Interaction between the helicases genetically linked to Fanconi anemia group J and Bloom's syndrome (original) (raw)

DNA helicases are important players in ensuring genome integrity, with mutations in their genes underlying various chromosomal instability disorders. Direct cross‐talk between human FANCJ and BLM helicases further interconnects the intricate network of genome maintenance pathways.

Introduction

Fanconi anemia (FA) is an autosomal recessive disorder characterized by multiple congenital anomalies, progressive bone marrow failure, and high cancer risk (Levitus et al, 2006; Taniguchi and D'Andrea, 2006; Wang, 2007). Cells from FA patients exhibit spontaneous chromosomal instability and hypersensitivity to DNA interstrand cross‐linking agents. Thirteen causative genes (FANCA/B/C/D1/D2/E/F/G/I/J/L/M/N) have been identified to date. Increasing evidence suggests that these genes encode proteins that function as signal transducers (Andreassen et al, 2004; Pichierri and Rosselli, 2004; Matsuoka et al, 2007; Wang et al, 2007) and/or DNA processing molecules (Cantor et al, 2001; Meetei et al, 2005; Gari et al, 2008) in a DNA‐damage response pathway. The FA proteins interact with a variety of DNA repair factors, and it is becoming increasingly evident that these interactions are critical for a robust DNA‐damage response. For example, the Bloom's syndrome (BS) helicase (BLM) is associated with the FA core complex (Meetei et al, 2003) and BLM mutations lead to a disease characterized by severe growth retardation, immunodeficiency, reduced fertility, and predisposition to cancer (Hanada and Hickson, 2007). In recent work, FANCM was shown to act as a bridging protein to connect the BLM complex and FA core complex through independent domains that mediate physical interactions with RMI1/TopoIIIα and FANCF, respectively (Deans and West, 2009). At the cellular level, the hallmark of BS is a high rate of sister chromatid exchange (SCE) (Chaganti et al, 1974; Ray and German, 1984). An emerging theme is that BLM and its associated factors collaborate with FA proteins to deal with replication stress and maintain chromosomal stability. Moreover, recent evidence suggests that the FA pathway and BLM collaborate during mitosis to prevent micronucleation and chromosome abnormalities such as those at fragile sites where sister chromatid bridging can occur (Chan et al, 2009; Naim and Rosselli, 2009).

One of the more recently identified FA proteins, shown to be defective in the FA complementation group J, is the _B_RCA1‐_a_ssociated _C_‐terminal _h_elicase (BACH1/BRIP; designated here as FANCJ) (Levitus et al, 2005; Levran et al, 2005; Litman et al, 2005). FANCJ was also identified as a protein associated with breast cancer (Cantor et al, 2001; Seal et al, 2006). FANCJ‐deficient cells are sensitive to DNA cross‐linking agents (Litman et al, 2005; Peng et al, 2007), and FANCJ‐depleted cells exhibit mitomycin C (MMC)‐induced chromosomal breakage accompanied by triradial and quadriradial chromosome formation (Bridge et al, 2005; Litman et al, 2005). Like the human counterpart, chicken brip1 (FANCJ) cells are also characterized by chromosomal instability, including an elevated level of SCE (Bridge et al, 2005) that is not seen in most human FA cells. FANCJ has been proposed to function downstream of FANCD2 monoubiquitination (Litman et al, 2005), a critical event in the FA pathway. Evidence supports a role for FANCJ in a homologous recombination (HR) pathway of double‐strand break repair. However, FANCJ is also required for timely progression through S phase (Kumaraswamy and Shiekhattar, 2007), and can unwind G‐quadruplex structures to maintain genomic stability (London et al, 2008; Wu et al, 2008), suggesting that FANCJ has additional roles in response to replication stress that may operate independently of the classic FA pathway (Wu et al, 2009; Hiom, 2010).

Since there is only limited information on the role of FA proteins in the proper DNA‐damage response, investigation of protein interactions between endogenous FANCJ and other DNA repair factors is likely to lead to new insights into the molecular pathogenesis of FA and the role of the FANCJ helicase in reacting to replication stress and participating in HR repair. Given the evidence that both BLM and FANCJ operate to maintain genomic stability in downstream events of HR repair, we investigated the potential interaction of the two helicases. Our studies demonstrate that FANCJ and BLM are associated with each other in vivo, and that the two helicases physically and functionally interact with each other. These results are discussed in light of the cross‐talk between the FA pathway and the BLM protein complex to prevent chromosomal instability associated with cancer and human disease.

Results

Reciprocal co‐immunoprecipitation of FANCJ and BLM helicases from human nuclear extracts

To explore whether endogenous FANCJ and BLM are associated with each other in vivo, immunoprecipitation experiments were performed using HeLa cell nuclear extracts (NEs). Anti‐FANCJ antibody precipitated both FANCJ and BLM from the NE (Figure 1A, lane 3). Neither FANCJ nor BLM was precipitated when anti‐FANCJ antibody was omitted or when normal rabbit IgG was used (Figure 1A, lanes 4 and 5). In control experiments, the FANCJ antibody immunoprecipitated BRCA1, a known protein partner of FANCJ. However, the FANCJ antibody failed to precipitate the Werner syndrome protein (WRN) (Figure 1A, lane 3), a RecQ helicase related to BLM in the conserved ATPase/helicase core domain (Sharma et al, 2006). FANCJ and BLM were co‐immunoprecipitated from NEs using an antibody directed against BLM protein (Figure 1A, lower panel).

Figure 1

Figure 1

The alternative text for this image may have been generated using AI.

Full size image

FANCJ and BLM are reciprocally co‐immunoprecipitated from human nuclear extracts and the C‐terminal domain of FANCJ binds BLM. (A) FANCJ antibody (top) or BLM antibody (bottom) was used to co‐immunoprecipitate FANCJ and BLM from HeLa NE (nuclear extract). The blot was probed with antibodies against indicated proteins. HeLa NE (lane 2) represents 5% of the input. (B) BLM antibody (top) or FANCJ antibody (bottom) was used to co‐immunoprecipitate BLM and FANCJ from NE of BLM null (BLM–/–) or BLM corrected (BLM+/+) cells. Blots were probed as described above. (C) FANCJ antibody co‐immunoprecipitates FANCJ and BLM from HeLa NE in the presence of ethidium bromide (10 μg/ml) or DNase I (2 μg/ml). The blot was probed with goat anti‐BLM (top), rabbit anti‐FANCJ (middle), and mouse anti BRCA1 (bottom) antibodies. (D) Whole‐cell extracts of HeLa cells that were prepared from untreated or treated with MMC (50 nM) or HU (3 mM) and immunoprecipitated with anti‐FANCJ or anti‐BLM antibody as indicated. The blots were probed with indicated antibodies. (E) Purified FANCJ and BLM proteins were reciprocally co‐immunoprecipitated from a mixture of the two proteins using antibodies against FANCJ or BLM. The blot was probed with anti‐BLM (top) or anti‐FANCJ (bottom) antibodies. (F) The different FANCJ constructs are indicated with a positive (+) or negative (−) to indicate binding to BLM. (G) Myc IP experiments were performed from 293T cells that were transfected with vector alone, full‐length FANCJ or the specified FANCJ construct. IP products were analysed by western blot with Myc or BLM antibodies. The asterisk denotes the migration of the different myc‐tagged FANCJ species.

As an additional control to address the specificity of the anti‐BLM antibody, we tested for the co‐precipitation of BLM and FANCJ in isogenic BLM null and BLM corrected cell lines. Representative results from these experiments are shown in Figure 1B. Using the same antibody directed against BLM that was used for co‐immunoprecipitation experiments with HeLa cell NEs, FANCJ failed to be immunoprecipitated by the BLM antibody from BLM null cells, which lacked detectable BLM protein. In the control experiments, BLM and FANCJ were co‐immunoprecipitated from the BLM corrected cells.

We attempted similar experiments with NEs from FANCJ null and corrected cells and a FANCJ antibody; however, the level of BLM protein in the FANCJ null extract input was dramatically reduced as detected by western blot analysis (see next section), making it technically difficult to perform the experiment.

Since FANCJ and BLM both interact with DNA, we tested if their association was DNA dependent. HeLa NEs were pretreated with either DNase I or ethidium bromide (EtBr), and subsequently incubated with anti‐FANCJ antibody. Western blot analysis of the immunoprecipitates demonstrated that both BLM and FANCJ were reciprocally co‐immunoprecipitated (Figure 1C). Thus, the association of the FANCJ and BLM helicases in NEs is resistant to DNA degradation by DNase I or EtBr intercalation, suggesting that their interaction is not dependent on DNA.

Since FANCJ and BLM are known to respond to interstrand cross‐links and replication stress, we examined if their interaction was affected when cells are exposed to MMC or hydroxyurea (HU). Results from experiments using either anti‐BLM or anti‐FANCJ antibodies demonstrated that their interaction was increased approximately two‐fold after cellular exposure to 50 nM MMC or 3 mM HU (Figure 1D).

FANCJ and BLM interact directly

To determine if FANCJ and BLM interact directly, we performed co‐immunoprecipitation experiments with purified recombinant FANCJ and BLM proteins. Using antibodies directed against either FANCJ or BLM, the two helicase proteins were reciprocally co‐immunoprecipitated (Figure 1E, lanes 4 and 7). In control experiments, neither FANCJ nor BLM was precipitated when primary antibody was omitted or when normal rabbit IgG was used (Figure 1E, lanes 5, 6, 8, and 9). Moreover, FANCJ antibody failed to precipitate BLM when FANCJ was absent from the binding mixture. Likewise, BLM antibody failed to precipitate FANCJ when BLM was absent from the binding mixture (data not shown). The co‐immunoprecipitation of purified BLM and FANCJ proteins was resistant to DNase I or EtBr, suggesting that the proteins can interact independently of DNA (Supplementary Figure S1A).

To confirm that FANCJ and BLM interact directly with each other, enzyme‐linked immunoabsorbent assay was used to test for the protein interaction (Supplementary Figure S1B). FANCJ was incubated in the presence of 3% bovine serum albumin (BSA) with BLM immobilized on microtiter wells, and the bound FANCJ protein was detected immunologically. The colorimetric signal was dependent on FANCJ protein concentration (Supplementary Figure S1C). Specific binding of FANCJ to BLM was analysed according to Scatchard‐binding theory. Data were analysed by a Hill plot and found to be linear, indicating a single set of binding sites for FANCJ with BLM. The apparent dissociation (_K_d) was determined to be 4 nM. The colorimetric signal from FANCJ binding to immobilized BLM was specific since a significantly reduced signal was observed in wells precoated with BSA (Supplementary Figure S1C). The colorimetric signal from the FANCJ–BLM interaction was resistant to pretreatment of both FANCJ protein and BLM with DNase I (2 μg/ml) (Supplementary Figure S1C), suggesting that a contaminating DNA bridge is not responsible for the signal.

The C‐terminal domain of FANCJ binds to BLM

To define the domain on FANCJ required for BLM binding, several FANCJ‐myc fusion proteins of varying length were expressed in 293T cells, and the corresponding FANCJ protein fragments were precipitated from the cell lysates with c‐Myc antibody tagged agarose beads (Figure 1F). Full‐length FANCJ and FANCJ expression construct FANCJ1−900 precipitated BLM, but FANCJ1−660 did not, suggesting that the FANCJ N‐terminal residues 1–660 are not required for BLM binding (Figure 1G). Expression of C‐terminal FANCJ fragments FANCJ881−1249 and FANCJ881−1060 were sufficient to co‐precipitate BLM, suggesting that a BLM interaction site resides in a C‐terminal non‐catalytic domain of FANCJ spanning residues 881–1060 (Figure 1G).

Cellular BLM protein levels are strongly dependent on FANCJ status

Since the evidence suggested that BLM and FANCJ interact with each other and may exist in a complex in vivo, we wanted to assess if their respective protein stabilities might be affected by each other. To address this possibility, we performed western blot analyses of BLM and FANCJ in either FA‐J mutant cells or HeLa cells that had been depleted of FANCJ protein using RNA interference. BLM protein levels were dramatically decreased by 80–90% in FA‐J cells or HeLa cells that had been depleted of FANCJ (Figure 2A). In control experiments, WRN protein levels were not significantly reduced in the FA‐J mutant cells or in cells that had been depleted of FANCJ (Figure 2A). Time course experiments demonstrated that BLM protein levels declined within 24 h after a single siRNA‐FANCJ transfection in HeLa cells (Figure 2B).

Figure 2

Figure 2

The alternative text for this image may have been generated using AI.

Full size image

Cellular BLM protein levels are strongly dependent on FANCJ status. (A) NE was prepared from HeLa, FA‐J, siRNA‐control‐treated HeLa, and siRNA‐FANCJ‐treated HeLa cells, and proteins were transblotted and probed with indicated antibodies. (B) HeLa cells were transfected with siRNA‐FANCJ or siRNA‐control and harvested at indicated time points after a single transfection. Whole‐cell extracts were prepared as described in Materials and methods. (C) RT–PCR experiments were performed as described in Materials and methods. Experimental data represent the mean of three independent experiments in triplicate with standard deviations indicated by error bars. (D) NE was prepared from FA‐J mutant and FA‐J corrected cells, and proteins were transblotted and probed with indicated antibodies. (E) Western blot showing FANCJ or BLM levels of either control or FANCJ‐siRNA transfected HeLa cells treated with indicated concentrations of MG132 for 6 h. Untreated cells were subjected to DMSO treatment alone. (F) NE was prepared from HeLa, BLM−/−, and BLM+/+ cells, and proteins were transblotted and probed with anti‐BLM and anti‐FANCJ antibodies. For all western blots, β‐actin served as a loading control.

The significant reduction in BLM protein levels in FANCJ‐depleted cells raised the possibility that BLM transcript abundance was reduced. Real‐time PCR analysis of BLM and FANCJ transcript levels were performed using siRNA‐FANCJ and siRNA‐control transfected HeLa cells. The relative BLM transcript levels were the same in FANCJ‐depleted and control cells (Figure 2C), indicating that the effect of FANCJ depletion on BLM protein levels was not manifested at the transcription level.

To confirm that the cellular level of BLM protein is dependent on FANCJ status, we examined BLM protein levels in an isogenic pair of FANCJ mutant and corrected cell lines. As shown in Figure 2D, BLM protein levels were significantly reduced in the FA‐J mutant cells compared with the same FANCJ corrected cells. Since BLM is known to interact with TopoIIIα (Wu et al, 2000; Johnson et al, 2000; Hu et al, 2001), RPA70 (Brosh et al, 2000), and FEN‐1 (Sharma et al, 2003), we examined the levels of these BLM‐interacting proteins in the FA‐J mutant and corrected isogenic pair of cell lines. TopoIIIα, RPA70, and FEN‐1 protein levels were comparable in the FA‐J mutant and corrected cell lines (Figure 2D), suggesting that the effect of FANCJ status on BLM protein level is highly specific.

To better understand the effect of FANCJ status on BLM protein stability, we examined the effect of the proteasome inhibitor MG132 on BLM protein level in FANCJ‐depleted or mock‐treated cells. In these experiments, HeLa cells transfected with control or FANCJ siRNA were subsequently treated with the indicated concentrations of MG132 for 6 h. Whole‐cell lysates were evaluated by western blot for FANCJ and BLM protein. In FANCJ‐depleted cells, BLM protein was restored in cells exposed to 10 μM MG132 (Figure 2E). These results suggest that BLM protein is degraded by a proteasome‐mediated pathway in FANCJ‐deficient cells.

Next, we examined FANCJ protein levels in an isogenic pair of BLM mutant and corrected cell lines. As shown in Figure 2F, FANCJ protein amounts remained comparable in the BLM mutant and corrected cells, indicating that cellular FANCJ protein levels are not affected by BLM status.

Interaction of FANCJ and BLM remains intact in FANCA, FANCD2, or BRCA1‐deficient cells

We next asked whether the FANCJ–BLM interaction was lost in FA‐deficient cells. Western blot analysis of NEs showed that BLM protein levels were reduced in FA‐A (Figure 3A, lane 2 compared with lane 1) or FA‐D2 (Figure 3A, lane 4 compared with lane 3) mutant cell lines compared with the corrected cells. FANCJ and BLM could be co‐immunoprecipitated from NE of the FA‐A or FA‐D2 mutant cell lines (Figure 3A), suggesting that FANCJ and BLM are associated with each other in cells that are defective in either a member of the FA core complex (FANCA) or FANCD2.

Figure 3

Figure 3

The alternative text for this image may have been generated using AI.

Full size image

The interaction between BLM and FANCJ helicases is not dependent on the integrity of the FA core complex, FANCD2, or BRCA1. (A) Immunoprecipitate from NE of the indicated FA mutant or corrected cell line using goat anti‐BLM antibody (top panel) or rabbit anti‐FANCJ antibody (middle panel) was analysed by western blotting using anti‐FANCJ and anti‐BLM antibodies. Input (bottom panel) represents 5% of NE input for co‐immunoprecipitation experiments. (B) Immunoprecipitate from NE of BRCA1 mutant, or BRCA1 corrected cell lines using anti‐FANCJ antibody (top panel) or goat anti‐BLM antibody (bottom panel) was analysed by western blotting using anti‐FANCJ and anti‐BLM antibodies.

Since FANCJ is known to interact with BRCA1 (Cantor et al, 2001), we asked if the FANCJ–BLM interaction would remain intact in BRCA1‐deficient cells. Co‐immunoprecipitation experiments were performed using BRCA1 mutant and corrected cells. FANCJ and BLM were reciprocally co‐immunoprecipitated from BRCA1 mutant or corrected cells (Figure 3B), suggesting that the FANCJ–BLM interaction is not dependent on BRCA1. Furthermore, since the binding site for BLM on FANCJ overlaps that of BRCA1, we sought to determine if a FANCJ‐S990A mutant defective in the BRCA1 interaction would still be able to bind BLM. The results from anti‐FANCJ immunoprecipitation experiments using extracts from transfected FA‐J cell lines expressing FANCJ‐S990A or FANCJ‐WT demonstrated that BLM was co‐immunoprecipitated with either FANCJ protein equally well, whereas BRCA1 failed to be co‐immunoprecipitated with FANCJ‐S990A (Supplementary Figure S2A).

The BRCA1‐associated complex (BASC) identified by Wang et al (2000) contains BRCA1 and BLM, as well as BLM‐interacting partners MLH1 and RPA. To determine if proteins found in the BASC complex associate with BRCA1 through FANCJ, anti‐BRCA1 co‐immunoprecipitation experiments were performed with extracts from FA‐J null and corrected cells. MLH1, RPA, and FANCA were all co‐immunoprecipitated with BRCA1 in the FA‐J null cells similar to FA‐J corrected cells (Supplementary Figure S2B).

To determine if the interaction of the BLM‐interacting protein TopoIIIα with the FA core complex (Deans and West, 2009) was altered when FANCJ was absent, anti‐FANCA co‐immunoprecipitation experiments were performed with extracts from FA‐J null and corrected cells. TopoIIIα was co‐immunoprecipitated with FANCA in the FA‐J null cells similar to FA‐J corrected cells (Supplementary Figure S2C).

Induction of FANCJ and BLM foci following replication stress

Evidence suggests that both BLM (Bachrati and Hickson, 2008) and FANCJ (Wu et al, 2008; Hiom, 2010) are important for the cellular response to replication stress. To address the possibility that the two helicases become associated with each other in subnuclear foci after replication stalling, we examined the co‐localization of endogenous BLM and FANCJ by immunofluorescence after exposure of HeLa cells to the replication inhibitor HU. Both BLM and FANCJ formed distinct nuclear foci after HU treatment (Figure 4; Supplementary Figure S3), a result consistent with earlier published data (Davies et al, 2004; Pichierri et al, 2004). Furthermore, BLM and FANCJ foci partially co‐localized after replication stress imposed by HU (Figure 4; Supplementary Figure S3). FANCJ and BLM foci also partially co‐localized in HeLa cells exposed to MMC (50 ng/ml) or ionizing radiation (IR; 10 Gy); however, the number and intensity of BLM and FANCJ foci formed was not as great in the case of IR treatment (data not shown). The stabilizing effect of FANCJ on BLM protein levels may help maintain independent recruitment of BLM to stalled forks or damaged sites.

Figure 4

Figure 4

The alternative text for this image may have been generated using AI.

Full size image

FANCJ and BLM foci formation after cellular exposure to the replication inhibitor hydroxyurea. HeLa cells were either left untreated or treated with HU (2 mM) as described in Materials and methods. After fixation and permeabilization, cells were stained with anti‐FANCJ (red) or anti‐BLM (green) antibodies. After treatment with HU, BLM partially co‐localizes in nuclear foci that coincide with FANCJ foci as shown in the merged images.

Although BLM protein levels are significantly reduced in FANCJ null or FANCJ‐depleted cells, the residual BLM protein in FANCJ‐deficient cells formed nuclear foci after DNA damage (MMC, IR) or replication stress (HU) (Supplementary Figure S4A and B). FANCJ foci formed after HU, MMC, or IR in BLM null cells (Supplementary Figure S4C). These results suggest that BLM or FANCJ foci are able to form after DNA damage or replication stress in the absence of the interacting helicase partner.

FANCJ‐depleted cells are sensitive to replication stress

We next sought to determine if FANCJ‐deficient cells, like BLM‐deficient cells, are sensitive to replication stress. To do this, FANCJ‐depleted and control‐siRNA HeLa cells (Figure 5A) were exposed to increasing concentrations of HU and cell proliferation was evaluated by the WST‐1 assay as a measure of metabolic activity. FANCJ‐depleted cells were significantly more sensitive to all HU concentrations analysed (Figure 5B). As a control, FANCJ‐depleted cells were also more sensitive to the DNA cross‐linking agent MMC compared with siRNA‐control cells (Supplementary Figure S5). We also examined HU sensitivity in cells depleted of FANCJ that were deficient in BLM (PSNG13) compared with BLM‐proficient cells (PSNF5) (Figure 5C and D). PSNF5 cells depleted of FANCJ displayed reduced BLM protein level and were more sensitive to HU compared with control‐siRNA PSNF5 cells (Figure 5D), a result that was consistent with FANCJ depletion studies in HeLa cells (Figure 5B). However, FANCJ‐depleted PSNG13 cells were more HU resistant than control‐siRNA PSNG13 cells (Figure 5D), suggesting that the presence of FANCJ in a BLM‐deficient background is toxic to the cells during replication stress.

Figure 5

Figure 5

The alternative text for this image may have been generated using AI.

Full size image

FANCJ‐depleted cells are sensitive to replication stress. HeLa cells were transfected twice with either control or FANCJ siRNA as described in Materials and methods and evaluated for FANCJ protein level with β‐actin serving as a loading control (A). One day after the second siRNA transfection, cells were exposed to the indicated concentrations of HU (B). Cell proliferation was measured using the WST‐1 assay (Roche, Indianapolis, IN) after 48 h. (C, D) PSNG13 (BLM–/–) or PSNF5 (BLM+/+) cells were transfected twice with either control or FANCJ siRNA as described under Experimental Procedures and evaluated for FANCJ protein level with β‐actin serving as a loading control (C). One day after the second siRNA transfection, cells were exposed to the indicated concentrations of HU (D). Cell proliferation was measured using the WST‐1 assay after 48 h. (E, F) FA‐J cells transfected with vector alone, wild‐type FANCJ, or FANCJ‐K52R were evaluated for expression of FANCJ/FANCJ‐K52R protein (E). The transfected cells were exposed to the indicated concentrations of HU and cell proliferation was measured using the WST‐1 assay after 48 h (F). Experimental data are the mean of three independent experiments done in triplicate and s.d. indicated by error bars.

The HU sensitivity of FANCJ‐deficient cells may be due to a reduced level of BLM. Alternatively, FANCJ may have a more direct role in HU resistance. To address the importance of FANCJ catalytic activity in HU resistance, FA‐J cells were transfected with a plasmid expressing the helicase‐inactive FANCJ‐K52R Walker A box mutant protein (Cantor et al, 2004) and HU resistance of these cells was compared with FA‐J cells expressing FANCJ‐WT or transfected with empty vector (Figure 5). As shown in Figure 5E, expression of FANCJ‐K52R in FA‐J cells restored BLM protein levels similar to FA‐J cells expressing FANCJ‐WT; moreover, HU resistance was partially but not fully restored (Figure 5F). These results suggest that FANCJ ATPase/helicase activity is required for complete HU resistance.

Domain mapping experiments demonstrated that BLM interacts with the C‐terminal region, but not the N‐terminal region, of FANCJ (Figure 1F and G). To determine if the physical interaction of BLM with FANCJ is important for HU resistance, Myc‐tagged FANCJ fragments were expressed in 293T cells (Figure 6A). The transfected cell lines were tested for HU sensitivity and co‐immunoprecipitation of endogenous FANCJ and BLM. 293T cells expressing the BLM‐interacting Myc‐FANCJ881−1249 or Myc‐FANCJ881−1060 were found to be sensitive to HU compared with cells expressing Myc‐FANCJ1−499 or transfected with empty vector expressing just the Myc epitope (Figure 6B). To determine if expression of FANCJ fragments affected the interaction of endogenous FANCJ with BLM, anti‐BLM co‐immunoprecipitation experiments were performed. The FANCJ–BLM co‐immunoprecipitation was decreased by 2.5‐fold for extracts from cells expressing Myc‐FANCJ881−1060 and 3.1‐fold for Myc‐FANCJ881−1249 (Figure 6C). These results suggest that expression of a C‐terminal fragment of FANCJ that interacts with BLM exerts a dominant negative effect on HU resistance by interfering with the FANCJ–BLM interaction.

Figure 6

Figure 6

The alternative text for this image may have been generated using AI.

Full size image

BLM‐interacting fragments of FANCJ expressed in 293T cells disrupt the FANCJ–BLM interaction and lead to HU sensitivity. 293T cells transfected with vector or indicated FANCJ constructs were evaluated for their expression (A). The transfected cells were treated with the indicated concentrations of HU. Cell proliferation was measured using the WST‐1 assay after 48 h (B). Experimental data are the average of three independent experiments done in triplicate and s.d. indicated by error bars. (C) Whole‐cell lysates prepared from 293T cells transfected with either vector or indicated FANCJ constructs were used for immunoprecipitation with anti‐BLM antibody. The immunoprecipitate was analysed by western blot using anti‐BLM or anti‐FANCJ antibody. Whole‐cell lysate from transfected 293T cells represents 5% of the input used for the co‐immunoprecipitation.

Increased SCE in FA‐J cells

A hallmark of BS is elevated SCE, raising the possibility that FANCJ mutated cells may display a similar form of chromosomal instability. Chromosomal spreads from wild‐type, FA‐J, and FA‐C fibroblasts were prepared and analysed for SCE. FA‐J fibroblasts displayed a significantly greater level of SCE compared with the wild‐type cells (Table I). In contrast, FA‐C cells displayed a level of SCE that was comparable to wild‐type cells.

Table I Quantification of SCE frequency

Full size table

To assess the effect of a combined BLM and FANCJ deficiency on SCE, FANCJ was depleted from an isogenic pair of BLM null and BLM corrected cells. The results from these assays demonstrated that FANCJ depletion in a BLM null background did not increase SCE frequency (Supplementary Table 1). However, FANCJ depletion in a BLM wild‐type background increased SCE frequency (Supplementary Table 1), a result that was consistent with what was observed for FA‐J cells (Table I). These results suggest that acutely depleting FANCJ in a BLM‐deficient background does not further elevate SCE frequency.

Functional interaction of BLM and FANCJ helicases

The physical interaction between FANCJ and BLM raised the possibility that the two helicases might functionally interact through their DNA unwinding functions. To test this, a mixture of purified FANCJ and BLM helicases was evaluated for their ability to unwind DNA substrates and compared with unwinding reactions performed using either helicase alone. FANCJ and BLM unwinding of a simple forked duplex substrate was additive (Supplementary Figure S6), suggesting that BLM and FANCJ do not synergistically interact on an undamaged forked duplex substrate.

Previously, we reported that FANCJ helicase activity was profoundly sensitive to a polyglycol backbone modification in either the translocating or non‐translocating strands of the duplex, suggesting that FANCJ interactions with the sugar phosphate backbone of both strands of the duplex are important as FANCJ tracks along the DNA molecule and separates the strands (Gupta et al, 2006). To test if FANCJ and BLM might work together to unwind the forked duplex with the backbone discontinuity, we performed experiments similar to those described for the undamaged forked duplex, except that the substrate contained a polyglycol linkage within the duplex in either the bottom (BLM translocating) or top (FANCJ translocating) strands (Figure 7A and C). BLM helicase alone failed to unwind the forked duplex substrate containing a polyglycol linkage in the bottom strand along which it translocates (Figure 7A); however, BLM was able to unwind the same forked duplex substrate when the polyglycol linkage was positioned in the top strand opposite to the one that BLM translocates along (Figure 7B). In contrast, FANCJ poorly unwound the DNA substrate with a polyglycol substrate in either the top (Figure 7A) or bottom (Figure 7B) strands, consistent with our previous observations (Gupta et al, 2006). These results suggest that FANCJ is uniquely sensitive to the backbone disruption and distinguishes its behaviour from that of the human RecQ helicase BLM.

Figure 7

Figure 7

The alternative text for this image may have been generated using AI.

Full size image

FANCJ and BLM helicases together more efficiently unwind a damaged DNA substrate with sugar phosphate backbone discontinuity. Helicase assays were performed as described under Materials and methods using the indicated concentrations of BLM and FANCJ helicases and forked duplexes of 31 bp, which contained a polyglycol linkage in the sugar phosphate backbone in either the bottom strand (A) or top strand (B) as indicated. Filled triangle, heat‐denatured DNA substrate control. (C) The quantitative analysis of unwinding data on the substrate with backbone discontinuity in the top strand (B) in the presence of BLM alone or BLM with FANCJ‐WT or FANCJ‐K52R. Note that FANCJ alone poorly unwound the forked duplex substrate with backbone discontinuity in either the top or bottom strands (lane 10). Data represent the mean of at least three independent experiments with standard deviations indicated by error bars. (D, E) The quantitative analysis of WRN and RECQ1 helicase reactions, respectively, in the presence of FANCJ (open circle) or the absence of FANCJ (filled circle) conducted with the 31‐bp forked duplex that had a polyglycol linkage in the top strand.

We next asked if the combined presence of BLM and FANCJ might result in a synergistic unwinding of the damaged DNA substrates. As shown in Figure 7A, a mixture of the FANCJ and BLM helicases completely failed to unwind the forked duplex substrate harbouring the polyglycol linkage in the bottom, BLM translocating, strand. However, incubation of both BLM and FANCJ helicases with the forked duplex harbouring the polyglycol linkage in the top (FANCJ translocating) strand resulted in a significantly increased level of unwinding throughout the BLM titration (1.2–10 nM) (Figure 7B, and shown quantitatively in Figure 7C). It was conceivable that FANCJ binding to the fork rather than its helicase activity was responsible for maximal promotion of BLM helicase activity on the damaged DNA substrate. However, results from BLM helicase assays on this substrate in the presence of the T4 helicase loading protein gp59 that preferentially binds forked DNA structures (Mueser et al, 2000) demonstrated that BLM unwinding activity was not stimulated (Supplementary Figure S7), suggesting that the effect is not attributed to simply forked DNA substrate binding by the FANCJ protein. Incubation of BLM and a helicase‐inactive FANCJ‐K52R mutant with the forked duplex substrate containing the polyglycol linkage in the top strand yielded a greater level of unwinding throughout the BLM titration as well; however, the unwinding at a given BLM concentration was not as high as the level observed when BLM and FANCJ‐WT was incubated with the damaged DNA substrate (Figure 7C). FANCJ‐K52R did not stimulate BLM helicase activity on the DNA substrate with the polyglycol linkage in the bottom strand (data not shown). These results demonstrate that under these reaction conditions, FANCJ can stimulate BLM helicase activity on a DNA substrate with a backbone discontinuity in the FANCJ translocating strand.

To determine if the interaction between FANCJ and BLM helicases was specific, we tested the effect of the human RecQ helicases WRN and RECQ1 on the ability of FANCJ to unwind the forked duplex substrate with the polyglycol linkage in the top strand. For these substrates, both RECQ1 and WRN behave like BLM, and are inhibited by the polyglycol linkage in a strand‐specific manner, that is, RECQ1 and WRN fail to unwind the forked duplex substrate when the polyglycol linkage is located within the duplex in the translocating (bottom) strand (manuscript in preparation). Incubation of either WRN (Figure 7D) or RECQ1 (Figure 7E) and FANCJ helicases with the forked duplex harbouring the polyglycol linkage in the top, FANCJ translocating, strand resulted in little or no increase in DNA unwinding compared with WRN or RECQ1 alone. These results suggest that WRN or RECQ1 do not interact with FANCJ in a manner to unwind the damaged DNA substrate as observed in reaction mixtures containing BLM.

Discussion

In this study, we report that the FANCJ and BLM helicases can be reciprocally co‐immunoprecipitated from human NEs, and that their association is not dependent on the integrity of the FA core complex, FANCD2, or BRCA1. We observed that the FANCJ–BLM interaction is enriched when cells are exposed to HU or DNA damaging agents and the helicases co‐localize in nuclear foci after HU, suggesting that they operate together in a pathway responsible for helping cells cope with replication stress. Indeed, we show that FANCJ‐deficient cells, like BLM‐deficient cells, are sensitive to the replication inhibitor HU and show elevated SCE. FANCJ status profoundly influences cellular BLM protein levels, suggesting their direct physical interaction in vivo. In support of this, biochemical interaction assays with purified recombinant proteins demonstrated that BLM and FANCJ helicases directly bind to each other in a manner that is not dependent on DNA. Thus, BLM and FANCJ interact with each other in vitro and in vivo.

One of the most striking observations from our studies is that BLM protein levels are significantly reduced in FANCJ null cells or in cells that are acutely depleted of FANCJ using RNA interference. Although the BLM protein level was markedly reduced in FANCJ‐deficient cells, the levels of interacting partners of BLM (RPA, FEN‐1, TopoIIIα) were not reduced, suggesting that the absence or reduction of intracellular FANCJ protein concentration exerted an effect specifically on BLM. Presumably, the physical interaction between BLM and FANCJ helps to stabilize the BLM protein in vivo; however, the converse is not true, that is, the FANCJ protein level was not different in BLM mutant and corrected cells. The observed decline in BLM protein level in FANCJ‐deficient cells suggests that cellular phenotypes associated with FANCJ deficiency may at least partly reflect an underlying reduction in BLM protein abundance. Interestingly, BLM protein was also reduced in FA‐A or FA‐D2 cell lines. BLM protein levels are known to be affected when certain interacting partners (e.g., RMI1 (Yin et al, 2005) and RMI2 (Xu et al, 2008)) are deficient, suggesting that BLM folding or susceptibility to proteasome degradation is highly influenced by interacting/associated proteins in the FA or related pathways. Both BS cells (Pichierri et al, 2004) and FA‐J cells are hypersensitive to DNA cross‐linking agents (Litman et al, 2005; Peng et al, 2007). A hallmark of BS is elevated SCE (Hanada and Hickson, 2007), also detected in chicken brip1 cells (Bridge et al, 2005) and, as shown here, in FANCJ‐deficient human cells. Although the ∼2‐fold SCE increase in FA‐J cells was not as great as the 7‐ to 10‐fold increase observed in cells from BS patients (Chaganti et al, 1974; German, 1993), it is comparable to that observed following depletion of BLM (Yin et al, 2005) or FANCM (Deans and West, 2009). In human cells, FANCJ knockdown by RNA interference resulted in elevated MMC‐induced chromosomal abnormalities including breakage, chromatid interchanges, triradials, and quadriradials (Bridge et al, 2005; Litman et al, 2005), a chromosomal structure typically observed in BS patients (van Brabant et al, 2000). Hence, the chromosomal instability associated with FANCJ deficiency may at least partly reflect impaired BLM function in FA‐J null cells.

The fact that DNA damage‐induced FANCD2 monoubiquitination is intact in FANCJ‐deficient cells (Litman et al, 2005) suggests that FANCJ helicase functions downstream in the FA pathway. Similarly, BLM is not required for FANCD2 monoubiquitination or focus formation (Pichierri et al, 2004), suggesting a downstream role as well. The FA core complex is required for BLM phosphorylation following treatment with a DNA cross‐linker but not HU (Pichierri et al, 2004), implicating the FA core complex as an upstream regulator of BLM function by setting the stage for homologous recombinational repair at blocked replication forks. Moreover, Hirano et al (2005) reported that MMC‐induced formation of GFP‐BLM foci was dramatically reduced in both human and chicken fancc or fancd2 cells. A functional linkage between FANCC and BLM was proposed since the level of SCE in chicken blm cells was similar to that of fancc/blm cells (Hirano et al, 2005). Cross‐talk between BLM and proteins of the FA pathway is supported by experimental evidence that the FA core protein complex and the BLM complex associate together in a supercomplex known as BRAFT (BLM, RPA, FA, and TopoIIIα) (Meetei et al, 2003). Although FANCJ was not identified by mass spec analysis of BLM‐interacting proteins from unstressed cells, it is possible that the interaction of FANCJ with BLM is transient and less stable than that of the BRAFT complex or that FANCJ and BLM exist in a protein complex that is regulated following cellular stress and is distinct from the BRAFT complex. In subsequent work, Singh et al (2008) identified FANCM in complexes containing BLM and RMI2, a novel factor required for the stability of the BLM complex and also important for chromosomal stability (Singh et al, 2008; Xu et al, 2008). Our results demonstrate that BLM interacts with FANCJ in addition to the FA core complex. Since BLM, but not TopoIIIα, is depleted in FANCJ‐deficient cells, it is possible that FANCJ regulates a very specific subset of BLM reactions (e.g. fork regression) not requiring the topoisomerase.

BLM co‐localizes with FANCJ (this study) and FANCD2 (Pichierri et al, 2004) in response to cellular exposure to HU or MMC. Therefore, it is likely that BLM, FANCJ, and monoubiquintylated FANCD2 preferentially associate with each other at stalled or blocked replication forks. The single‐stranded DNA‐binding protein RPA, which interacts physically and functionally with both BLM and FANCJ, strongly co‐localizes with FANCJ after replication stress as well (Gupta et al, 2007). The concerted action of BLM and FANCJ with their protein partners such as RPA and FANCD2 is likely to be important for their collective role in the recovery from S‐phase progression and maintenance of genomic stability. The observations that FANCD2 monoubiquitination is required for its association with chromatin and subnuclear localization with BRCA1, RAD51, MRE11‐RAD50‐NBS1, RPA, PCNA, and BRCA2 (Gurtan and D'Andrea, 2006), together with the current results, suggest a scenario in which the critical DNA repair steps of the FA pathway are orchestrated by FANCD2 and enacted by the FANCJ and BLM helicases with HR and DNA repair proteins.

The concerted action of FANCJ and BLM helicases to unwind a DNA structure associated with a stalled replication fork or HR repair may enable subsequent processing by structure‐specific nucleases to ensure genome integrity or serve to clear proteins bound to DNA (e.g. Rad51; Bugreev et al, 2007; Sommers et al, 2009) to complete recombinational repair. Both BLM and FANCJ recognize and unwind a D‐loop structure (Gupta et al, 2005; Bachrati et al, 2006), an early intermediate of HR. However, only BLM unwinds a Holliday Junction structure (Karow et al, 2000; Mohaghegh et al, 2001; Gupta et al, 2005), and FANCJ does not stimulate BLM unwinding of this late HR intermediate (our unpublished data).

The functional interaction between FANCJ and BLM on the DNA substrate with a polyglycol linkage in the BLM translocating strand suggests that the helicase interaction may be important for traversing backbone damage. The ability of FANCJ to interact with BLM may facilitate unwinding past the lesion in a specific manner, since DNA unwinding by WRN or RECQ1 on the same substrate was not increased by the presence of FANCJ. The ability of the catalytically inactive FANCJ‐K52R mutant protein to stimulate BLM helicase activity suggests that the protein interaction is an important factor in the functional interaction. However, further mechanistic studies are necessary to understand the importance of the physical interaction between BLM and FANCJ for DNA unwinding by the two helicases acting together perhaps in a protein complex. A paradigm for two helicases with opposite polarities functioning together was provided by the molecular characterization of the RecBCD complex in which RecB and RecD helicases, which translocate in opposite directions along ssDNA, unwind duplex DNA in the same direction by moving along complementary strands of the Watson–Crick DNA double helix (Singleton et al, 2004).

Although BLM and FANCJ interact, the distinctive clinical features of BS and FA group J patients indicate that the two helicases also have independent roles. It will be important to quantify BLM protein levels in individuals with FANCJ mutations and relate this to their clinical symptoms. According to the Fanconi Anemia Mutation Database maintained by The Rockefeller University, FANCJ protein variants exist that fall into three classes: substitutions, frameshifts, and nonsense mutations. Clinical heterogeneity in FA‐J patients may be dependent on the nature of the FANCJ mutation as it relates to the stability of the dysfunctional FANCJ protein it encodes. For example, western blot analysis of FA‐J patient cell lines demonstrated the absence of full‐length FANCJ protein in cells harbouring frameshift, nonsense, or deletion mutations, but residual full‐length FANCJ protein in a cell line harbouring a FANCJ missense mutation (Levitus et al, 2005). The interaction of FANCJ with BLM and the strong effect of FANCJ deficiency on BLM protein stability suggest that FA‐J patient cell lines should be screened for BLM protein by western blot analysis. BLM deficiency in FA‐J patients may be a contributory factor to the nature and degree of chromosomal instability, cancer, and associated phenotypes observed in the individuals.

Materials and methods

Proteins

Baculoviruses encoding FANCJ or FANCJ‐K52R with a C‐terminal FLAG tag were used to infect high five insect cells and the recombinant FANCJ protein was purified as described previously (Cantor et al, 2004). Recombinant WRN (Sharma et al, 2004), RECQ1 (Sharma et al, 2005), and BLM (Karow et al, 1999) proteins were purified as described previously. Purified gene 59 protein was a kind gift from Dr Stephen Benkovic (Pennsylvania State University).

Cell lines

HeLa, FA‐J (EUFA30‐F), and 293T cells were grown in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin‐streptomycin, and 1% l‐glutamine at 37°C in 5% CO2. FA‐J cells infected with the pOZ retroviral vector containing no insert or wild‐type FANCJ were prepared as described (Peng et al, 2007). BLM mutant (PSNG13) and BLM corrected (PSNF5) were grown in the same media but supplemented with 350 μg/ml G418. Immortalized fibroblasts FA‐A (PD220), FA‐D2, and their respective corrected cells, provided by the Fanconi Anemia Research Fund, were grown in the same medium as HeLa but supplemented with 0.2 mg/ml puromycin. HCC1937, which carries a hypomorphic mutation in BRCA1, and the corresponding BRCA1‐reconstituted cells were grown in RPMI1640 supplemented with 10% FBS and 1% penicillin‐streptomycin.

For FANCJ knockdown, FANCJ siRNA (100 nM; Dharmacon, Chicago, IL) was transfected into HeLa cells using Lipofectamine‐2000 (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. After two transfections on consecutive days, cells were harvested 48 h after the second transfection.

GFP‐vector or GFP‐FANCJ plasmid DNA (2 μg) was used for transfecting BLM null and BLM corrected cells using Lipofectamine‐2000 (Invitrogen). FA‐J null cells were transfected using the Nucleofector Kit (Lonza) with plasmid DNA encoding Myc‐FANCJ‐WT, Myc‐FANCJ‐S990A, or vector according to the manufacturer's protocol.

MG132 (Sigma, St Louis, MO) was dissolved in DMSO (10 mM) and small aliquots (30 μl) were stored at −80°C. HeLa cells transfected with control or FANCJ siRNA were subsequently treated with the indicated concentrations of MG132 for 6 h. Untreated cells were subjected to DMSO treatment alone.

SCE analysis

Cells were incubated with 10 μM BrdU for two cell cycles and treated with 0.1 μg/ml Colcemid for the last 2 h. After harvesting, cells were incubated in 0.075 M KCl solution for 20 min, and subsequently fixed in methanol and acetic acid (3:1 ratio) for 30 min. Cells were dropped onto 50%‐ethanol‐wet slides and dried. Slides were then incubated with 10 μg/ml Hoechst 33258 in phosphate buffer (pH 6.8) for 20 min. Slides were exposed to 352 nm black light for 30–60 min, followed by incubation in 2 × SSC at 62°C for 1 h. They were then stained using 3% Giemsa solution (pH 6.8) for 10 min.

Co‐immunoprecipitation experiments and western blot analyses

NEs were prepared from exponentially growing HeLa, FA‐A, and FA‐D2 cells as described previously (Dignam et al, 1983).

NE (1 mg protein) was incubated with rabbit anti‐FANCJ polyclonal antibody (1 μg; Sigma), goat anti‐BLM (1 μg C‐18; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti‐BRCA1 (1 μg 9010; Cell Signaling Technology, Danvers, MA; rabbit anti‐FANCA (kindly provided by Dr Weidong Wang, NIA‐NIH), normal rabbit IgG antibody or normal goat IgG antibody (1 μg, Santa Cruz Biotechnology) in Buffer D (50 mM HEPES (pH 7.5), 100 mM KCl, 10% glycerol) for 2 h, and subsequently mixed with 20 μl protein‐G agarose (Roche) for 2 h at 4°C. Beads were washed with Buffer D containing 0.1% Tween‐20. Immunoprecipitations from FA mutant and BRCA1 null cell lines were performed using whole‐cell lysates prepared as described previously (Gupta et al, 2007). Proteins were eluted by boiling in SDS sample buffer, resolved on 10% polyacrylamide Tris‐glycine SDS gels, and transferred to PVDF membranes (Amersham Biosciences, Piscataway, NJ). Membranes were probed for FANCJ or BLM using the rabbit polyclonal anti‐FANCJ (1:2000) or goat anti‐BLM (1:500) antibodies, respectively. Proteins on immunoblots were detected using ECL (Amersham Biosciences).

Co‐immunoprecipitation experiments using purified proteins were performed by incubating purified FANCJ (100 ng) and purified BLM (100 ng) proteins together at 4°C for 2 h. This binding mixture was used as the input for immunoprecipitation experiments, as described above.

For FANCJ mapping studies, plasmids encoding Myc‐FANCJ fragments were used to transfect 293T cells as described previously (Peng et al, 2007). 293T cells were harvested and lysed in 150 mM NETN lysis buffer (20 mM Tris (pH 8.0), 150 mM NaCl, 1 mM EDTA, 0.5% NP‐40, 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, 10 μg/ml aprotinin) for 30 min on ice. Cell extracts were clarified by centrifugation. For immunoprecipitation assays, cells lysates were incubated with c‐Myc antibody tagged agarose beads (Sigma) at 4°C for 2 h. Beads were subsequently washed and boiled in SDS loading buffer. Proteins were separated using SDS–PAGE and electrotransferred to nitrocellulose membranes. Membranes were blocked in 5% milk phosphate‐buffered saline/Tween and incubated with primary Ab against the Myc epitope (9e10, Santa Cruz, 1:500) or BLM (Abcam, 1:1000) for 1 h. Membranes were washed, and incubated with horseradish peroxidase‐linked secondary Abs (Santa Cruz, 1:5000), and detected by chemiluminescence (Amersham).

For western blot detection of other proteins, the following primary antibodies were used: WRN, mouse monoclonal against Glutathione‐S‐Transferase tagged WRN1072−1432, Spring Valley Labs (Sykesville, MD), 1:1000; BRCA1, mouse monoclonal, Santa Cruz Biotechnology, 1:1000; RPA70, mouse monoclonal, Calbiochem (San Diego, CA), 1:1000; TopoIIIα, rabbit polyclonal, Corgen (Taipei, Taiwan), 1:1000; FEN‐1, mouse monoclonal, Abcam (Cambridge, MA), 1:1000; FANCA (kindly provided by Dr Weidong Wang, NIA‐NIH); MLH1, mouse monoclonal, BD Biosciences Pharmingen (Basel, Switzerland), 1:1000.

DNA substrates

The 22 and 31 bp duplex DNA substrates were constructed as described previously (Gupta et al, 2006, 2007). PAGE‐purified oligonucleotides used for the preparation of DNA substrates were purchased from Loftstrand Labs (Gaithersburg, MD).

Radiometric strand displacement helicase assays

BLM and FANCJ helicase assay reaction mixtures (20 μl) contained 40 mM Tris–HCl (pH 7.6), 25 mM KCl, 5 mM MgCl2, 2 mM dithiothreitol, 2% glycerol, 100 ng/μl BSA, 2 mM ATP, 10 fmol of the specified duplex DNA substrate (0.5 nM DNA substrate concentration), and the indicated concentrations of FANCJ or BLM helicases. WRN (Sharma et al, 2004) and RECQ1 (Sharma et al, 2005) reaction mixtures were prepared as described previously. Helicase reactions were initiated by the addition of FANCJ, and were then incubated at 30°C for 15 min unless otherwise indicated. Reactions were quenched in the presence of a 10‐fold excess of unlabelled oligonucleotide with the same sequence as the labelled strand to prevent reannealing, and the products were resolved on non‐denaturing 12% (19:1 acrylamide:bisacrylamide) polyacrylamide gels. Unwinding was quantified as described previously (Gupta et al, 2005).

References

Download references