bio) according to manufacturer’s instructions. Cells were washed 2× with PBS 16–20 h posttransfection for flow cytometry or Western blot.

Native RAG1 Western blotting was performed as previously described (13) using rabbit anti-mouse RAG1 Ab from David Schatz (Yale). For RAG1 expressed in 293T cells, transfected cells were lysed using M-PER lysis buffer (Thermo Fisher Scientific) with protease inhibitor mixture (Thermo Fisher Scientific). Fifteen micrograms of protein was boiled in loading buffer with 5% β-ME for 10 min and run on a 4–12% SDS-PAGE gel before transfer to nitrocellulose membrane (Bio-Rad). Membranes were probed with anti-HA (Clone 3F10; Sigma-Aldrich) and anti-FLAG (Clone M2; Sigma-Aldrich) and then AF700 and AF780 conjugated secondary Abs (Invitrogen) before imaging on a Licor Odyssey.

All data were analyzed in GraphPad Prism 8 using statistical tests indicated in the figure legends. Error bars indicate mean ± SEM. The p values < 0.05 are considered significant (*p < 0.05, **p < 0.01, ***p < 0.0001, and ****p < 0.0001). A p value > 0.05 was not significant.

Routine analysis of PBMCs in a line of C57BL/6 mice within our colony revealed B and T cell lymphopenia of unknown cause (data not shown). Whole-exome sequencing of these mice was performed, and 1386 de novo mutations were identified compared with the reference C56BL/6 genome (mm10). A homozygous nonsense mutation early in the Rag1 gene at nucleotide position 187 (c187C > T) was identified as a mutation likely to drive the observed phenotype. This mutation was confirmed by Sanger sequencing (Fig. 1A, 1B). These mice were backcrossed a minimum of three generations to WT C57BL/6 mice acquired from The Jackson Laboratory while selecting for the mutant RAG1 allele to decrease the likelihood that the other de novo mutations could influence the observed phenotype. This mutation introduces a stop codon that is predicted to terminate translation at aa 60 of RAG1 (Q60X) and result in no functional RAG1 (Fig. 1A, 1B). We named this the RAG1^NX mouse. However, unlike complete RAG1 deficiency, this homozygous mutation does not result in complete absence of T and B cells. We therefore checked for RAG1 mRNA by qPCR and RAG1 protein by Western blot from homozygous RAG1^NX thymocytes and show an increase in RAG1 mRNA and the formation of a number of proteins of smaller sizes than full-length, WT RAG1 (Fig. 1C, 1D). Truncating frameshift mutations in the 5′ region of the Rag1 gene in human patients are unexpectedly associated with OS rather than SCID (5). Thus, we considered that the RAG1^NX mouse may represent a model OS to investigate how such mutations retain enough RAG activity to prevent complete SCID.

Given the mutation in RAG1, we sought to determine the effect on T and B cell development. RAG1^NX thymuses demonstrate a significant reduction in total cellularity, accounted for by a ∼3-fold reduction in the number of double-positive (DP) (CD4⁺CD8⁺) and mature single-positive (SP) cells (Fig. 2A, 2B). The major developmental block occurs at the DN3 to DN4 transition (Fig. 2C, 2D). We find an analogous block in B cell development, at the pro-B to pre-B transition (Fig. 2E, 2F). These phenotypes are inherited in an autosomal recessive manner, as RAG1^NX heterozygotes are phenotypically indistinguishable from homozygous Rag1^WT mice.

The DN3-DN4 transition for developing T cells and pro-B to pre-B transition for B cells require the productive rearrangement of the TCRβ allele or IgH chain for T and B cells, respectively, and subsequent signaling through either the pre-TCR or pre-BCR (14). Blocks at these developmental stages can be due to defective recombination efficiency; therefore, we sought to determine whether RAG1^NX cells had decreased VDJ recombination.

To examine recombination efficiency and Vβ repertoire in RAG1^NX mice, we sorted DN3 thymocytes and assayed Vβ-DβJβ rearrangements by TaqMan PCR. We find substantially lower-than-normal levels of rearrangements of all Vβ gene segments in RAG1^NX DN3 cells (Fig. 3A, 3B). We also determined the Vβ repertoire of thymocytes that produce a functional TCRβ by surface staining and show that there is an altered repertoire at both DP and SP stages (Fig. 3C, 3D, Supplemental Fig. 1A, 1B). Finally, by analyzing the ratio of Vβ usage at the DP versus the SP stage, we show that there is altered selection of certain Vβ proteins, such as less selection of Vβ4 and greater selection of Vβ5 (Fig. 2E). Correspondingly, Vβ usage by bulk thymocytes is altered as evidenced by TaqMan PCR of Vβ-DβJβ rearrangements on total thymocytes (Supplemental Fig. 1C, 1D).

RAG1^NX mice have an ∼50% reduction in peripheral T and B cell counts. This phenotype is inherited in an autosomal recessive manner, as RAG1^NX heterozygotes are phenotypically indistinguishable from homozygous Rag1^WT mice (Fig. 4A). There is no effect on dendritic cells, neutrophils, and inflammatory monocytes, but there is a slight increase in the numbers of splenic NK cells (Fig. 4A). RAG1^NX mice had higher frequencies of T regulatory cells (Tregs) (Fig. 4B) and memory phenotype CD4⁺ and CD8⁺ T cells (Fig. 4C, 4D). The total numbers of Tregs and memory phenotype cells were equivalent across genotypes (Fig. 4B, 4D). This phenotype is consistent with the leaky-SCID/OS phenotype observed in humans and other mouse models of OS (3, 15–17).

Given the similarity of the Q60X mutation present in RAG1^NX mice to lesions found in a number of OS patients, we assessed phenotypes typically associated with development of autoimmunity (5). Ex vivo–stimulated CD4⁺ T cells from young (4- to 6-wk-old) RAG1^NX mice show a heightened propensity to express IFN-γ and IL-17A, and CD8⁺ T cells express more IFN-γ and TNF-α than WT counterparts (Fig. 5A, 5D). OS patients often present with erythroderma, eosinophilia, high circulating IgE levels, and Th2 skewing, including T cell IL-4 and IL-5 production. Although it is clear that young RAG1^NX mice preferentially adopt a Th1-skewed T cell phenotype, more characteristic of the C57BL/6 strain (18), we were interested to see whether canonical OS phenotypes arose as the mice aged. RAG1^NX mice aged 20–22 mo have no difference in immune cell populations, and T and B cell counts have equilibrated in mice by this age (Fig. 5E). There is no evidence of increased eosinophils. Overall RAG1^NX mice have no overt autoimmune pathology (data not shown) but feature increased CD11b⁺CD11c⁺ age-related B cells (ABCs) (Fig. 5F). ABCs are often associated with a predisposition toward ANA-positive autoimmune diseases (19). Accordingly, aged RAG1^NX mice have higher ANA levels (Fig. 5G, Supplemental Fig. 2). Overall, these data suggest that although RAG1^NX mice do not have overt OS disease as seen in humans, they may have a predisposition toward autoimmunity.

Given the position of the RAG1^NX nonsense mutation (Fig. 1A) and the presence of smaller-than-normal RAG1 proteins in RAG1^NX cells (Fig. 1C), we were intrigued as to how these smaller proteins form. The description of OS patients with 5′ RAG1 frameshift mutations hypothesized that alternate N-truncated isoforms arise via translation from alternative TIS (5). Although this report showed that N-terminally truncated RAG1 proteins function in recombination assays, whether alternative translation initiation creates these isoforms was not directly assessed. Interestingly, even RAG1^WT thymocytes have expression of smaller proteins (Fig. 1C). These proteins have been noted and designated previously as “break-down products” (20). We hypothesized that these smaller proteins are naturally occurring isoforms created from translation of alternative TIS. To test this, we created RAG1 proteins epitope tagged at N-terminal (FLAG) and C-terminal (HA) ends. Probing for FLAG only detects full-length RAG1 translated from the canonical TIS, whereas probing for HA would reveal all possible internally translated isoforms (Fig. 6A). Site-directed mutagenesis was performed to introduce the Q60X mutation and to mutate the canonical initiating methionine (iM) codon, AUG, to isoleucine (AUC, iMI). RAG1-WT, RAG1-Q60X, and RAG1-iMI constructs were transfected into 293T cells, and FLAG and HA expression was analyzed. FLAG expression is only detected in the WT-transfected cells (Fig. 6B), whereas the C-terminal HA tag is detected in all RAG1-transfected cells (Fig. 6C). The full-length band at ∼120 kDa is detected by the FLAG tag in WT by Western blot, whereas Q60X and iM1I mutants are absent (Fig. 6D). Analysis of the C-terminal HA tag reveals a number of proteins in all cases. The Q60X and iM1I mutant RAG1 constructs could only produce the smaller isoforms (Fig. 6E). To interrogate whether the smaller bands were indeed produced by internal translation, the first five internal methionines (AUG) were mutated to isoleucines (AUC) in a sequential and additive manner (i.e., M3I = M3I + M2I + M1I). As before, RAG1 constructs were transfected into 293T cells and only the full-length, ∼120-kDa RAG1 isoform is detected via FLAG probing (Fig. 6F). As hypothesized, the potential TIS at M1, M4, and M5 are required for the expression of their respective smaller isoforms as observed by the loss of specific bands upon AUG to AUC mutation (Fig. 6G). Whether the M2 TIS gives rise to the band at ∼80 kDa is inconclusive, as M2I mutation shows a slight decrease in protein quantity at the putative M2 band but not complete ablation as with M1, M4, and M5. It is clear that mutation of the TIS at M3 does not impact expression of the putative M3 isoform. We also interrogated the strength of the Kozak sequences surrounding the putative TIS sites using a published equation (21) and term these “Noderer scores.” We show that M1, M2, M4, and M5 all have extremely strong Kozak sequences (Fig. 4H). Comparison of the Noderer scores (21) between Kozak sequences surrounding the putative TIS sites, and sequences surrounding random, out-of-frame AUGs within the RAG1 transcript, highlights that these high scores are unlikely to be random and may be selected for (Fig. 4I). Evolutionary selection for these alternative TIS sites is supported by conservation of strong internal TIS across species (Supplemental Fig. 3). Given the degenerate nature of codon specificity, it may not be expected that these Kozak sequences would be selected on the basis of the full-length protein alone, as synonymous mutations would not impact this, yet they may impact the natural formation of N-truncated RAG1 isoforms. Overall, these data highlight that RAG1 can be translated from internal TIS and that these N-truncated RAG1 isoforms may have been evolutionarily selected.

In this article, we describe a mechanism for the formation of alternative RAG1 isoforms. These RAG1 isoforms arise as a result of internal translation initiation and lack N-terminal regions of RAG1. They were revealed upon identification of a novel (to our knowledge) nonsense mutation in RAG1, a genetic lesion that is very similar to that present in a number of OS patients. We therefore present a novel (to our knowledge) murine model of OS-like disease that describes mechanistically how functional RAG1 protein can be made despite early nonsense or frameshift mutations and a more-detailed description of T and B cell development in mice with such mutations. This provides a model to further interrogate how autoimmunity manifests concurrently with immunodeficiency and the effect that RAG1 protein or defective recombination has to play in this.

The Q60X nonsense mutation in RAG1 would upon initial observation be expected to result in a nonfunctional peptide that could not participate in VDJ recombination, as it should lack the catalytic domain of RAG1 (the RAG1^core domain). However, some internal translation initiation products provide N-terminal truncated RAG1 proteins that contain the RAG1^core domain. In fact, many of the phenotypes displayed by the RAG1^NX mouse replicate that of the RAG1^core mouse, including blocks in development at Ag rearrangement stages, likely because of inefficient recombination activity (6, 22). Interestingly, altered selection of Vβ proteins from the DP to SP stage is also seen in the RAG1^core mice (6). This altered Vβ usage is independent of choice of TCRβ rearrangement but may be due to altered TCRα repertoire or reduced secondary Vα-to-Jα rearrangements. Thus, both in terms of lymphocyte numbers and VDJ rearrangement, the RAG1^NX mouse phenocopies RAG1^core, consistent with the RAG1^NX mutation unexpectedly generating smaller N-truncated isoforms.

As a result of decreased recombination efficiency, young RAG1^NX mice expectedly display peripheral B and T cell lymphopenia. The increased frequencies of memory phenotype T cells are likely a result of homeostatic expansion of T cells within lymphopenic hosts (23) and increased Treg frequencies because Tregs outcompete conventional T cells (24) or convert from conventional T cells in a lymphopenic environment (25). Treg and T memory populations are equivalent in number in young mice, and this homeostatic expansion is sufficient to fill the entire T/B lymphocyte niche upon aging of the mice. The role of lymphopenia in driving autoimmune disease has been postulated a number of times, yet it is difficult to tease apart the relative contributions of lymphopenia from altered lymphocyte development, defective VDJ rearrangement, and altered Ag receptor repertoires. It is, however, attractive to speculate that homeostatic expansion of few T cells drives oligoclonality and that those T cells that escape development with higher self-reactivity can expand disproportionally.

Although the RAG1^NX mouse has a similar genetic lesion to humans with OS, and similar lymphopenia, the pathology experienced by OS humans is not entirely replicated. However, there is some evidence of immune dysregulation in that RAG1^NX T cells can more efficiently make effector cytokines ex vivo, and aged RAG1^NX mice have an increase in age-associated B cells and increased ANAs. It is also important to note that, in comparing the RAG1^NX mouse to two other murine models of OS, the MM mouse (15) and the R229Q RAG2 mouse (16), we see a number of similarities and differences. All three models display a block in development of B and T cells at Ag receptor gene rearrangement steps, increases in memory phenotype T cells, and altered TCR repertoires. The MM and R229Q RAG2 mice show some facets of canonical OS, such as erythroderma and high IgE, but differ substantially in other aspects, suggesting the specific genetic lesion, mouse strain (which differs between all three mice), and environment in which they are housed all may contribute to the observed autoimmune phenotypes. Of note, the C57BL/6 strain of mice that our RAG1^NX mutation is present in is notoriously resistant to autoimmunity in the absence of additional environmental stimuli (26). Similarly, the R229Q RAG2 mutation on the C57BL/6 background does not present with overt autoimmunity (27).

The most important finding described in this report is that a 5′ nonsense mutation does not lead to loss of functional RAG1 protein. This raises a number of important conceptual points. First, that RAG1 has many naturally occurring, N-truncated isoforms is of considerable interest. RAG1 has been studied for more than 30 y as a single protein, yet it is intriguing to consider that naturally occurring smaller RAG1 isoforms may have important roles in normal RAG function. Already we know that the N terminus of RAG1 has important functions in aiding recombination activity. For example, the RAG1 N terminus contains an E3 ubiquitin ligase that modifies histones to enhance RAG cleavage (7, 10) and regions that bind the VprBP kinase to help repair RAG DSBs (8). Translation from M1 would retain the RAG1 RING domain but may ablate VprBP binding, whereas translation from M2 would likely ablate RING domain activity. Translation from M4 and M5 would result in loss of portions of the RAG1^core domain and will likely have no recombination activity. Whether these internally translated proteins are present at biologically significant levels is of considerable interest because smaller RAG1 isoforms could form part of the RAG1_2/RAG2₂ (RAG) heterotetrametric complex and be important for the regulation of normal function.

Second, that alternative translation initiation may be a natural phenomenon occurring widely within eukaryotic cells is still an underappreciated occurrence and may have many important consequences. Ribosome profiling in eukaryotic cells implies internal translation initiation generates N-truncated isoforms for ∼15% of the ∼20,000 human proteins (28–30), yet our understanding of mechanisms that control internal translation initiation and how resulting N-truncated isoforms function normally and contribute to disease are at infancy. Identification of alternative translation of RAG1 contributes to the expanding database of alternatively translated proteins, and the exact mechanism of translation of RAG1 alternative isoforms is a subject of future investigation.

Another major consequence of internal translation initiation is that early nonsense mutations or frameshift mutations leading to early translation termination may not always result in the absence of a protein. Indeed, it is attractive to speculate that internal TIS may exist as a mechanism by which to protect from deleterious nonsense or frameshift mutations, as is the case with the RAG1^NX mice that do not have the expected SCID phenotype. This is important to consider during analysis of clinical exome/genome sequencing as interpretations could vary drastically. For example, N-terminal truncations of certain proteins may delete autoregulatory domains and result in constitutively active proteins, rather than complete absence. Finally, it is an important factor to consider when designing conditional knockout mice using the cre-lox system, in which typically exons are flanked by loxP sites and excised following cre expression, leading to frameshifted proteins. Although the assumption is that this frameshifted mRNA will be degraded by nonsense-mediated decay, it is possible that translation initiation downstream of the excised region may result in protein expression. Continued investigation of internal translation initiation using RAG1 as a model system will provide a means to understand all of these possibilities.

We thank David Schatz (Yale University, New Haven, CT) for providing the RAG1 Ab and Philip Zoltick (Children’s Hospital of Philadelphia) for the pZHK expression vector.

The authors have no financial conflicts of interest.