Defective Artemis nuclease is characterized by coding joints with microhomology in long palindromic-nucleotide stretches

In vivo D_H-J_H junctions in Artemis-deficient SCID patients

As expected, the frequency of D_H-J_H junctions in bone marrow mononuclear cells (BMMC) of Artemis-deficient patients was lower than in healthy donors. Therefore, a second round of PCR was performed to obtain sufficient quantities of PCR products for further analysis. The presence of D_H-J_H junctions in BMMC of all patients, including the patients with large genomic deletions, indicates that D_H-J_H junctions can be formed even in the complete absence of Artemis protein. Therefore, residual activity of a mutated Artemis protein cannot be the explanation for these junctions.

The majority (90%) of sequences of the D_H-J_H junctions of the Artemis-deficient patients showed long stretches of P-nucleotides (Table 1). The average length of the P-nucleotide stretches was 6.5 nucleotides per junction, which is significantly more than the 0.4 nucleotides in healthy controls (Table 2). Long P-nucleotide stretches have been identified previously in Artemis-deficient mice and in an in vitro V(D)J assay 23, 25. Long stretches of P-nucleotides result from aberrant hairpin opening. In Artemis-proficient cells, the coding end hairpins are opened preferentially near the tip 7, 26. This results theoretically in the presence of a few P-nucleotides, but exonuclease activity often deletes these P-nucleotides. In the absence of Artemis, hairpin opening apparently occurs more downstream of the tip, giving rise to much longer P-nucleotide stretches.

A second characteristic of D_H-J_H junctions of Artemis-deficient patients was a high frequency of microhomology in the P-nucleotide regions. In 25% of the analyzed junctions, two to six P-nucleotides could be derived either from the D_H segment or from the J_H segment. These junctions most probably occur via annealing of the long 3′overhangs that are generated through aberrant hairpin opening (Fig. 1). Apparently, microhomology of P-nucleotide stretches plays an important role in D_H-J_H junction formation in Artemis-deficient patients, presumably by facilitating ligation of two ends with long (P-nucleotide) overhangs.

V(D)J recombination assay

We developed a V(D)J recombination assay to further investigate the role of hairpin opening and P-nucleotide microhomology use in coding joint formation. The V(D)J recombination substrate contains two RSS in inversion orientation that were flanked by a four-base pairs direct repeat (Fig. 2A). This recombination substrate was cotransfected with RAG1 and RAG2 expression constructs into wild-type or Artemis-deficient fibroblasts. The excised DNA fragment is inverted, resulting in a signal joint and a coding joint in the same plasmid. The coding joints can be amplified by a nested PCR approach (Fig. 2A). Depending on the way of hairpin opening and coding joint formation, three different types of coding joints can be generated and discriminated from each other by restriction enzyme digestion (Fig. 2B). Coding joints with long stretches of homologous P-nucleotides give an NgoMI restriction site, while a four-base pair microhomology of the coding ends creates a NotI restriction site. Junctions that have been processed differently will not contain either of the two restriction sites.

Coding joints in C5RO wild type were of diverse nature: of these junctions, approximately 30% were NgoMI sensitive, approximately 30% were NotI sensitive and the rest had a different composition (Fig. 2C). Artemis-deficient fibroblasts (including fibroblasts of Artemis-8) yielded only NgoMI-sensitive coding joints with microhomologous P-nucleotide stretches, consistent with the high frequency of microhomology of P nucleotides in BMMC-derived D_H-J_H junctions. Cotransfection of a wild-type Artemis expression construct restored the normal distribution of junctions.

This assay can be used for easy discrimination between fibroblasts derived from Artemis-deficient SCID patients and LIG4-deficient SCID patients, because the latter did not show the characteristic shifted pattern (data not shown).

Validation of the V(D)J recombination assay for Artemis activity studies

To evaluate whether our assay can be used to test Artemis mutants, which can be either generated via site-directed mutagenesis or identified in a patient and subsequently cloned into an expression vector, we performed complementation experiments with Artemis mutants that have been described in the V(D)J recombination literature 23, 24. These two independent studies have shown that eight conserved amino acids are indispensable for the Artemis endonucleolytic activity. In addition, they showed that the C terminus could be deleted without losing V(D)J recombination activity, whereas small N-terminal deletions are already deleterious for Artemis function. We therefore generated four N-terminal deletion mutants (NΔ7, NΔ16, NΔ45, and NΔ66), four C-terminal mutants (CΔ80, CΔ135, CΔ232, and CΔ331), and four point mutants (D17A, H33A/H35A, H115A, and H319A) and tested whether these mutants could restore the shifted junction pattern of 100% NgoMI-sensitive to a normal distribution (Fig. 3A, B).

In line with published data, the N-terminal deletion or point mutants did not show complementation, indicating that the N-terminal region is indispensable for V(D)J recombination. All C-terminal deletion mutants complemented the Artemis defect. The C-terminal mutant with the largest deletion (CΔ331), however, did not show full complementation (Fig. 3A), probably because a small part of the β-CASP domain was lost. Based on these results, we conclude that our V(D)J recombination assay can reliably be used for Artemis activity studies.

Further mapping of the Artemis-active site

Artemis is assigned to the metallo-β-lactamase superfamily, based on sequence homology. Therefore, the mechanism of this structure-specific endonuclease has been proposed to function similarly. In this model, H33, H35, D37, and H38 form the conserved metallo-β-lactamase HxHxDH signature. H115 and H319 have been proposed to coordinate two active site metals, while D17, D136 and/or D165 may form salt bridges to the HxHxDH motif and stabilize its conformation. However, alternative models may also be considered to explain some observations that cannot be explained easily by this model. First, D136 and D165 have not been found in the sequences of available metallo-β-lactamase crystal structures. Furthermore, metallo-β-lactamases coordinate Zn²⁺ ions in the active site, whereas biochemical experiments show that Mg²⁺ ions are required for activity 23.

Therefore, we considered another possible nuclease mechanism. Retroviral integrases and several transposases coordinate Mg²⁺ in the active site for DNA cleavage using an array of three acidic amino acids, i.e. two aspartates (D) and one glutamate (E) residue, referred to as a DDE motif 27. The acidic amino acids in this motif cannot be exchanged for any other amino acid residue without loss of activity. Even the most conservative mutations (D to E or E to D) result in inactive proteins. We reasoned that changing D to E or E to D would inactivate the Artemis protein if an array of acidic amino acids is required for divalent metal ion coordination, while a metallo-β-lactamase active site would probably not be disrupted (completely) by such a conservative substitution.

Therefore, we generated five Artemis mutants in the D and E amino acids that have been conserved between Artemis and Snm1 (D17E, D37E, D136E, D165E, and E213D). Except for E213D, cotransfection of these Artemis mutants did not result in complementation of the Artemis defect (Fig. 3A, B), suggesting that they may indeed be involved in metal ion coordination. The importance of D37 is supported by the finding of a mutation in this amino acid in SCID Artemis-9, i.e. p.D37G, also results in radiosensitive SCID. We therefore propose in an alternative model that the Artemis-active site might contain an array of conserved acidic amino acids that coordinate Mg²⁺ ions. This motif might be stabilized by the metallo-β-lactamase fold, in which the histidines could aid in coordination of the acidic residues and/or accept a proton from the water molecule that acts as the nucleophile that cleaves the phosphodiester bond in the DNA.

SCID patient with a mutation in the Artemis C terminus

One of the SCID patients, Artemis-8, had a mutation affecting the C-terminal region of the Artemis protein resulting in a frame-shift and a premature stop codon (p.G464AfsX18). This seemingly contradicts the observation that C-terminal deletion mutants complement the Artemis defect. We hypothesized that the expression level of the C-terminally mutated Artemis protein in patient cells is insufficient to carry out non-homologous end joining (NHEJ) and V(D)J recombination resulting in radiosensitive SCID. Therefore, we cloned this mutation in the pDVG90 expression vector and tested this Artemis-8 mutant in our V(D)J recombination assay. The Artemis-8 mutant was indeed able to complement the Artemis defect in Artemis-1 fibroblasts and even in Artemis-8 fibroblasts (Fig. 3C), showing that complementation was caused by overexpression. Therefore, the results of activity assays with transient overexpression of mutant Artemis protein should be interpreted with caution, since disease-causing Artemis mutations with residual activity might be masked by the transient overexpression.

Patient samples

Five radiosensitive (RS) SCID patients were studied. Informed consent for patient material was obtained according to the guidelines of the medical ethics committees of Hacettepe University Children's Hospital and Erasmus MC. All patients were diagnosed with SCID within the first 6 months of life based on failure to thrive, recurrent infections and absence of B and T cells. The previously described Artemis-2 patient carries a homozygous G118V mutation 16, 28, Artemis-4 and 5 had a homozygous deletion of exons 1 and 2. Artemis-8 has a homozygous deletion of five nucleotides (c.1391_1395delGAATC) resulting in a frame shift and a premature stop codon (p.G464AfsX18) and Artemis-9 had a homozygous missense mutation (c.A110G; p.D37G).

Analysis of D_H-J_H junctions in Artemis-deficient SCID patients

DNA was isolated from BMMC of four Artemis-deficient SCID patients and healthy donors. From the fifth Artemis-deficient patient (Artemis-9) no BMMC were available. Analysis of D_H-J_H gene rearrangements was performed as previously described 20. Due to the low frequency of D_H-J_H junctions, a second PCR was performed to generate sufficient amounts of PCR product for cloning and sequencing.

V(D)J recombination assay

C5RO wild type or Artemis-1 fibroblasts containing a homozygous deletion of exons 10–12 16 were transfected with a recombination substrate (pGG49.93), RAG1 and RAG2 expression constructs and one of the Artemis mutant constructs 16, 29. The pGG49.93 was made by cloning a 1.4-kb AflIII (blunted)-BamHI fragment of pDvG93 into the SpeI (blunted)-BamHI digested vector pGG49 16, 29, 30. After 48 h, the fibroblasts were harvested and extrachromosomal DNA was isolated 29. To amplify the coding joints, a nested PCR of 2 × 25 cycles was performed. The first round of PCR was performed with oligonucleotides NV05F and DG147. Subsequently, 1 µL of this PCR reaction was used as a template for the second round of PCR using oligonucleotides DG89 and radioactively labeled FM30 31. PCR products were digested with NotI or NgoMI. After electrophoresis, the gel was dried and PCR products were visualized by phosphorimaging.

Artemis mutant expression constructs

For in vitro PCR-based mutagenesis, the Artemis wild-type construct pDVG190 (based on pCDM8, Invitrogen) was used 16. Several deletion and point mutants were generated. Details are available upon request. The protein expression levels of the Artemis-mutants were checked by Western blot analysis of transiently transfected COS cells.