Ku Heterodimer-Independent End Joining in Trypanosoma brucei Cell Extracts Relies upon Sequence Microhomology

T. brucei growth and cell extract preparation.T. brucei procyclic-form cells of strain EATRO795 were grown in SDM-79 medium at 27°C. Approximately 1 liter of cells at densities between 0.8 × 10⁷ and 2.0 × 10⁷ cells·ml⁻¹ was used to prepare a nuclear or cell extract. Bloodstream-stage cells of strain Lister 427 or ILTat1.2 were grown in HMI-9 medium at 37°C and used to infect adult female ICR mice (Harlan, United Kingdom) that had been immunocompromised with cyclophosphamide (250 mg·kg of body weight⁻¹). At a parasitemia of 0.5 × 10⁹ to 1.0 × 10⁹ cells·ml⁻¹, the mice were sacrificed, and 0.5 ml of blood was used to infect adult female Wistar rats (immunocompromised as described above; Harlan, United Kingdom). When the parasitemia here reached 0.5 × 10⁹ to 1.0 × 10⁹ cells·ml⁻¹, ∼10 ml of blood was withdrawn by cardiac puncture into Carter's balanced salt solution containing 5% sodium citrate anticoagulant, and T. brucei was then purified by DE52 anion-exchange chromatography. Nuclear extracts were prepared in glycerol-containing buffers as described previously (5). Whole-cell extracts were obtained by the procedure of Laufer et al. (48), using sucrose-containing buffers. RAD51 and KU70 homozygous mutants were generated in procyclic-form T. brucei EATRO795 by using constructs and transformation conditions described previously (19, 56).

End joining.Standard reactions proceeded for 10 min at 37°C and used 10 to 20 μg of cell extract and 200 to 500 ng of DNA in a total volume of 100 μl containing 50 mM Tris·HCl (pH 7.5), 20 mM potassium acetate, 3 mM magnesium acetate, 1 mM ATP, 1 mM dithiothreitol, and 100 mg·ml⁻¹ bovine serum albumin. A 5-min preincubation of the extract preceded addition of the DNA. For reactions involving ATP regeneration, 10 mM creatine phosphate and 20 μg·ml⁻¹ creatine kinase were included (both from Roche). To deplete ATP, 10 U of apyrase (New England Biolabs) was added to each reaction mixture, and the extract was incubated for 10 min at 37°C prior to substrate addition. Reaction products were prepared for analysis by phenol-chloroform extraction and ethanol precipitation and were normally examined by Southern blotting and hybridization with [α-³²P]-labeled substrate DNA. For experiments using whole-cell extracts, the products were treated with 0.2 mg·ml⁻¹ RNAse A (Sigma) for 2 min prior to phenol-chloroform extraction. Hybridization was visualized using a PhosphorImager (Typhoon 8610; Molecular Dynamics) and quantified by densitometric analysis using ImageQuant.

pBluescript (Stratagene) and other plasmids were prepared for end joining by digestion of ∼20 μg of DNA with the appropriate restriction enzyme (see below) at 37°C for 2 h and were then purified by phenol-chloroform extraction and ethanol precipitation; the concentration and extent of digestion were analyzed by agarose gel electrophoresis prior to end joining. PCR products to assay end joining were amplified from the NEO or HYG gene in T. brucei strain 3174 (56) genomic DNA by using Taq polymerase (ABgene) and oligonucleotide primers that had been gel purified; after PCR, the products were purified by phenol-chloroform extraction and ethanol precipitation or by spin purification (QIAGEN) and were assessed for specificity and quantity by agarose gel electrophoresis. Joints were analyzed by purifying the end-joining reaction products as described above, separating half the reaction product by agarose gel electrophoresis, and performing PCR amplification on the other half with Taq polymerase. The resultant PCR products were TOPO cloned (Invitrogen), multiple clones were isolated, and products of representative sizes were sequenced. Control reactions were performed with the HYG joint primers on the end-joined NEO substrate, with NEO joint primers on the end-joined HYG substrate, and with both primer pairs on substrates that had not been incubated in the nuclear extract; the reaction products were analyzed by gel electrophoresis and were TOPO cloned and sequenced, showing either no amplification or small amounts of a single-primer product. All oligonucleotide primers used for this work can be provided on request.

Bioinformatic analysis of NHEJ factors.To search for homologues of the core NHEJ components in the kinetoplastids, polypeptide sequences for Ku (the 70 and 80 subunits), DNA ligase (I, III, or IV), or XRCC4/Lif4 from a number of species were identified from annotations within the NCBI database and used initially to search the T. brucei, Trypanosoma cruzi, and Leishmania major genome project databases by BlastP (default settings). Iterative searches were also used to attempt to identify remote homologues. To ensure that the searches were sensitive, and to attempt to exclude false positives, bespoke hmmer profiles were created from multiple sequence alignments of the different protein sequences by the hmmbuild and hmmcalibrate programs, and conceptual translations from individual genome projects were analyzed using the hmmsearch program. The NR database was mined for homologues with the bespoke profiles using the Kyoto University Bioinformatics Center MOTIF service (http://motif.genome.jp), demonstrating the validity of the profiles. All potential hits were checked manually against known sequence motifs and domain structures.

For phylogenetic analyses, multiple alignments of the different protein sequences were performed using each of the following methods: ClustalW (default parameters), Hmmalign (displaying match states to the profile only), Matchbox (default parameters) (25), T-coffee (default parameters) (59), and Fugue (68) alignment against a pdb structural file or the Homstrad profile appropriate to the case. Where subsequences of known domains were required, sequences were aligned against the appropriate profile from the Pfam database using the hmmalign program. The leading and trailing portions of the alignment outside the match to the profile were then removed and the sequences converted to fasta format for subsequent analysis. The alignments made using the different methods were then combined by the T-coffee COMBINE function. For the ATP-dependent DNA ligases, the consensus alignment was written in phylip format, and a phylogenetic tree was generated by the Phylip package utilizing protdist and fitch. The significance of the different DNA ligase groupings was determined by bootstrap analysis of 100 replicates, and each had a >90% probability (data not shown).

T. brucei nuclear extracts catalyze efficient joining of linear DNA molecules.To look for end-joining activity in T. brucei, we examined the ability of nuclear extracts to catalyze the joining of linear DNA molecules. First, ScaI-digested pBluescript DNA was incubated with an extract from ILTat1.2 bloodstream-stage cells (Fig. 1A). DNA joining was detected readily, resulting in dimer- and trimer-sized linear molecules, as well as higher forms that appear to result from the substrate joining to genomic DNA (the molecules are >12 kb). All the reaction products were joined by covalent linkages, as evidenced by the fact that their electrophoretic mobilities were unaffected by heating of the reaction products to 90°C (data not shown). To examine the efficiency of the reactions, time courses were then performed (an example is shown in Fig. 1B), revealing that DNA joining was rapid, with dimer-sized products appearing after 30 s. Quantification of a number of independent experiments using distinct plasmids (Fig. 1C) showed that the dimer product became most abundant after 5 to 10 min and diminished thereafter. Most likely, this is due to nuclease digestion by the extract, since the decrease in the amount of the substrate over time (only around 50% remained after ∼20 min) was greater than could be explained by conversion to joined products. In all DNA-joining reactions, using nuclear extracts and plasmids or smaller PCR products (see below) as substrates, we observed intermolecular joining; only in whole-cell extracts (see below) did we see any evidence for intramolecular joining.

• Open in new tab
• Download powerpoint

FIG. 1.

End joining of linear plasmid DNA in T. brucei nuclear extracts. (A) DNA incubated for 10 min at 37°C with a nuclear extract from bloodstream-stage cells (+), compared with a control without the extract (−). The substrate monomer (M) and dimer (D)- and trimer (T)-sized molecules are indicated. (B) Time course of end joining. (C) Quantification of the DNA substrate and the dimer-sized product over time, relative to the abundance of each species at the earliest time point (30 s, arbitrarily assigned a value of 1.0). Plotted values are averages from three experiments. Error bars, standard deviations.

To determine if end joining is strain or life cycle stage specific, the reactions in nuclear extracts from Lister 427 bloodstream-stage cells and from EATRO795 procyclic-form cells were compared. Equal efficiency was found in each (data not shown). We next attempted to optimize the reaction conditions (data not shown). A linear increase in the quantity of the reaction product was seen as the amount of the extract was increased over the range 5 to 100 μg. The reaction occurred with optimal efficiency between 35 and 37°C (becoming less efficient at higher and lower temperatures over the range 25 to 45°C), and maximum joining was found between 20 and 40 mM potassium diacetate and was undetectable above 100 mM.

To assess whether ATP and magnesium are required for DNA joining, a 798-bp DNA fragment was PCR amplified using 5′-phosphorylated oligonucleotide primers. DNA joining was not observed if Mg²⁺ was omitted from the reaction mixture (Fig. 2A). Moreover, joining of plasmid DNA occurred efficiently up to 3 mM Mg²⁺ but became less efficient at higher concentrations (Fig. 2B). Since the substrate appeared to be degraded more rapidly at higher Mg²⁺ concentrations, the reduced end joining may be due to increased nucleolytic digestion. Surprisingly, end joining was equally efficient in nuclear extracts whether exogenous ATP (1 mM) was added or omitted (Fig. 2A). Although we cannot rule out the possibility that sufficient ATP for the reaction is present in the nuclear extract, this seems unlikely; T. brucei contains 22 nmol of ATP·mg of protein⁻¹ (21), meaning that maximally 4.4 μM ATP could be present in an unsupplemented reaction mixture using 20 μg of extract (assuming, erroneously, that ATP is not depleted during extract preparation). In fact, T. brucei contains significant glycerol kinase activity (39), which would deplete the nuclear extract of ATP due to the use of glycerol in the preparation buffers (A. Gunzl, personal communication). We therefore generated whole-cell extracts that were prepared in a sucrose-containing buffer (and should retain ATP) and compared joining on a linear pBluescript plasmid with or without added ATP (1 mM) or with addition of an ATP regeneration system (Fig. 3A). Overall, end joining was slightly less efficient in whole-cell extracts than in nuclear extracts. In addition, the product profile in whole-cell extracts was more variable and somewhat distinct from that in nuclear extracts: although dimer-sized products were still detectable, other molecules, both larger and smaller than the substrate, were generated in some reactions. It is possible that these products result from intramolecular joining, or potentially from truncation of the substrate, but we have not verified this, because such products were not generated uniformly in every experiment (e.g., compare Fig. 3A and B). Nevertheless, exogenous ATP did not increase end joining substantially, as determined by the amounts of the dimer product generated over time. Addition of an ATP regeneration system also did not increase the amount of the product, nor did it increase the longevity of the reaction. In fact, the substrate was depleted more rapidly under these conditions. To examine this further, we asked if depletion of putative ATP in the extract by the addition of apyrase would affect the reaction. Again, we found no evidence that the end-joining activity was reduced by this treatment (Fig. 3B).

• Open in new tab
• Download powerpoint

FIG. 2.

Magnesium and ATP requirements of T. brucei end joining. (A) A 798-bp HYG PCR product was incubated with (+) or without (−) a T. brucei nuclear extract in a reaction buffer containing 3 mM magnesium acetate and 1 mM ATP (lane 1) and compared with equivalent reactions in which ATP (lane 3) or magnesium acetate (lane 5) was omitted. (B) Time courses of linear plasmid end joining in increasing concentrations of magnesium acetate [Mg(OAc)₂]. The substrate monomer (M) and dimer (D)- or trimer (T)-sized molecules are indicated.

• Open in new tab
• Download powerpoint

FIG. 3.

Assessment of the ATP requirement of end joining in T. brucei whole-cell extracts. (A) EcoRI-digested pBluescript plasmid DNA is shown incubated in a T. brucei whole-cell extract in a reaction buffer without added ATP (ATP−) or containing either 1 mM ATP (ATP+) or 1 mM ATP plus an ATP regeneration system (ATP+, ATP reg). Samples of the reaction products taken during a time course are shown. (Note that the 15-, 30-, and 45-min samples for the ATP+, ATP reg time course are shown somewhat overexposed to compensate for substrate degradation.) (B) Time courses of linear plasmid end joining in a whole-cell extract using a buffer without added ATP, where the extract had either been preincubated with apyrase for 10 min prior to addition of the substrate (APY+) or, as a control, left untreated (APY−). The substrate monomer (M) and dimer (D)-sized molecules are indicated, as are the positions of other putative products.

T. brucei DNA joining is mediated by microhomology.To determine if the end-joining activity occurs precisely, as has been described for error-free Ku-dependent NHEJ reactions (4), or involves sequence changes indicative of error-prone Ku-dependent NHEJ (12) or Ku-independent joining (50), further substrates were derived by PCR. In one PCR, a 400-bp DNA fragment was amplified with primers containing terminal recognition sites for Acc65I, KpnI, or EcoRV, which create termini with a 5′ overhang, a 3′ overhang, or a blunt end, respectively. The different restriction-digested substrates were incubated with a T. brucei nuclear extract under “standard” conditions (see Materials and Methods). An example of this analysis is shown in Fig. 4A, indicating that the DNA end configuration did not influence the efficiency of joining. To ensure that this finding was not due to inefficient restriction digestion of the PCR products, we compared the end-joining efficiency of plasmid DNA that had been similarly restriction digested, with the same result (data not shown). The lack of correlation between joining efficiency and DNA end configuration in T. brucei contrasts with end joining in mammalian nuclear extracts, where 3′ overhangs are joined most efficiently, 5′ overhangs are joined slightly less efficiently, and blunt ends are the poorest substrates (4, 12). In fact, in T. brucei nuclear extracts, the same PCR product that had not been restriction digested was joined as efficiently as the other three substrates (Fig. 4A). This indicates strongly that the end configuration has little influence on end joining, since such Taq-generated PCR products lack 5′-terminal phosphates and contain 3′-phosphorylated A overhangs, both of which are conditions refractory to ligation and NHEJ (23). To examine this further, a 391-bp substrate was generated by PCR with nonphosphorylated primers, each containing a terminal HindIII site. HindIII-digested or undigested substrates were then incubated with the nuclear extract, the reaction products were deproteinized, and half of each was redigested with HindIII. The majority of the products could not be digested with HindIII (Fig. 4B), showing that T. brucei DNA end joining is predominantly error prone, removing terminal sequences irrespective of whether complementary overhangs are present.

• Open in new tab
• Download powerpoint

FIG. 4.

DNA end requirements of T. brucei end joining. (A) The end-joining efficiency of a 400-bp NEO PCR product digested at each terminus with Acc65I (lanes 3 and 4), KpnI (lanes 5 and 6), or EcoRV (lanes 7 and 8), or left undigested (lanes 1 and 2), was compared following incubation with (+) or without (−) a T. brucei nuclear extract. The form of DNA end is indicated: M, substrate monomer; D, dimer-sized molecule. (B) A 391-bp HYG PCR product, either digested at terminal HindIII sites (lanes 1 and 2) or left undigested (lanes 3 and 4), was incubated in a T. brucei nuclear extract and electrophoresed (uncut). The remainder of the reaction products were purified and digested with HindIII prior to electrophoresis (lanes 5 to 8, corresponding to lanes 1 to 4, respectively).

To determine how DNA joining occurs in T. brucei extracts, a 400-bp 5′-nonphosphorylated substrate (NEO) and a 798-bp 5′-phosphorylated substrate (HYG) were incubated in separate, standard reactions using either nuclear or whole-cell extracts. In each case, half of the reaction product was analyzed by gel electrophoresis to confirm joining (data not shown). To characterize the joins, PCR was performed on the other halves of the reaction products using primers internal to the substrates, which should amplify across the joins of head-tail, head-head, or tail-tail product configurations. For both substrates, several different-sized DNA fragments were PCR amplified (data not shown), and cloning and sequencing revealed that these were derived only from head-tail joins (Fig. 5). Most likely, the other joining configurations occur, but PCR is impeded by the formation of hairpins during the annealing of the PCR products. In control experiments where the substrates had not been incubated with an extract, very limited PCR amplification occurred by fortuitous, single-primer amplifications (data not shown).

• Open in new tab
• Download powerpoint

FIG. 5.

Sequences of joints during T. brucei end joining. DNA sequences of head-tail-ligated products of NEO (N) or HYG (H) PCR substrates following end joining in T. brucei cell extracts were determined following PCR amplification using primers internal to the substrates. For each product, the top sequence depicts the upper (head) stand (5′-to-3′ orientation) of one molecule where it has been ligated to another (tail) molecule (bottom sequence, same orientation); the joints formed between the molecules were, in each case, within regions of sequence homology (underlined) revealed by nucleolytic degradation of the DNA ends (the numbers of bases deleted [Δ] are given). The relative abundance of each product is shown on the right, expressed as the number of clones isolated with a given joint sequence (from a total of 27) and the percentage of the total that this represents.

All intermolecular joins occurred at regions of microhomology within the DNA substrates, irrespective of whether the reaction had been conducted in nuclear or whole-cell extracts. These regions ranged from 6 to 16 bp and virtually always (7 of 8 joins) contained at least 1 mismatched base. Mismatch repair does not appear to occur during joining, however; the sequences of the microhomologous joins were derived from one or the other substrate molecule and were not hybrids (data not shown). In creating the joins, sequence was almost always lost from the ends of the DNA substrates, explaining why terminal restriction sites were not conserved (Fig. 4B) and why substrates without terminal 5′ phosphates can be ligated efficiently (Fig. 4A). Gel electrophoresis of the joining-reaction products suggested that discrete products, compatible with dimerized or trimerized substrates, were created, while the PCR analysis revealed multiple-sized products that often had significant terminal deletions (up to 301 bases). Most likely this dichotomy can be explained as an experimental artifact: joining reactions that delete large regions of sequence may, in fact, be infrequent, but their products are PCR amplified, or cloned, preferentially. Note, however, that even with this putative bias, the most common products for each substrate had lost the smallest amounts of terminal sequence (Fig. 5).

KU70 and RAD51 are dispensable in T. brucei end joining in vitro.T. brucei RAD51 and the Ku heterodimer have been described previously, including the generation and analysis of homozygous mutants (19, 20, 44, 56). To determine if the loss of either of these factors affects DNA end joining, rad51 and ku70 homozygous mutants were generated in procyclic-form cells (data not shown). Nuclear extracts were prepared from each mutant strain and from wild-type cells, and the phosphorylated 798-bp substrate was assayed for joining using standard conditions (Fig. 6). No difference in reaction efficiency was detected in any of the nuclear extracts. This indicates, first, that the use of sequence homology in the reaction does not result from a role for RAD51 or homologous recombination and, second, that the reaction is distinct from Ku-dependent NHEJ.

• Open in new tab
• Download powerpoint

FIG. 6.

T. brucei end joining does not require RAD51 or KU70. The efficiency of end joining in nuclear extracts of wild-type procyclic-form T. brucei was compared to that in extracts from homozygous mutants (−/−) of RAD51 or KU70. Increasing amounts of the extract from each cell line, compared with control reactions where no extract was added (−), were used. The substrate monomer (M) and dimer (D)-sized molecules are indicated.

Core NHEJ components may not be conserved in the Kinetoplastida.Given the absence of a detectable role for Ku in DNA joining in T. brucei nuclear extracts, we undertook a bioinformatic analysis of NHEJ factors in the genomes of three related kinetoplastid parasites: T. brucei, T. cruzi, and Leishmania major. Orthologues of KU70 and KU80, previously identified in T. brucei (19, 44), were detected readily in T. cruzi and L. major (data not shown). In contrast, DNA Lig IV and XRCC4 were not discernible in any of the genomes. Though this may be because both factors are highly divergent in the Kinetoplastida and escaped detection during searches of the genomes, we find this unlikely. In searching for DNA Lig IV, four DNA ligases were apparent in each kinetoplastid genome. One of these is well conserved with other eukaryotes and is most homologous to DNA Lig I (Fig. 7). The other three proteins are substantially divergent from the range of DNA ligases described for eukaryotes but were identified readily as ligases. Two of these correspond to mitochondrion-specific DNA ligases, Lig Kα and Lig Kβ (30), characterized previously in T. brucei and encoded by tandemly linked genes. The fourth DNA ligase groups with Lig Kα and Lig Kβ in phylogenetic analysis, suggesting a common origin (Fig. 7), though it is encoded from an unlinked locus. DNA Lig IV is distinguished from other eukaryotic ATP-dependent DNA ligases by two C-terminal BRCT domains surrounding an XRCC4 interaction domain (29, 69). These motifs are absent from all of the predicted kinetoplastid DNA ligases, and searches of the kinetoplastid genomes using these sequence elements in isolation revealed no other putative DNA ligases. Taking these data together, it is likely that we have identified all the ATP-dependent ligases encoded by these three kinetoplastids, and none of those present are recognizably related to DNA Lig IV. Searches for XRCC4/Lif1 are more problematic, because the primary sequence of this protein is poorly conserved in eukaryotes (29, 69). Perhaps not surprisingly, initial BlastP searches of the kinetoplastid genomes failed to indicate any XRCC4/Lif1 homologues. To examine this further, we generated an hmm profile from the multiple alignment of the DNA Lig IV interacting domain published by Sibanda et al. (69). This confirmed the presence of a conserved motif ([V or L]LN[I or E]KK) that was considered diagnostic for XRCC4 (29, 69). Searches using this hmm profile also failed to identify any protein in the kinetoplastid genomes. Although the NHEJ factors beyond the core machinery are less widely conserved, we also searched the kinetoplastid genomes for each of these proteins (DNA-PKcs, Artemis, NEJ1, XLF), and no detectable homologue was found (data not shown).

• Open in new tab
• Download powerpoint

FIG. 7.

Phylogenetic comparison of eukaryotic ATP-dependent DNA ligases. An unrooted phylogenetic tree, derived from an optimized multiple sequence alignment of the ATP-dependent DNA ligase A-M domain (Pfam01068), is shown. DNA ligases are classified by sequence identity as either Lig I, Lig III, or Lig IV enzymes, found in many eukaryotes, or as a kinetoplast-specific group of DNA replication/repair ligases (Lig K) (the asterisk indicates an inference made in this work). Organisms, abbreviations, and accession numbers for the sequences used are as follows: P. falciparum (Pfa) Lig I (NP_704940); Cryptosporidium parvum (Cpa) Lig I (XP_626929); Homo sapiens (Hsa) Lig I (P18858), Lig III (P49916), and Lig IV (NP_002303); Drosophila melanogaster (Dme) Lig I (AAK93363) and Lig IV (AAF48298); Caenorhabditis elegans (Cel) Lig I (Q27474) and Lig IV (NP_498653); Arabidopsis thaliana (Ath) Lig I (CAA66599) and Lig IV (NP_568851); T. cruzi (Tcr) Lig I (EAN97122, EAN91701), Lig Kα (EAN93728), Lig Kβ (EAN93729, EAN92478), and Lig K* (EAN94758, EAN97945); T. brucei (Tbr) Lig I (AAZ12036), Lig Kα (AAQ72485), Lig Kβ (AAZ12125), and Lig K* (EAN79636); L. major (Lma) Lig I (CAJ06835), Lig Kα (CAJ05226), Lig Kβ (CAJ05229), and Lig K* (CAJ02559); S. cerevisiae (Sce) Lig I (P04819) and Lig IV (NP_014647); Schizosaccharomyces pombe (Spo) Lig I (NP_595345) and Lig IV (CAA21085); Giardia lamblia (Gla) Lig I (XP_768805). For Tetrahymena thermophila (Tth) and Toxoplasma gondii (Tgo), protein sequences were derived from gene predictions made by the genome projects as follows: T. thermophila (www.ciliate.org/) Lig I-a (36.m00193), Lig I-b (154.m00112), and Lig IV (41.m00206); T. gondii (http://www.toxodb.org/toxo/home.jsp) Lig I (TgTwinScan_0903) and Lig IV (TgTwinScan_0375).