DNA double-strand breaks (DSBs) create harmful lesions that are typically generated in response to extrinsic sources like ionizing radiation (IR) and chemotherapeutic drugs or intrinsic sources such as DNA replication fork collapse, transcription, and oxidative stress. If the DSBs are not efficiently joined, unrepaired DNA ends could lead to severe consequences such as cell cycle arrest, apoptosis, or senescence. On the other hand, if the DSB ends are joined improperly, mis-repaired DNA ends could lead to abnormalities such as genomic instability, chromosome translocation, and subsequent carcinogenesis (1–5). In eukaryotic cells, two major pathways are responsible for the repair of DSBs: homologous recombination (HR) and classical non-homologous end-joining (cNHEJ). HR uses an intact DNA as a template to accurately repair the breaks to keep the fidelity of the genome. HR is considered to be mediated by a couple of factors including the single-strand binding protein RPA and the human homologs of Saccharomyces cerevisiae Rad51 (6). Due to the requirement of the sister chromatid or a homologous chromosome that provides the template, HR is generally restricted to late S or G2 cell cycle (7). cNHEJ is the other major DNA repair pathway in eukaryotic organisms including humans (4, 8–14). Characteristically, cNHEJ directly joins broken ends together and does not require a homologous template; thus, it is not restricted to a certain phase of the cell cycle (15). Recently, the alternative end-joining (Alt-EJ) is emerging as another important repair pathway that describes end joining events lacking cNHEJ factors.

The general mechanism of cNHEJ is the recognition of DSBs and bridging of the broken ends, assembly, and stabilization of different repair factors at the damaging sites, and the dissolve of these factors after the repair (4, 15). cNHEJ comprises core KU70 and KU80 subunits that form the KU heterodimer, which recognizes double-strand DNA ends and promotes the recruitment of downstream cNHEJ core factors, including the nuclear serine/threonine kinase DNA-dependent protein kinase catalytic subunit (DNA-PKcs), Artemis nuclease, X-ray repair cross-complementing protein 4 (XRCC4), XRCC4-like factor (XLF), and DNA ligase IV (LIG4). In recent years, several new factors were discovered to play crucial roles in cNHEJ such as modulator of retroviral infection (Mri) and Paralogue of XRCC4 and XLF (PAXX) (16).

Interestingly, DNA repair factor 53BP1, DNA damage response kinases ATM and DNA-PKcs, the histone variant H2AX, Snf2-family helicase-like ERCC6L2, the RNF8 and RNF168 ubiquitin ligases, the mediator of DNA damage checkpoint protein 1 (MDC1), and RAG2 were found to have redundant roles with cNHEJ factor XLF, even though single deficiency of either protein did not abolish cNHEJ (17–22). Once the DSBs are generated, the KU heterodimer and DNA-PKcs form the DNA-PK holoenzyme that associates with the DNA ends. DNA-PK autophosphorylates DNA-PKcs to license further ligation activities (23, 24). Artemis nuclease is essential for opening loops, bubbles, flaps, gaps, and hairpins, which are obstacles for ligation (25–27). Activation of Artemis requires both the DNA-PKcs protein and the kinase activity of either DNA-PKcs or ATM (24). XRCC4 and PAXX are requisite for the stabilization and activation of LIG4, which essentially ligates the DNA ends (28–30). XLF stimulates the ligase activity by interacting with XRCC4-LIG4 (31), while Mri prevents components of HR or A-EJ pathway from access of broken DNA ends (32).

Out of these core NHEJ factors, PAXX is relatively a new member of XRCC4 and XLF-superfamily that possess an identical structure of N-terminal domain. It is widely distributed in all human body with low tissue specificity, despite that it is highly expressed in the blood dendritic cells and lymphocytes such as B and T cells, according to the database of The Human Protein Atlas. In addition, Paxx gene exists in all vertebrates but not in most invertebrates or yeast. This evolutionary distribution of Paxx is different from Xrcc4 or Xlf that is conserved from yeast to vertebrates (33). However, Paxx seems to have co-evolved with Prkdc gene that encodes DNA-PKcs protein and Dclre1c gene that encodes Artemis protein, which are restricted to higher eukaryotes (34), suggesting that PAXX may function in the same manner as DNA-PKcs and Artemis, connecting complex rather than simple DNA ends (35). On the basis of experimental data, it has been proven that even though PAXX itself does not bind DNA, it directly interacts with the KU complex and promotes KU-dependent DNA ligation (33, 36–38). However, the underlying mechanism how PAXX associates with cNHEJ still needs to be further explored.

The human PAXX protein contains a total of 204 amino acids. Seven beta-strands (S1–S7) and three helixes (H1–H3) are observed in the N-terminus of PAXX protein ranging from the 1st to 145th amino acid, in which a spherical head domain and a coiled-coil stalk make the protein primarily as a dimer in solution. This structure is similar to the dimeric interface formed by XRCC4, XLF, and SAS6 (33, 39–43) ( Figure 1 ). Interestingly, despite that PAXX is called “Paralogue of XRCC4 and XLF,” it shows more structural similarities with XRCC4 than XLF. In the case of XLF, the C-terminus of the XLF is folded back to the junction between H3 and the head domain, forming a 90°C angle between them. However, the angle between the helix H3 and the head field of XRCC4 and PAXX is about 45°C due to lack of tension. Different from XRCC4, PAXX has a much shorter coiled-coil stalk. Since LIG4 binding requires an extended stem of XRCC4, it implies that PAXX protein may not be able to directly interact with LIG4 (33). Besides, the C-terminal domains (CTRs) of the XRCC4, XLF, and PAXX superfamily are more diverse with different functions. Deletion (AA1–170) or point mutation (RRR177–179AAA, IN186–187AA, or F201A) of the PAXX CTRs resulted in loss of its interaction with KU. An electrophoretic mobility shift assay (EMSA) showed that V199A/F201A mutants of PAXX abrogated its binding with Ku molecules in vitro. In addition, the F201A mutation compromised the nuclear localization of PAXX. These results suggested that the C-terminus of the PAXX might be the regulatory region of the protein, with the underlying mechanism to be further explored (33, 43).

Interestingly, hypomethylation of Paxx gene promoter and higher expression level of mRNA were observed in tumors compared with normal tissues, implying that overexpression of PAXX might be correlated with tumorigenesis. In summary, given the importance of this new cNHEJ factor, we will discuss the role of PAXX in DNA repair and its biological functions in lymphocyte development and tumorigenesis in this review.

To explore the role of PAXX in DNA repair, PAXX-deficient chicken DT40 cells were treated with different sources of DNA-damaging agents, including irradiation (IR), bleomycin, etoposide (VP16), ICRF193, and camptothecin (CPT). Cell survival assay showed that PAXX deficiency increased sensitivity to most of these stimuli except for CPT (33), the inhibitor of DNA topoisomerase I. Human colon cancer cell line HCT116 in which PAXX was depleted using CRISPR-Cas9 technology or osteosarcoma cell line U2OS that was treated with small interfering RNA (siRNA) against PAXX also showed hypersensitivity to IR treatment and delayed γH2AX foci resolution (43). Mouse B-lymphocyte cell line CH12F3 in which PAXX was knocked out was hypersensitive to zeocin, but its class switch recombination (CSR) efficiency was not affected (44). Among these agents mentioned above, CPT induces replication-dependent DNA damage that relies on HR to repair (45). In contrast, repair of ICRF193 caused DNA damage replies more on cNHEJ rather than HR pathway. In addition, G1 arrested Paxx ^−/− cells were more sensitive to IR, compared with unsynchronized cells, suggesting that PAXX played a unique role in G1 phase. Moreover, PAXX-deficient U2OS or human retinal pigment epithelial RPE-1 cells were defective in random-plasmid integration, which was dependent on cNHEJ (43). All this evidence suggests that PAXX plays an essential role in cNHEJ, and we will further discuss the mechanism that PAXX plays roles in DNA repair in the next few paragraphs.

As discussed extensively above, PAXX contributes to cNHEJ by interacting with several cNHEJ-related factors and promotes DSB repair through cNHEJ pathway (33, 43). Mouse model also confirmed that PAXX played a critical role in cNHEJ that was masked by XLF (36, 47). In-depth study showed that PAXX and XLF play distinct roles in cNHEJ (36, 47, 49, 53, 54).

To find out whether PAXX played a role in other parallel DNA repair pathways independent of cNHEJ, such as HR or base excision repair (BER), Yang et al. used plasmid-based cNHEJ and HR reporter assays to detect the cNHEJ and HR efficiency. They observed that in PAXX-deficient U87 cells, HR efficiency was increased by 18%–21% (75). However, since deficiency of cNHEJ factors could significantly enhance HDR (76), whether PAXX plays a direct role in HR still needs to be explored.

Interestingly, in a temozolomide (TMZ)-resistant U87 cell line, PAXX protein level was increased, implying that PAXX might contribute to TMZ resistance in glioma cell line (75). TMZ can alkylate DNA and add a methyl group to purine base to introduce O6-methylguanine (O6-MeG) specifically (77). Notably, over 80% of DNA damage caused by TMZ are N-methylated bases that are substrates for BER pathway. This means that BER may be responsible for TMZ resistance, at least partially, but importantly (78, 79). Thus, the increased expression levels of PAXX in TMZ-resistant cells indicated that PAXX might contribute to TMZ resistance via BER pathway. In order to verify this conjecture, a GFP-based reporter assay was used to detect BER efficiency in PAXX-deficient cells, and results showed that the BER efficiency was reduced by 50% in PAXX-deficient U87 cells. Previous studies had identified that polymerase beta (pol β) was crucial for BER pathway, and knocking out pol β caused significant BER reduction (78, 80). Consistently, PAXX interplayed with pol β in vitro, indicating that PAXX might contribute to BER via interacting with pol β, which conduced TMZ drug resistance in glioma cells (75).

Interestingly, HAP1 cells that contained one single copy of almost every human chromosome knocked out by PAXX were less sensitive to DNA damage reagents such as zeocin and etoposide, with no significant change in genome stability (44). These results suggested that PAXX may have an extremely complex regulatory mechanism under physiological conditions in different cell lines.

In summary, these findings enlarge the dimensions of the PAXX functions: it not only plays an important role in cNHEJ but also contributes to BER and would not be surprising in other repair pathways such as MMEJ. Thus, the mechanism that PAXX plays roles in these cNHEJ-independent repair pathways requires to be further determined ( Figure 3 ).

cNHEJ is essential for maintaining the genomic stability and the development of the mammalian immune system. The deletion or mutation of the cNHEJ factors causes abnormal V(D)J recombination or CSR, leading to defects in lymphocyte development. Defects in cNHEJ factors were also observed in human patients. For example, patients with XLF deficiency usually have microcephaly and growth delay (81). Artemis null-deletion patients lacks T and B cells, and patients with abnormally low expression have spontaneous EBV-associated leukemia and progressive immunodeficiency (82, 83). Patients with DNA-PKcs deletion have microcephaly, developmental delay, and postnatal neurodegeneration (84).

To assess the physiological functions of PAXX in mammalian development, several groups including us generated PAXX-deficient mice. Unexpectedly, PAXX-deficient mice were viable and grew normally. Knocking out of Paxx gene does not affect the weight, size, fecundity, or development of the central nervous system (47, 85). Besides, the immune systems of the mice have been extensively analyzed. The spleen, thymus, bone marrow, and lymphocyte development showed no significant difference with that of WT mice, in terms of the organ appearance and flow cytometry analysis of the surface markers (36, 47, 85). In addition, no significant role of PAXX was observed in genome stability maintenance, V(D)J recombination, or CSR in PAXX-deficient mice (47, 85).

Since double deficiency of two cNHEJ factors or one cNHEJ factor with other DNA-repair-associated factor in mouse models exhibited both synthetic viability (KU and LIG4) or synthetic lethality (XLF and H2AX) (86); Balmus et al. and other labs crossed the Paxx ^−/− mice into ATM-, KU80-, LIG4-, or XLF-deficient backgrounds. Paxx^−/−Atm^−/− mice were born at frequencies similar to Atm ^−/− mice and did not show any overt phenotypes compared with Ku80^−/− mice. Similarly, Paxx^−/−Ku80^−/− mice displayed no additional phenotypes compared with Ku80^−/− mice. In addition, after multiple rounds of breeding, no viable Paxx^−/−Lig4^−/− offspring was observed, indicating that PAXX deficiency cannot rescue the embryonic lethality caused by LIG4 depletion. These results suggested that deficiency of PAXX was epistatic with ATM, KU, and LIG4 deficiency in mammalian development (36). However, it is not quite clear whether PAXX has redundant functions with other DNA repair factors such as 53BP1, given the fact that both genes located on the same chromosome in mice. Surprisingly, PAXX and XLF double-deficient mice are lethal during embryonic development (36, 47). Xlf^−/−Paxx^−/− embryos were observed at the expected Mendelian ratios in E14.5, but the size was smaller than the littermate controls. However, by E18.5, the Xlf^−/−Paxx^−/− embryos were no longer obtained at the expected rates. A few survival embryos showed reduced body weight and smaller spleen and thymus. Similar to XRCC4-deficient mice that showed severe neuronal apoptosis, the brain of the Xlf^−/−Paxx^−/− mice contained significant increased number of apoptotic inclusions in the post-mitotic intermediate zone, which might be the cause of the embryonic death. Consistently, in both E10.5 and E14.5, an increase in the number of γH2AX-positive cells was observed in the central nervous system (CNS) of the Xlf^−/−Paxx^−/− mice, compared with the WT littermates. These results revealed a critical role of PAXX in the survival neurons and embryonic development in XLF-deficient mice. Recently, Castañeda-Zegarra et al. showed that p53 deficiency rescued embryonic lethality of Xlf^−/−Paxx^−/− mice, the Xlf^−/−Paxx^−/− p53^−/− mice lived up to 60 days and died for unknown reasons without tumors. In addition, Xlf^−/−Paxx^−/− p53^+/− mice were also live-born and possessed reduced body weight, reduced size of spleens and thymus, and severe lymphocytopenia (87). Conditional knockouts may help us to further study the physiological function of PAXX in specific tissues. In fact, Musilli et al. conditionally knocked out XLF in hematopoietic stem cells or mature B cells in PAXX knockout mice and confirmed that PAXX and XLF are redundant in the V(D)J recombination, but not in CSR (88). This result was consistent with the results performed in the CH12F3 B-cell lymphoma cell line based CSR assay showing that PAXX was dispensable for CSR cNHEJ in the presence or absence of XLF (53).

In summary, present data suggested a very important role of PAXX in mouse neuron and lymphocytes development that was masked by XLF.

cNHEJ is required for V(D)J recombination during the lymphocyte development and is important for class switch recombination (CSR) process (89, 90). Deletion of cNHEJ factors such as LIG4, XRCC4, DNA-PKcs, and Artemis leads to defects in V(D)J recombination, resulting in severe combined immunodeficiency (SCID) (91). In contrast, PAXX-deficient mice had normal V(D)J recombination and CSR, and the number and proportion of T- and B-lymphocytes are comparable to those of WT mice (47). 5′ RACE RT-PCR and next-generation sequencing revealed that PAXX mice had a well-diversified TCRα repertoire, comparable to WT mice (48).

However, Xlf^−/−Paxx^−/− embryos exhibited a 10-fold decrease in thymocyte number, mainly accounted for the CD4⁺CD8⁺ double-positive (DP) cells. Analysis of the existing CD4⁻CD8⁻ double-negative (DN) thymocytes showed that the CD44⁻CD25⁺CD28⁺ cells, which had undergone successful rearrangement of the TCR-β locus, were significantly reduced in Xlf ^−/− Paxx ^−/− E18.5 embryos. Consistently, PCR assay that detected TCR-β rearrangements involving Vβ10 and the Dβ2-Jβ2 cluster was performed, and no products were found in DNA from Xlf ^−/− Paxx ^−/− thymocytes. This observation indicated an impairment of V(D)J recombination at the TCR-β locus in thymocytes from Xlf ^−/− Paxx ^−/− embryos.

As for B-cell development, V(D)J recombination took place in the fetal liver (FL) at around E17.5 at which time μH chain became detectable after successful rearrangement (92). Whereas ~20% of pro-B cells expressed the intracytoplasmic μH chain (iIgM) in FL of littermates Xlf^+/− Paxx^+/− embryos, only approximately 5% of this population was observed in the Xlf ^−/− Paxx ^−/− embryos, implying major defects in rearrangements (48). Conditional knockout of Xlf in Paxx^−/− mice showed similar results that recapitulate the profound immune deficiency caused by a block in the V(D)J process noted previously in E18.5 PAXX/XLF double knockout fetuses (88). Not surprisingly, CD21-Cre-mediated deletion of Xlf in Paxx^−/− mature B cells showed a similar CSR defect compared with Xlf^−/− B cells, excluding the role of PAXX in this particular stage during B-cell development.

Aberrant DNA damage repair is a major cause of genomic instability, and the latter is the most common feature of human tumor formation. Therefore, it is reasonable to ask whether PAXX is associated with cancer genesis. Unfortunately, we and other groups did not observe early spontaneous tumor development in Paxx^−/− mice (47, 88). Consistently, lymphocytes and MEF cells from PAXX-deficient mice did not show severe genomic instability. However, although PAXX itself does not seem to play a role in tumor suppression, it is still possible that loss of PAXX might reduce the fidelity of the end-joining products of cNHEJ, which in turn may cause abnormal gene expression and promote cancer development in the late stage of the lifespan. Furthermore, PAXX-deficient cells were more sensitive to ionizing radiation. This might be because PAXX loss was prone to accumulate more harmful mutations, which may eventually lead to cell death, or evolved to the other extreme—immortality. Meanwhile, it is not clear whether PAXX deficiency would accelerate tumorigenesis in tumorigenesis models such as p53-deficient or KRAS mice.

PAXX single knockout has no significant effect on mouse development (85), but PAXX/XLF double knockout mice show severe genomic instability and neuronal cell apoptosis and eventually lead to embryonic death (48). Therefore, in PAXX knockout mice, XLF can rescue most of the biological defects caused by the loss of PAXX. Intriguingly, in colon cancer patients, Xlf expression was negatively correlated with PAXX expression. Hence, whether the role of PAXX in tumorigenesis is hidden in the presence of XLF is also to be explored, which could only be accessed by generating conditional knockout mouse models. In a mouse model generated in which Cre was expressed under control of the iVav promoter, Xlf was conditionally knocked out in a Paxx^−/− background, turned as XlfΔiVav mice. Mortality of the XlfΔiVav mice was sharply increased compared with the Paxx^−/− mice. In addition, XlfΔiVav mice showed decreased weight and moderate to severe alopecia. More importantly, three of five analyzed mice developed a thymic mass composed of cells that had larger than normal nucleus, showing the tumorigenic nature. Further flow cytometric analysis showed increased proportion of CD44+ CD62L+ cells with negative TCRβ+ staining. These observed results suggested that, at least in in hemopoietic system, PAXX prevents tumorigenesis in XLF-deficient background in aging mice.

It is also not clear whether the mutation of PAXX had some gain of function in tumorigenesis. This is not surprising since protein kinases ATM, DNA-PKcs, and ATR, which play essential roles in DNA repair, all have unexpected structural function (93, 94).

Surprisingly, different from what was predicted in mouse model, PAXX was found to be highly expressed in the TCGA colon cancer dataset due to the hypomethylation status of the PAXX gene promoter region in tumor cells. In addition, some studies have shown that PAXX was highly expressed in several drug-resistant cancer cells, such as osteosarcoma (OS) and glioma cells. Even though the underlying mechanism was not quite known, it was suggested that PAXX may be involved in some proto-carcinoma signaling pathways through its function beyond DNA repair. Meanwhile, there is a correlation between the overexpression of PAXX and the survival rate of patients, and the expression of PAXX may be used as a prognostic marker for disease-specific survival (95).

At present, multi-channel target combination therapy has received more and more attention and has achieved ideal clinical effects, such as poly(ADP-ribose) polymerase (PARP) inhibitors for the treatment of HR-deficient tumors (96, 97). Interestingly, in OS cells that are resistant to doxorubicin or cisplatin, the expression of PAXX was upregulated, and when PAXX was knocked out or specifically inhibited, drug sensitivity was restored. In this scenario, PAXX might be used as a target to effectively promote the effects of chemotherapy drugs (95). In addition, for tumors with XLF mutations, we could inhibit PAXX,which has less damage to the body after knockdown, as adjuvant therapy. It can not only improve the lethality of radiotherapy/chemotherapy to tumors but also effectively reduce the damage of radiotherapy/chemotherapy to normal cells, thereby achieving the effect of improving the sensitivity of tumor radiotherapy/chemotherapy and reducing toxic side effects. In addition, in view of the drug resistance of tumors with high expression of PAXX, whether the drug resistance can be reversed by finding suitable targets and designing combination drugs will play an important role in improving the efficiency of tumor treatment. Therefore, in combination with the biological functions of PAXX, the design of small molecule inhibitors targeting PAXX and the way of combination drugs are still scientific issues worth exploring.

JT and XL conceived and wrote the manuscript. JT, ZL, QW, MI, WL, and XL read and approved the final manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by the Guangdong Provincial Key Laboratory of Regional Immunity and Diseases (Grant Number 2019B030301009), National Natural Science Foundation of China (81972661), and high-level university phase 2 construction funding from Shenzhen University (860-00000210).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.