L1 expression and regulation in humans and rodents

Abstract

Long interspersed elements type 1 (LINE-1s, or L1s) have impacted mammalian genomes at multiple levels. L1 transcription is mainly controlled by its 5’ untranslated region (5’UTR), which differs significantly among active human and rodent L1 families. In this review, L1 expression and its regulation are examined in the context of human and rodent development. First, endogenous L1 expression patterns in three different species—human, rat, and mouse—are compared and contrasted. A detailed account of relevant experimental evidence is presented according to the source material, such as cell lines, tumors, and normal somatic and germline tissues from different developmental stages. Second, factors involved in the regulation of L1 expression at both transcriptional and posttranscriptional levels are discussed. These include transcription factors, DNA methylation, PIWI-interacting RNAs (piRNAs), RNA interference (RNAi), and posttranscriptional host factors. Similarities and differences between human and rodent L1s are highlighted. Third, recent findings from transgenic mouse models of L1 are summarized and contrasted with those from endogenous L1 studies. Finally, the challenges and opportunities for L1 mouse models are discussed.

2. INTRODUCTION

Mammalian genomes are replete with transposable elements. Long interspersed elements type 1 (LINE-1s, or L1s) belong to the most abundant class of autonomous transposable elements in the sequenced human, mouse, and rat genomes, and account for roughly 17%, 20%, and 23% of their genomic DNA, respectively (1-3). Humans and rodents diverged about 75 million years ago. A comparative analysis of the age (i.e., the extent of sequence divergence) distribution of L1 sequences indicates that, in the past 25 million years, L1 activity has undergone significant decline in the human but not in the mouse and rat genomes (1-3). This overall decline of L1 activity in the human lineage is also evident by comparing the relative contribution of lineage-specific L1 sequences to each genome. Since the primate-rodent split, human-specific L1 subfamilies contribute to a mere 8% of the human genome as compared to over 16% by mouse-specific L1s (3). L1 activity is slightly higher in rat than mouse since the mouse-rat split (12–24 million years ago): rat-specific L1 subfamilies contribute to 12% of the rat genome, while 10% of the mouse genome is derived from mouse-specific L1s (1). However, highly active L1s do exist in the modern human genome. In the last 25 million years, five distinct human L1 families have emerged and generated more than 35,000 copies (4, 5). The youngest family, termed Ta (transcriptionally active), arose approximately 5 million years ago after the divergence of humans and chimpanzees. They have accumulated, per generation, at about the same rate in humans as the currently active L1 families in mice and rats (6). The distribution of polymorphic Ta1 elements in human populations suggests that the Ta1 subfamily continues to decrease the fitness of modern humans (7).

L1 transcription is mainly controlled by its unique, internal promoter (8). The promoter activity resides in the 5’ untranslated region (5’UTR) of a full-length L1 sequence (6–8 kb), i.e., the promoter sequence is transcribed (9-12). The 3’ untranslated region (3’UTR) contains a polyadenylation signal (13, 14). Two open reading frame proteins (ORF1p and ORF2p) are encoded by the L1 body, and both are essential for L1 retrotransposition (15). The vast majority of genomic L1 sequences are 5’ truncated (13, 14) and thus cannot be autonomously expressed. The number of full-length elements is in the order of 6,000 to 7,000 in human and mouse genomes (16-18). Among these, approximately 100 and 3,000 copies are estimated to be potentially active for retrotransposition in human and mouse genomes, respectively (19, 20). Proteins expressed from active full-length L1s preferentially mobilize the source template L1 mRNA via a phenomenon termed cis-preference (21-23). However, non-autonomous retrotransposons, such as human Alu elements (24) and mouse B1 and B2 elements (25), can be efficiently mobilized by L1 retrotransposition machinery. L1 can also generate processed pseudogenes but at a relatively lower frequency (21).

Non-LTR (long terminal repeat) retrotransposons are drivers of mammalian genome evolution (26). L1, as the sole autonomous member, exerts its impact on mammalian genomes at multiple levels. First, existing L1 sequences can profoundly modulate the quantity and quality of mammalian gene expression (27). On the genome scale, there exists an inverse correlation between gene expression level and the amount of L1 sequence present (28). Human genes carry an average of 8 kb of L1 sequences, but the amount of L1 sequence is drastically different among highly and poorly expressed genes (approximately 3 kb and 14 kb, respectively) (28). Second, the expression of L1 proteins causes multiple types of toxicity in cultured cells (29-32). For example, ectopic expression of a full-length L1 results in high levels of double-strand DNA breaks (DSBs) (30, 31), cell cycle arrest at G2/M (30, 31), induction of apoptosis in p53-competent cancer cells (31, 32), and induction of a senescence-like cellular state (29). The endonuclease domain of ORF2p is critical for DSB formation (30), but both endonuclease and reverse transcriptase domains of ORF2p contribute to deleterious effects on cellular proliferation (29). Third, L1-mediated insertional and post-insertional mutagenesis is responsible for a wide spectrum of human diseases (33, 34). While the majority of observed L1 cases are germline events, two somatic L1 insertions have been reported (35, 36). The first case is an insertion in the intron of one proto-oncogene myc allele; the insertion is tumor-specific because it is only found in the ductal adenocarcinoma sample but not in normal breast tissue of the same patient (36). The second case is an insertion that disrupts the last exon of the tumor suppressor gene APC in a colon cancer. In contrast to the myc case, the insertion in the APC gene bears all hallmarks of de novo L1 insertion and is very likely the causative event (35). More recently, several somatic L1 insertions have been found in human lung cancer genomes through next-generation sequencing but their contribution to tumorigenesis is unknown (37).

In this review, we examine L1 expression and its regulation in the context of human and rodent development. In Section 3, endogenous L1 expression patterns in three different species—human, rat, and mouse—are compared and contrasted. A detailed account of relevant experimental evidence is presented according to the source material, such as cell lines, tumors, and normal somatic and germline tissues at different developmental stages. In Section 4, factors involved in the regulation of L1 expression at both transcriptional and posttranscriptional levels are discussed, including transcription factors, DNA methylation, PIWI-interacting RNAs (piRNAs), RNA interference (RNAi), and posttranscriptional host factors. Similarities and differences between human and rodent L1s are highlighted. In Section 5, recent findings from transgenic mouse models of L1 are summarized and contrasted with those from endogenous L1 studies. Finally, the challenges and opportunities for transgenic L1 studies are discussed.

3. ENDOGENOUS L1 EXPRESSION

The expression of endogenous L1 RNA and proteins is spatiotemporally regulated during animal growth and development. Our understanding of its spatial and temporal expression patterns originates mostly from studies of endogenous human, rat, and mouse L1 sequences. In this section, we will provide an overview of experiments that profile endogenous L1 expression in the form of L1 RNA and/or proteins in cell lines and tissues. Two different techniques, i.e. reverse transcriptase-polymerase chain reaction (RT-PCR) and Northern blot analysis, have been featured in L1 RNA detection studies. RT-PCR is a very sensitive method and has been frequently used to quantify the level of L1 RNA. However, Northern blot analysis has the advantage in discerning full-length L1 transcripts from abundant read-through or truncated transcripts. Further, the interpretation of RT-PCR can be confounded by L1-related RNA species resulting from premature polyadenylation (28, 38), aberrant splicing (39), and cellular transcripts from an antisense promoter located in the 5’UTR (40, 41). Thus, an emphasis will be placed on studies that provide evidence for full-length L1 transcripts.

3.1. Endogenous human L1 expression

The examination of human L1 expression is limited by the lack of accessibility for normal tissue samples, and thus most work has focused on L1 expression in cell lines and tumor biopsies, such as germ cell tumors (GCTs), GCT-derived embryonal carcinoma cell (ECC) lines, and breast cancers. Recently, fetal/adult testes and a panel of poly (A)+ RNA from normal tissues have also been evaluated for endogenous L1 expression. Overall, L1 expression is predominantly detected in tumors, and cell lines derived from malignant tumors. Additionally, L1 expression is detected by immunohistochemistry in both germ cells and selected somatic cells of human fetal and adult testes, and by Northern blot analysis from selected normal tissue RNA samples.

3.1.1. Human cells lines

L1 expression is prevalent in GCT-derived ECC lines, such as 2102Ep, GH, JEG-3, NCCIT, NTera2D1, and PA1 (42-45). They are derived from two types of GCTs: teratocarcinoma and choriocarcinoma (GCTs are detailed in Section 3.1.2.). However, the first glimpse of potential full-length L1 transcripts was provided by gradient sedimentation of the cytoplasmic RNA from Jurkat cells, an immortalized human T lymphocyte cell line (46). Later, a discrete 6.5 kb L1 RNA signal was detected in the cytoplasmic fraction of a human teratocarcinoma cell line (NTera2D1) (44). This RNA species only hybridizes to antisense probes, is polyadenylated (44), and has been subsequently characterized at the sequence level (47). Most interestingly, its expression is regulated by the cellular differentiation process as evidenced by a strong correlation between its abundance and the differentiation status of these cells: the 6.5 kb L1 RNA signal is present at high levels in high-density NTera2D1 culture when the cells display the characteristic appearance of ECCs, and is absent when cells are differentiated by retinoic acid (44). Full-length L1 transcripts are also detected in additional teratocarcinoma cell lines, including NCCIT, GH and PA1 (42). The functional relevance of these full-length L1 transcripts is corroborated by the detection of L1 protein products. Western blot analyses with a polyclonal antibody reveal the presence of human L1 ORF1p in two teratocarcinoma cell lines (NTera2D1 and 2102Ep) and one choriocarcinoma cell line (JEG-3) (43). This antibody has served as a key reagent in numerous subsequent studies on L1 expression in both cell lines and formalin-fixed tissues (Sections 3.1.2., 3.1.3., and 3.1.4.).

Besides cell lines derived from GCTs, several other tumor cell lines also show expression of L1 RNA and/or ORF1p. At least seven breast cancer cell lines have been tested positive for ORF1p by Western blot analysis, including three infiltrating ductal carcinoma cell lines, three adenocarcinoma cell lines, and one carcinoma cell line (48, 49). Recently, a full-length L1 signal has also been detected in the poly (A)+ RNA fraction from a cervical adenocarcinoma cell line (HeLa; detailed in Section 3.1.4.) (50). However, L1 expression is not restricted to tumors of epithelial cell origin as previously claimed, because ORF1p can be detected in cell lines derived from human bladder carcinoma (J82), colon carcinoma (K12M), melanoma (A375), and fibrosarcoma (HT-1080) (49). Two other bladder carcinoma cell lines (HT1376 and 5637) and one T-cell lymphoma cell line (Molt4) show full-length RNA transcripts, albeit at a much reduced level as compared to teratocarcinoma cell lines (43).

In contrast to tumor cell lines, human embryonic stem cells (ESCs) are the only non-transformed human cell lines displaying endogenous L1 expression. ESCs are derived from the inner cell mass of blastocyst-stage embryos, and are pluripotent in that they can differentiate into cell types of all three primary germ layers. ESCs represent the normal counterpart of ECCs. Undifferentiated human H9 ESCs show levels of ORF1p comparable to ECCs (51).

3.1.2. Human germ cell tumors

Human GCTs are a heterogeneous group of neoplasms with germ cell origin. They mainly occur in gonads and in the midline of a human body. These locations represent the primordial germ cell (PGC) migration route from the yolk sac to the genital ridge (52). Several histotypes are observed in the testis: teratomas and yolk sac tumors are often found in neonates and infants, seminomas and nonseminomas in adolescents and adults, and spermatocytic seminomas in the elderly. Histologically, seminomas are homogeneous tumors. In contrast, nonseminomas are often heterogeneous, and can contain different histological elements, including embryonal carcinoma (the undifferentiated stem cell component), teratoma (the differentiated somatic component), yolk sac tumor and choriocarcinoma (differentiated extraembryonic components) (52). The precursors of infant teratomas and yolk sac tumors are probably early PGCs or gonocytes that carry partially-erased genomic imprinting. These precursors are arrested before meiosis I and become pluripotent embryonic germ cells (EGCs) after reprogramming. Both seminomas and nonseminomas originate from neoplastic PGCs/gonocytes, which have undergone complete erasure of genomic imprinting. These precursors occupy the spermatogonial niche of the seminiferous tubule in adult testes, and their default pathway is to form seminomas. Nonseminomas result from the reprogramming of a seminomatous tumor cell to a pluripotent ECC (the malignant counterpart of EGC) (53).

Several studies have examined ORF1p in adult and pediatric GCTs (54-56). Earlier results indicate that the prevalence of ORF1p-positive cells is approximately 5-10% in both adult and pediatric germ cell tumors (59 and 58 cases, respectively) (54, 55). However, a more recent study has detected variable ORF1p expression in all 162 cases of pediatric germ cell tumors (56). Among them, 18.5%, 37.7%, and 43.8% show strong, moderate, and weak immunoreactivity to ORF1p antibody, respectively. The differences between these studies may be explained by the more sensitive immunohistochemistry detection method adopted in the latter study (56); it is likely that only cells expressing high levels of ORF1p were detectable in the earlier studies (54, 55).

Some common themes emerge from these studies regardless of the methodological differences. First, strong ORF1p expression is frequently detected in embryonal carcinoma, choriocarcinoma, and yolk-sac tumors, but not in teratoma and germ cell tumors without embryonal carcinoma and yolk sac tumor components (55). Second, embryonal carcinoma or yolk sac tumor cells are the reactive component in all positive cases for both adult and pediatric germ cell tumors (54, 55). Normal cells within the testicular tumor mass do not react with ORF1p antibody (54). Normal testicular tissue adjacent to tumors show negative or weaker ORF1p expression than tumor cells (56). Third, most immunoreactive cells appear to be undifferentiated epithelial cells. Multivariate analyses show that strong expression of ORF1p is associated with poor differentiation, with extragonadal sites, and with the yolk sac tumor histotype (56). Fourth, metastases to the lung and lymph nodes display an ORF1p expression profile similar to that seen in primary tumors (54). Lastly, it is noted that an immunohistochemical screening of over twenty different normal tissues fail to detect any ORF1p signal (54).

3.1.3. Human breast cancers

ORF1p is also expressed in a significant number of human breast cancers (48, 57, 58). In one study, nine out of twelve infiltrating ductal carcinomas are positive for ORF1p by immunohistochemistry, but nonmalignant breast epithelial cell lines and normal breast tissues are all negative (48). In a separate study, the epithelia of all twelve breast cancers express ORF1p by immunohistochemistry, although one tissue is very weakly reactive (57). In contrast, the epithelia of four out of six nontumor tissues adjacent to matched cancers are negative or very weakly positive (57). For most positive cases, the ORF1p signal is heterogeneous among tumor cells, varying from negative to strongly positive (48, 57). Interestingly, Western blot analysis appears to be much more sensitive than immunohistochemistry in detecting ORF1p from breast tissues. Western blot analysis, with the same ORF1p antibody, detects a specific signal from all human breast tissues examined, including tumors, matched nontumor breast tissues, and reduction mammoplasty tissues. Nonetheless, matched nontumor breast tissues and reduction mammoplasties have significantly lower levels of expression than primary tumors (57). In general, there is no correlation between ORF1p levels and clinicopathologic outcomes (57, 58), but nuclear localization of L1 ORF1p is frequently associated with a poor prognosis of breast tumors (58).

3.1.4. Human somatic and germline tissues

Due to the lack of an effective ORF2p antibody, the examination of ORF2p expression was not possible until very recently (59). Using immunohistochemistry, both ORF1p and ORF2p are found in fetal and adult testes (the latter are from orchiectomized patients with prostate cancer). Germ cells positive for both ORF1p and ORF2p include gonocytes from fetal testes, secondary spermatocytes, and immature spermatids with residual bodies from adult testes (59). In addition, L1 proteins are found in some somatic cells, including Leydig cells in both fetal and adult testes and Sertoli cells of the adult testis. L1 proteins are also found in vascular endothelial cells of both fetal testis and adult epididymis and in syncytiotrophoblasts of the placenta (59). Similar but slightly differing findings were previously reported in mouse studies (60, 61) (discussed in Sections 3.3.2. and 3.3.3.). In contrast to males, L1 protein expression in female germ cells has not been reported. Recently, using an RNA-fluorescence in situ hybridization technique, L1 transcripts have been detected in human immature diploid oocytes from in vitro fertilization donors (62). However, it is unclear whether such transcripts are full-length.

L1 expression in normal human tissues has recently been evaluated at the RNA level by a sensitive Northern blot protocol (50). This protocol was previously used to reveal full-length, as well as variably processed, L1 transcripts from two cancer cell lines that have shown evidence for ORF1p expression (MCF7 and SK-BR-3) (39). A similar L1 RNA expression pattern is demonstrated in HeLa cells, although a higher proportion of L1 signals in HeLa cells are from processed RNA species as compared to MCF7 cells (50). Importantly, the improved sensitivity and the rational design of 5’UTR probes enable the detection of specific spliced L1 transcripts. Two of such spliced transcripts are predicted to encode a functional ORF2p (39). When the spliced 5’UTR sequences are cloned between an SV40 promoter and an active ORF2 sequence, the expressed product displays functions that are expected from an authentic ORF2p (50). For example, their overexpression supports Alu mobilization in HeLa cells and induces DNA damage and a senescence-like phenotype in normal human fibroblasts (50). However, the most provocative finding comes from the examination of L1 RNA in a poly (A)+ RNA panel of normal human tissues (50). Full-length L1 RNA is readily detected in several tissues, including esophagus and placenta, at a level comparable to that found in HeLa cells. Processed L1 RNA species are detected in nearly all tissues examined. The highest level of L1-related transcripts is found in the testis, but interestingly it lacks a discernable full-length L1 signal (50). In addition, the bulk of these “processed” L1-related RNA species are significantly shorter than the spliced ORF2 transcripts. This L1 RNA profile in the human testis sample is reminiscent of that in the mouse testis (Section 3.3.2.). The identity and functional significance of these short L1 RNAs warrant further investigation.

3.2. Endogenous rat L1 expression

There are only a limited number of studies on L1 expression in rats and rat cell lines. An elevated level of ORF1-containing transcripts is observed in rat neural progenitor cells (NPCs) by microarray and RT-PCR studies (63). A peculiar pattern of L1 expression is found in rat chloroleukemia (RCL) cells. Chloroleukemia is an uncommon condition in both humans and rats. Rat chloroleukemia can be experimentally induced by gastric instillation of a chemical carcinogen, methylcholanthrene (64). The chloroma cells consist of an immature population of myeloblasts, promyelocytes, and neutrophilic myelocytes (65). These cells represent the first three morphologic stages in the development of mature neutrophils, and, in contrast to later stages of neutrophil development, are able of self-renew. They can be maintained by repeated transplantation into young rats, or in suspension cell culture. It has been reported that L1 transcription is suddenly activated when RCL cells reach a critical population density (i.e., at approximately half of the maximal population density) (66). A ten-fold increase of cytoplasmic L1 transcripts over the course of four hours is detected by an RNA dot blot analysis (66), although it is not clear to what extent full-length L1 transcripts contribute to such an increase. Cells continue to grow beyond the critical population density while L1 transcript levels remain high (66). However, in less than thirteen hours, cell growth is inhibited, and all cells die within two days due to apoptosis. DNA dot blot analysis suggests a sudden incorporation of about 300,000 L1 copies into the genome of dying cells. Interestingly, this growth inhibition can be released by subculturing. If subculturing is performed at or near the maximal cell density, the level of L1 transcripts quickly return to the baseline level typical of the exponential growth phase. If the subculturing is delayed after reaching the maximal density, the level of L1 transcripts does not decrease (66). In addition to cell density, ultraviolet and ionizing radiation can induce rapid, premature activation of L1 during the exponential growth phase (67). In fact, discrete, full-length, and poly (A)+ L1 transcripts can be readily detected from ultraviolet-irradiated RCL cells as well as RCL cells reaching maximal population density; one such complementary DNA clone is competent in retrotransposition in cell culture (68). L1 expression in RCL cells has also been confirmed at the protein level by Western blot analysis of both ORF1p and ORF2p. In particular, two ORF1 polypeptides are detected, and they correspond to the predicted molecular masses of two ORF1p classes, due to differences in the length polymorphism region (LPR) (68). The molecular mechanism underlying the observed L1 expression dynamics has not been explored. Nevertheless, RCL cells represent a unique case that involves drastic changes in L1 transcription and retrotransposition activities.

Very little is known about L1 expression patterns in normal rat tissues. The presence of L1-related transcripts in nuclear RNA from rat liver, kidney, and a rat neural cell line was initially revealed by reverse Northern blot analysis, in which radio-labeled rat RNA is used as a probe and hybridized to restriction endonuclease-digested genomic DNA (69). Recently, the endogenous expression of rat L1 proteins has been detected in rat heart tissue maintained ex vivo (70). ORF1p is localized predominantly in the nucleus of cardiomyocytes, and its level increases in response to ischemia (70). The antibody against ORF2p detects a strong signal exclusively in the endothelial cells of the heart (70); however, this could simply be an artifact as the same antibody cross-reacts strongly to HSP90 in Western blot analysis (68).

3.3. Endogenous mouse L1 expression

Human and mouse L1 sequences are discovered in tandem (71), and many seminal discoveries about L1 structure and function originate from studies on mouse L1 retrotransposons. In fact, the mouse has served as an indispensable model for studying L1 expression in the past two and a half decades, and much we know about L1 expression patterns in vivo comes from mouse studies. Specifically, endogenous mouse L1 is expressed in different stages of mouse development in a tightly regulated manner.

3.3.1. Mouse cell lines

The initial evidence of full-length mouse A-type L1 transcripts was found in several mouse lymphoid cell lines (72). These transcripts are sense-strand specific and enriched in poly (A)+ cytoplasmic RNA fractions. Subsequently, full-length, sense-stranded mouse L1 transcripts were detected in two mouse teratocarcinoma cells: F9 (73, 74) and C44 (75). They are found to be associated with ORF1p in sucrose gradient fractions (73, 75). In addition, the younger A-type L1 transcripts are more abundant than F-type transcripts when hybridized with type-specific 5’UTR probes (74). Indirect immunofluorescence reveals a punctate distribution of ORF1p in the cytoplasm of F9 cells with a marked variation in the intensity of staining from cell to cell (75). Unlike F9 cells, C44 embryonal carcinoma cells are maintained in cell culture as embryoid bodies (EBs) (75). Normally, embryoid bodies are formed when ES cells undergo differentiation in cell culture. They are three-dimensional multicellular aggregates consisting of differentiated and undifferentiated cells. C44 cells form either simple or cystic EBs, which resemble mouse embryos at either the morula or blastocyst stage, respectively. Two distinct cell lineages are present in C44 EBs: the inner cells resemble the inner cell mass of a normal blastocyst and ECCs, and are the only cells positive for ORF1p; in contrast, the outer cells are histologically heterogeneous, with the majority reminiscent of endoderm cells. Similar to studies with human teratocarcinoma cells, no full-length RNA and ORF1p can be detected when F9 and C44 cells are differentiated by retinoic acid (75).

3.3.2. Mouse germ cells

The pursuit of L1 expression in normal mouse tissues started with an effort to detect full-length transcripts in mouse oocytes and blastocysts (76). In this study, a discrete 8 kb A-type transcript was found in mouse blastocysts (76). Two subsequent studies investigated the developmental and cell type specificity of L1 expression during mouse embryogenesis by examining postimplantation embryos, prepuberal and adult testes, and adult ovaries (60, 61). In the male germ line, L1 ORF1p is detected by immunohistochemistry in gonocytes of E15.5 (i.e., embryonic day 15.5), E16.5, and E17.5 embryonic testes but not in E14.5 or earlier stages (61). In addition, ORF1p is observed by immunohistochemistry in primary spermatocytes (only in leptotene and zygotene stages) of postnatal mouse testes from P10 (i.e., postnatal day 10), P14, P18, and P25 animals, and in round spermatids, elongating spermatids, and residual bodies of the adult mouse testis (60). Similar to the timeline of the male germ line, in the female germ line, L1 ORF1p is first detected at E15.5 in the primary oocytes and continues through E16.5 and E17.5 (61). However, no germ cell immunoreactivity is found in adult ovaries (61). Several recent studies have reexamined the expression of L1 ORF1p by indirect immunofluorescence with the same antibody and have extended their analyses to neonatal testes (P2, P6 and P10) (77). Similar to the early study, prominent ORF1p signals are detected in E14.5 and E16.5 wild-type embryonic testes; E18.5 is also positive but with a reduced signal. A minor ORF1p signal is observed in P2 gonocytes, and no ORF1p protein is present in P6 testes. A minor signal reappears in P10 testes (77). In adult testes, L1 ORF1p is detected in leptotene, zygotene, and pachytene stages of primary spermatocytes, but not in preleptotene and diplotene stages (78).

L1 ORF1p expression in mouse testes as demonstrated by immunohistochemistry has been confirmed by Western blot analysis. Additionally, Western blot analysis reveals both quantitative and qualitative details of mouse ORF1p expression during development from P10 to adult (60). Three ORF1p variants (42 kD, 43 kD, and 44.6 kD) appear in testis extracts in a developmentally regulated manner. Weak 42 kD and 43 kD ORF1p signals are detected at P10. Both signals are increased at P14. At P18, the 42 kD signal is no longer detectable, the 43 kD signal is at its peak level, and a faint 44.6 kD signal starts to appear. In both P25 and adult testis extracts, the 44.6 kD form is the predominant signal detected (60). Distribution of ORF1p variants has also been studied in specific cell fractions from the adult mouse testis. Spermatocytes at the pachytene stage display a weak 43 kD signal and a strong 44.6 kD signal. The 43 kD signal maintains its low intensity in round spermatids, but increases in elongating spermatids and peaks in residual bodies. In contrast, the 44.6 kD signal decreases rapidly in round spermatids and becomes undetectable in elongating spermatids and residual bodies. Consistent with round spermatids and elongating spermatids being the most abundant cell types in adult testes, homogenized seminiferous tubules show an ORF1p profile that is similar to round spermatids: 43 kD and 44.6 kD signals coexist, with the 43 kD signal being slightly stronger than 44.6 kD. Interstitial cells display a single strong 44.6 kD signal, equivalent to that detected in whole testis extract (60). Thus, it seems that different cell types in the mouse testis express distinct ORF1p variants. Studies of ORF1p expression in mouse teratocarcinoma cell lines indicate that members of A and T_F subfamilies of mouse L1 are responsible for the 42 kD and 43 kD signals, respectively (79). In vitro expression of the 44.6 kD ORF1p isoform has not yet been reported, but a third active L1 subfamily (20), termed G_F, is predicted to be its source. Thus, at least three mouse L1 subfamilies—A, G_F, and T_F—are being actively transcribed and translated during mouse development. How they are differentially regulated remains unexplored.

In general, there is a lack of correlation between full-length L1 RNA and ORF1p in male germ cells (60). A full-length L1 transcript can be readily detected in total RNA from P14 testes but not in any other developmental stages. Likewise, a specific 2.8 kb L1 RNA signal can only be detected in P25 and adult testes, but its identity and functional significance remain to be studied.

3.3.3. Mouse somatic cells

There is scant evidence that L1 is expressed in somatic tissues. The initial evidence is from immunohistochemical studies of gonads. Several somatic cells in the gonads are positive for mouse ORF1p, including Leydig cells (in E17.5 (61), prepuberal and adult testes (60)), myoid cells (prepuberal and adult (60)), but not in Sertoli cells (prepuberal and adult (60)) or Theca cells in the adult ovary (61). In addition, cyncytiotrophoblast cells of the placenta are ORF1p positive at all developmental stages examined (E13.5, E14.5, E15.5, E16.5, and E17.5) (61). A potential link between steroid hormone production and L1 expression has been suggested because, with the exception of myoid cells, all ORF1p-positive somatic cells (i.e., Theca cells, Leydig cells and cyncytiotrophoblast cells) are involved in active androgen production (61). Interestingly, a later study on L1 expression in corresponding human tissues reveals a partially overlapping set of somatic cells that express human ORF1p and ORF2p (Section 3.1.4.) (59). In light of the data from human gonads, this view has been revised since not all somatic cells are involved in androgen production; rather they all share receptors for steroid hormones, androgen and/or progesterone (59). The significance of this link remains to be determined.

4. REGULATION OF L1 EXPRESSION

Recent evidence suggests that L1 expression is regulated at multiple levels, both transcriptional and posttranscriptional. Mammalian L1 families have repeatedly acquired novel 5’ sequences during evolution (8). 5’UTR differs significantly in both sequence and organization between rodent and primate L1 families. In contrast to human L1 5’UTR, both rat and mouse L1s carry a varied number of monomers. As unique internal promoters, L1 5’UTRs from all three species—human (9, 12), rat (10), and mouse (11)—are necessary and sufficient in directing reporter gene expression in transfected cells. In this section, we evaluate the role of transcription factors, DNA methylation, PIWI-interacting RNAs (piRNAs), small-interfering RNA (siRNAs), and posttranscriptional host factors in L1 regulation.

4.1. Transcription factors

Several transcription factors that interact with the human L1 promoter have been identified. Human L1 5’UTR contains binding sites for YY1 (9, 80-82), RUNX3 (83), SOX proteins (63, 84), and p53 (85). The Yin Yang-1 (YY1) binding site is located close to the 5’ end of human L1 5’UTR on the antisense strand (+21 to +13) (80). Mutations in the YY1-binding site have only minor effects on L1 transcriptional activation; instead, the YY1-binding site functions as a component of the L1 core promoter to direct accurate transcription initiation and thus to maintain the integrity of the 5’ end of human L1 (82). Notably, the YY1-binding site is conserved in the 5’UTR of all extant full-length L1 sequences in the modern human genome (17). In addition, a consensus YY1-binding site is located in monomers of the mouse F and T_F subfamily, although it is mutated in the G_F subfamily (20). Human L1 5’UTR has three putative runt-domain transcription factor 3 (RUNX3) binding sites: the first two on the sense strand (+83 to +101 and +389 to +407), and the third on the antisense strand (+526 to +508). Only the first RUNX3-binding site promotes L1 transcription and retrotransposition in vitro. The third site may regulate the antisense promoter activity (83). Despite its functional significance, however, it is unlikely that RUNX3 dictates the tissue specificity of L1 expression, because RUNX3 is expressed in a wide range of tissues. In addition, a RUNX3-binding site is absent in the most active mouse L1 subfamily T_F (83). The tumor suppressor p53 is a key regulator in maintaining genomic integrity. Human L1 5’UTR contains a three-quarter p53-binding site (+452 to +467) and a half p53-binding site (+489 to +498) (85). L1 expression is increased in a reporter assay when p53 is overexpressed; only the three-quarter site is required for this p53 response. The expression of an integrated full-length L1 sequence is also increased in response to p53 activation (85).

It is puzzling as to how different L1 families achieve similar levels of cell-specific expression with divergent promoters. It is commonly believed that cell- and stage-specific epigenetic mechanisms play a pivotal role (Section 4.2.). Recent studies indicate that transcription factors of the SOX (SRY-related HMG box containing) family are involved in L1 transcriptional regulation at least during neurogenesis. The 5’UTR of human L1 contains two functional binding sites for a recombinant protein encompassing the SRY HMG box domain: +472 to +477 and +572 to +577 (84). Ectopic expression of SOX11 results in a 10-fold increase of L1 promoter activity not only in human embryonic kidney 293 cells using a luciferase-based reporter assay, but also in an embryonal rhabdomyosarcoma cell line when endogenous full-length L1 transcripts are measured (84). However, the physiological relevance of SOX11 in regulating L1 expression has yet to be explored. Another SOX family member, SOX2, represses L1 transcription in cultured adult rat hippocampus neural (HCN) stem cells (63). When HCN cells are differentiated toward neuronal lineage, SOX2 expression is down-regulated, which is concomitant with a transient stimulation of L1 transcription (63). Paradoxically, SOX2 has no effect on full-length 5’UTR of human L1 in a luciferase-based reporter assay. Instead, it stimulates reporter gene expression when the first 100 bp are deleted from L1 5’UTR, suggesting that other factors are involved in SOX2-mediated transcription repression of human L1 promoters (63). Adding another layer of complexity, both SOX-binding sites in human L1 5’UTR overlap with T-cell factor/lymphoid enhancer factor (TCF/LEF)-binding sites. TCF/LEF family transcription factors are effectors of the canonical Wnt pathway and can promote specific gene expression when, in association with beta-catenin, bound to their cognate binding sites. Indeed, the overexpression of either WNT3 or beta-catenin stimulates L1 promoter activity in neural stem cells (NSCs) using a reporter assay (86). Thus, it seems that SOX2 and TCF/LEF antagonistically regulate L1 promoter activity in NSCs. L1 promoters from all three species—human, rat, and mouse—contain at least one overlapping SOX2 and TCF/LEF-binding site (86), possibly as the result of convergent evolution.

4.2. DNA methylation and PIWI/piRNA pathway

DNA methylation is correlated with transcriptional gene silencing. It acts by directly impeding the binding of transcription factors to their target sites, and/or altering chromatin structure via histone modification and nucleosome occupancy within the promoter regions of genes (87). L1 5’UTR promoters from human (47, 88), rat (10, 89, 90), and mouse (both A-type (91, 92) and F-type (93)) are readily recognized as CpG islands. The correlation between DNA methylation and L1 expression has been substantiated by a number of studies. An active role of piRNA in de novo methylation of L1 sequences has also been revealed.

4.2.1. L1 methylation in cell lines and its correlation with L1 expression

The implication of DNA methylation in regulating L1 expression originates from multiple studies. Expression of L1 full-length transcripts is correlated with differential methylation of the promoter region in various cell lines (45): in teratocarcinoma cells and medulloblastoma cells that express abundant L1 protein, 5’UTR sequences are unmethylated; in somatic cells expressing little or no L1 protein, most or all are methylated (45). When DNA methylation is inhibited by 5-azacytidine or 5-aza-2’-deoxycytidine, a full-length L1 RNA signal becomes detectable in 3T3-derived A cells (94), and the level of 5’UTR-containing transcripts is increased in human embryonic fibroblasts and Ntera2D1 cells (95). Furthermore, individual L1 5’UTR sequences may be differentially methylated in different tumors. For example, the methylation status of nine individual 5’UTR sequences varies not only from 5’UTR to 5’UTR but also among different tumor types (96). The majority of these 5’UTR sequences are heavily methylated in teratocarcinoma-derived cell lines (2102Ep, NTera2D1, and PA-1), but only a varying subset are methylated in cell lines derived from non-germ cell malignancies and non-malignant breast tissues (96). In cell-based reporter assays, human (97) and rat (10) L1 promoter activity is significantly inhibited by CpG methylation. It is possible that not all CpG sites on 5’UTR are required for transcriptional inhibition, because only a subset of these CpG sites affect L1 promoter activity in an in vitro transcription assay when mutated (97).

DNA methylation mediates transcriptional silencing by impeding transcription factor binding and/or altering chromatin modification. In the latter case, methylated CpG sites are recognized by members of the methyl-CpG-binding domain (MBD) protein family, which in turn recruit histone deacetylases and generate a transcriptionally inactive chromatin structure (98). The effect of DNA methylation has only been shown for one of the L1-related transcription factors (Section 4.1.): YY1 binding is unaffected by promoter methylation in a gel shift assay (97). On the other hand, as a member of the MBD protein family, methyl-CpG-binding protein 2 (MeCP2) can bind to methylated L1 5’UTR and represses its transcription in reporter assays (99, 100).

4.2.2. L1 methylation in mouse tissues during development

Genomic methylation patterns undergo at least two cycles of reprogramming during development (101): in preimplantation embryos and during gametogenesis. In general, the genomes of mature sperm and oocytes are highly methylated. The paternal genome of the zygote is actively demethylated shortly after fertilization. Thereafter, the diploid genome is passively demethylated in a stepwise, DNA replication-dependent fashion up to the morula stage. Genomic methylation is restored by de novo methylation in a lineage-specific fashion in blastocysts: the inner cell mass is significantly more methylated than trophectoderm cells (destined to form extraembryonic tissues). PGCs originate from proximal epiblast cells at around E7.25, and undergo genome-wide demethylation as early as E8.0. A second wave of demethylation occurs around E11.5 when PGCs arrive at the genital ridge (termed gonocytes from then on). Subsequently, de novo methylation starts in the gonocyte genome and specific methylation patterns are reestablished. It should be noted that different classes of sequences might deviate from this general description of methylation dynamics (L1 as an example below).

L1 methylation was initially examined by using Southern blot and methylation-sensitive restriction enzymes (10, 102-106). This method surveys the methylation status of a limited number of CpG sites that overlap with the enzyme recognition sites. In addition, earlier studies employed a probe that corresponds to the body of mouse L1, and consequently, the dynamics of the observed methylation change may only reflect that of the body of endogenous L1 elements (102-106). The body of endogenous L1s is fully methylated in somatic tissues, such as brain (104), kidney (104, 105), liver (102, 103) and spleen (104, 106). Male and female germ cells differ in the degree of methylation. Mature sperm are fully methylated (102-106). In contrast, primary oocytes at the diplotene stage show little methylation (103, 106), secondary oocytes (ovulated but unfertilized) show intermediate level of methylation (42%) (106), and one-cell embryos (fertilized eggs) show higher but incomplete methylation (72%) (106). During the development of preimplantation embryos, L1 methylation gradually decreases (103, 106). When cultured in vitro, 2-cell embryos display 44% of methylation, 8-cell embryos 36%, calvitating embryos 32%, and expanded blastocysts 13–23% (106). By E7.5, L1s are fully methylated (102). As the result of demethylation in the developing germ cells, by E11.5, germ cells are nearly devoid of any methylation in both male and female fetuses (105). Remethylation can be seen as early as E12.5, and continues through E13.5 and E14.5 (105). By E16.5, female germ cells remain partially demethylated (103, 105), but male germ cells have become fully methylated (105). Specifically, the body of L1 sequences is fully methylated in pachytene spermatocytes, round spermatids, and epididymal sperm from the adult testis (103). Female germ cells are never fully methylated (103, 105, 106). In extraembryonic tissues, yolk sac mesoderm has a level of methylation similar to somatic tissues (102, 105), while yolk sac endoderm (102, 105), placental DNA (105), and the ectoplacental cone (102) are partially unmethylated.

The characterization of L1 5’UTR made it possible to specifically investigate the methylation status of L1 promoters. In addition, a large number of CpG sites in L1 5’UTR have been surveyed by bisulfite sequencing (107, 108). In general, the methylation patterns of 5’UTR during development are consistent with those of the body of L1. Specifically, 5’UTR sequences are heavily methylated in adult somatic tissues, such as rat liver (10) and mouse brain (109). For germ cells, sperm are highly methylated (98%), and ovulated oocytes are modestly methylated (29%) (107). Methylation of the zygote genome is comparable to oocytes (25%) (107), indicating active demethylation has occurred in the paternal genome. No significant methylation changes occur during preimplantation development, and the level of methylation persists through the blastocyst stage (27%) (107). The methylation status of 5’UTR between blastocyst and E11.5 is not known. An elevated level of methylation is observed in PGCs by E11.5 (65%). It rapidly decreases to 32% by E12.5, and further to 17% by E13.5 (107, 108). The dynamics of 5’UTR demethylation and remethylation differ between male and female. No difference is detected in 5’UTR methylation between male and female at E12.5 (109). In males, no further demethylation occurs beyond E16.5, and 5’UTR sequences are completely remethylated by E17.5 (109, 110). This high level of methylation is maintained in male germ cells from P0–P2 (110, 111), P8–P12 (110), P15 (112) and P17 (113) animals. In females, demethylation extends beyond E15.5, and some remethylation is observed at E17.5, although a substantial portion of L1s remain unmethylated (109).

4.2.3. Correlation of L1 methylation and expression during development

Methylation studies of mouse L1 reveal two developmental time windows when L1 promoters are substantially demethylated: first in preimplantation embryos, and second during gametogenesis (Section 4.2.2.). It is plausible that DNA methylation is not the sole mechanism in place for L1 repression, because, otherwise, the organism would have become very vulnerable during these two time periods due to uncontrolled L1 expression and activity. Possible candidates include repressive histone modifications that are normally associated with transcriptional gene silencing. An alternative, but not necessarily mutually exclusive, possibility is that the relaxed control allows a physiological level of expression that is tolerated by the host. In certain cases (i.e., during gametogenesis and neurogenesis), such level of expression may even be beneficial to the host. Here we discuss the correlation or the lack thereof between DNA methylation and L1 expression during three stages of mouse development. First, no correlation can be drawn for preimplantation embryos (the first window of opportunity for L1 expression), as little is known about the level of endogenous L1 expression during early embryogenesis in wild-type animals. This developmental stage should be revisited when more information becomes available. In fact, recent transgenic animal studies detect transcripts of human L1 transgene by PCR-based techniques (Section 5.1.2.). Second, there is a strong correlation between DNA methylation and endogenous L1 expression during fetal germ cell development (the second window of opportunity; Section 3.3.2.). In the mouse, L1 5’UTR sequences undergo rapid demethylation in male germ cells between E11.5 and E13.5 (from 65% to 17%) (107, 108), and then become fully remethylated by E17.5 (109). In good agreement with DNA methylation dynamics, prominent ORF1p signals are detected in E14.5–E17.5 embryonic testes (61, 77). Third, there is in general a lack of correlation between DNA methylation and L1 expression in the male germ cell lineage beyond E17.5 (discussed below in the context of meiosis progression).

Typically, germ cells in the embryonic mouse testis enter mitotic arrest at around E14.5, resume cell proliferation at around P2, and develop into spermatogonia via mitotic division. The latter are subsequently transformed into preleptotene spermatocytes, starting at P8–P10 and marking the first wave of meiosis. Leptotene, zygotene, pachytene, and diplotene spermatocytes from the prophase of meiosis I start to appear approximately at P10, P12, P14, and P17, respectively. Haploid round spermatids first appear at around P20 after the second meiotic division. They further differentiate into mature spermatozoa at around P30 (114, 115). Spermatogenesis continues in the adult testis in an asynchronous fashion. The majority of L1 5’UTR sequences are presumably hypermethylated in postnatal and adult germ cells for the following reasons, although DNA methylation has not been examined in germ cells of specific meiotic stages. First, there is no evidence for additional genome-wide demethylation of mouse 5’UTR sequences during spermatogenesis once they become fully methylated by E17.5 (109). Second, mouse L1 5’UTR is highly methylated in male germ cells from P0–P2 (110, 111), P8–P12 (110), P15 (112) and P17 (113) animals, and in adult sperm cells (107). Third, the body of L1 sequences is known to be fully methylated in pachytene spermatocytes, round spermatids, and epididymal sperm from the adult testis (103).

Despite the apparent hypermethylation status of 5’UTR, endogenous L1 is expressed in postnatal and adult testes (Section 3.3.2., and summarized below). A minor ORF1p signal is observed in P2 gonocytes, and reappears in P10 preleptotene spermatocytes (77). Prominent ORF1p signals are observed in primary spermatocytes (leptotene and zygotene stages) of the postnatal mouse testis (P10, P14, P18 and P25) (60). In the adult testis, L1 ORF1p is detected by immunofluorescence in leptotene, zygotene, and pachytene spermatocytes (78), and by immunohistochemistry in round spermatids, elongating spermatids and residual bodies in a separate study (60). During meiosis the germ cell genome undergoes extensive chromatin remodeling (116). Thus, it is likely that other epigenetic changes such as histone modifications are responsible for the activation of L1 expression in postnatal and adult male germ cells, although DNA demethylation at 5'UTR cannot be excluded.

4.2.4. L1 expression in mice deficient in piRNA/DNA methylation pathways

DNA methylation is a key mechanism in controlling endogenous transposable elements, and multiple cellular factors have been recently identified (117). DNA methyltransferase 3-like (DNMT3L) appears to regulate L1 remethylation specifically in the male germ line. DNMT3L is expressed in the non-dividing gonocytes in a brief perinatal period, which coincides with de novo methylation of retrotransposons including L1 (113). In Dnmt3l knockout mice, an elevated level of full-length L1 RNA is observed by Northern blot analysis in the P25 testis, and it is corroborated by in situ hybridization on cross sections of the P2 testis (113). DNA methylation analyses reveal significant demethylation at L1 5’UTR in germ cell but not somatic cell fractions isolated from Dnmt3l knockout P0–P2 (111) and P17 testes (113).

piRNAs belong to a class of small RNAs that are highly expressed in germ cells, and are so termed because of their association with PIWI proteins (118). Two mouse PIWI family proteins, MILI and MIWI2, are expressed during germ cell development, and their associated piRNAs display a strong strand bias (119). Primary piRNAs correspond to sense-strand transcripts from retrotransposons, and predominantly associate with MILI in the cytoplasm. Secondary piRNAs are mainly antisense-strand relative to retrotransposon transcripts, and associate with MIWI2 in the nucleus. Together, MILI and MIWI2 act to amplify piRNAs in a Ping-Pong cycle (119). Loss-of-function studies in mice suggest that piRNAs guide de novo methylation to retrotransposons (110, 119). L1 RNAs are up-regulated in male germ cells of Mili and Miwi2 knockout mice, and this change in L1 expression is correlated with significant demethylation of L1 5’UTR sequences (110, 120, 121). Furthermore, the PIWI/piRNA pathway likely acts upstream of known mediators of DNA methylation, because piRNAs are still produced in Dnmt3l-deficient mutants (119). Details of piRNA-meditated retrotransposon silencing have started to emerge. Multiple cofactors functioning in the piRNA/DNA methylation pathway have been recently revealed by mouse knockout experiments, including MAEL (a murine homolog of Drosophila maelstrom) (78), TDRD1 (tudor domain-containing protein-1) (112), TDRD9 (tudor domain-containing protein-9) (122), GASZ (a germ cell-specific protein with four ankyrin repeats, a sterile alpha motif, and a basic leucine zipper domain) (123), GTSF1 (gametocyte specific factor 1) (124), and MOV10L1 (Moloney leukemia virus 10 like-1) (125, 126). Mice deficient in any of these genes inevitably manifest meiotic arrest of spermatogenesis and male sterility. Such phenotype is commonly interpreted as a direct consequence of derepression of transposable elements, but mechanistic insights on how L1 derepression begets meiotic arrest remain lacking.

4.3. Posttranscriptional regulation of L1 expression

The second layer of L1 regulation is posttranscriptional. The term posttranscriptional is broadly used here to describe any regulatory mechanisms that act downstream of RNA transcription. Candidates for posttranscriptional regulation of L1 expression include RNA processing (discussed in Section 3.1.4.), RNA interference (RNAi), piRNAs, and potential translational and post-translational mechanisms.

4.3.1. Host factors for posttranscriptional regulation

Several cytidine deaminases and TREX1 (3’ repair exonuclease 1) are promising host factors for posttranscriptional regulation of L1 activity. APOBEC3 (apolipoprotein B mRNA-editing enzyme, catalytic polypeptide-like 3), APOBEC1 and AID (activation-induced deaminase) belong to a family of RNA or DNA cytidine deaminases, which can mutate RNA (or DNA) by converting cytidine to uridine (or deoxycytidine to deoxyuridine) (127). The mouse genome has a single APOBEC3 gene (mA3), while humans possess seven APOBEC3 genes (hA3A, hA3B, hA3C, hA3DE, hA3F, hA3G, and hA3H). The expansion of the APOBEC3 family during mammalian evolution coincides with a reduced level of overall retrotransposition activity from endogenous retrotransposons in primates (127). Several human APOBEC3 genes (128-133), but not mouse mA3 (129, 134), inhibit L1 retrotransposition in vitro. Similarly, the overexpression of APOBEC1 or AID from human, rat and mouse can inhibit the retrotransposition activity of a cloned human L1 in human cell lines (135, 136). It is puzzling, however, that L1 inhibition by all these deaminase appears to be independent of their deaminase activity. It is also noteworthy that the effect of these cytidine deaminases varies greatly from study to study (127, 137). TREX1 is an abundant 3’ to 5’ single-stranded DNA (ssDNA) exonuclease found in both human and mouse cells. Overexpression of TREX1 reduces L1 retrotransposition in vitro by 5-fold (138). In addition, Trex1-deficient mice display an approximately 6-fold accumulation of L1-related ssDNA in the heart. These L1-related ssDNAs potentially reflect a failure to degrade reverse-transcribed L1 ssDNAs in these Trex1 knockout animals (138). It remains to be investigated whether any of these host factors plays a role in controlling endogenous L1 retrotransposition, especially in tissues that naturally have high levels of L1 expression. For example, AID is expressed in both mouse PGCs and early embryos at a time window when demethylation occurs. Aid-deficient mouse PGCs undergo inefficient genome-wide erasure of DNA methylation, but the level of L1 expression/retrotransposition has not been evaluated in these cells (139).

4.3.2. RNA interference (RNAi)

The RNAi pathway is a conserved eukaryotic mechanism for post-transcriptional regulation, and involves two types of small RNA molecules (140). MicroRNAs (miRNAs) are usually derived from endogenously expressed precursor transcripts. Small interfering RNAs (siRNAs) were initially found to be of exogenous origin, but endogenous siRNAs have been recently discovered in a variety of species. In mammals, both types of small RNAs are processed by Dicer, a member of the RNase III nuclease family. One of the two strands, termed the guide strand, is incorporated into the RNA-induced silencing complex, which subsequently targets and cleaves complementary RNA species.

Human L1 5’UTR contains both a sense and antisense promoter, and bidirectional transcription of L1 5’UTR has the potential to generate double-stranded RNA specific to L1 5’UTR. Indeed, both sense- and antisense-strand small RNAs of ~21 bp are detected by Northern blot analysis of human 293 cells (141). In a reporter assay, the presence of the antisense promoter in 5’UTR decreases the level of 5’UTR-derived transcripts through post-transcriptional mRNA degradation (141). These results suggest that human L1 may have evolved to modulate the abundance of its own transcripts by co-opting the conserved cellular RNAi pathway. However, whether or not the observed L1 siRNAs arise from double-stranded precursors has not been determined. Additionally, it remains to be investigated whether siRNAs can be generated at other regions of an L1 sequence (142). There appears to be an inverse correlation between Dicer function and L1 expression level in Dicer-deficient ESCs (143), but it is not known whether such a correlation is simply due to an indirect effect. When Dicer function is selectively knocked out in mouse PGCs, the level of L1 RNA expression is reduced by 2-fold at E13.5, and shows no significant change at P3 (144). The latter results suggest Dicer-mediated processes may be dispensable for early germ cell development.

4.3.3. piRNA-mediated posttranscriptional silencing in germ cells

In addition to its role in de novo DNA methylation (Section 4.2.4.), recent studies suggest that piRNA machinery is also involved in posttranscriptional silencing of transposable elements. piRNA pathway components localize to two distinct types of cytoplasmic granules in fetal gonocytes: pi-bodies contain the MILI-TDRD1 module, and piP-bodies contain the MIWI2-TDRD9-MAEL module (77). The latter are considered as a specialized form of P-bodies because piP-bodies and P-bodies share several mRNA degradation/translational repression components. MAEL plays a critical role in piP-body formation. In Mael-deficient gonocytes, pi-body composition is not affected, but MIWI2, TDRD9 and MVH (mouse VASA homolog protein) fail to localize to piP-bodies (77). As a consequence, secondary piRNA production is abrogated, and de novo DNA methylation of L1 5'UTR sequences is delayed in Mael-deficient gonocytes. Interestingly, the defect in piP-body formation is accompanied by a prolonged accumulation of L1 ORF1p signals in postnatal spermatogonia even after DNA methylation has been fully restored by E18.5. This temporal decoupling of de novo DNA methylation and L1 protein expression can be explained by a failure in piRNA-mediated postranscriptional/translational silencing (77), the remarkable stability of L1 ORF1p as demonstrated in solution (145), or both. Additional support for piRNA-mediated posttranscriptional silencing of L1 comes from an experiment in which the expression of a single specific antisense L1 piRNA is down-regulated (146). In contrast to global piRNA knockout studies, the down-regulation of this specific L1 piRNA leads to L1 derepression in the absence of any overt effect on spermatogenesis and male fertility (146).

4.3.4. Cytoplasmic sequestration of L1 ribonucleoprotein (RNP) complexes

The formation of a cytoplasmic RNP complex between L1 proteins and L1 mRNA is a proposed intermediate step for mouse and human L1 retrotransposition (73, 75, 147, 148). Genetic and biochemical studies indicate that proteins expressed from active full-length L1s preferentially mobilize the source template L1 mRNA via a phenomenon termed cis-preference (21-23). Both ORF1p and ORF2p localize predominantly to cytoplasm when overexpressed in human cell lines. In addition, the colocalization of L1 RNA and proteins is often associated with markers of cytoplasmic stress granules (149-151). Stress granules represent sites of stalled translation initiation complexes that accumulate in response to a variety of stress conditions (152). Thus, it is possible that cells recognize overexpression of L1 as a stressor and consequently sequester L1 RNP in stress granules for degradation (150). However, it remains unknown whether transit through stress granules by an L1 RNP is a requisite step for productive L1 retrotransposition (150, 151).

5. L1 TRANSGENIC MOUSE MODELS

Endogenous L1 studies have provided a wealth of information about L1 expression and regulation, but they are not without limitations. First, high levels of background transcriptional signals are often produced from the vast number of preexisting L1 sequences in the genome. As a consequence, the reliable detection of endogenous L1 RNA relies heavily on methodology that can differentiate full-length, authentic L1 transcripts from abundant transcriptional noises. Northern blot analysis is well suited for this purpose but it lacks the level of sensitivity that quantitative RT-PCR offers. Low but physiologically significant levels of L1 expression might have escaped detection by Northern blot analysis in earlier studies. Second, de novo L1 insertions by endogenous L1s are intractable by conventional techniques, because it is equivalent to “finding a needle in a haystack” of preexisting insertions. Third, standard loss- or gain-of-function studies by targeted mutagenesis cannot be performed on endogenous L1s due to its repetitive nature. These limitations, along with the prospect of developing L1 as useful insertional mutagens, prompted transgenic studies of tagged L1s in rodents (153). So far, at least nine different L1 transgenes, regulated by endogenous L1 promoters and/or heterologous promoters, have been studied (63, 153-160). These transgenic models of L1 have started to provide exciting new insights in L1 biology, especially on how L1 functions in vivo. In this section we will highlight the progress made concerning these models, with a focus on transgenic mouse models using endogenous L1 promoters.

5.1. L1 expression from transgenes

Under the regulation of their native 5’UTR promoters, human L1 transgenes provide a rare opportunity for tracking L1 expression from a specific L1 locus (63, 153, 158). On one hand, the expression of L1 transgenes mimics that of their endogenous counterparts, i.e., in male and female gonads (153). On the other hand, these studies also reveal some pleasant surprises, such as the unexpected expression of L1 in the brain (63) and an abundance of L1 transgene transcripts in preimplantation embryos (158). However, a thorough survey of L1 transgene expression that parallels endogenous L1 studies is still lacking.

5.1.1. Expression in germ cells

Human L1 transgenes are expressed in both male and female germ cells. In a study featuring an active human L1 isolated from a patient with X-linked retinitis pigmentosa (L1_RP) (161), strand-specific transcripts from the 5’UTR-L1_RP transgene are found only in the testis and ovary among all tissues examined (153). Specifically, L1 transcripts are detected in pachytene spermatocytes, round spermatids, and elongating spermatids from adult testes (153). These results are consistent with observations from endogenous L1 studies (Section 3.3.2.): the endogenous mouse ORF1p is detected in the pachytene spermatocytes by immunofluorescence (78), and in round spermatids and elongating spermatids by immunohistochemistry (60). The above germ cell expression pattern has been replicated by a subsequent study using independently derived L1_RP transgenic mouse and rat lines (158). However, in contrast to endogenous L1 studies, L1 transgene expression has not been examined in germ cell populations from fetal and prepuberal mouse testes.

5.1.2. Expression in preimplantation embryos

One surprising finding from 5’UTR-driven L1 transgenes is the detection of abundant L1 transgene transcripts in preimplantation embryos (158). This particular study features several human L1 transgenes, including L1_RP transgenic mouse and rat lines (the same mouse line reported in reference (63)), a transgene based on a different human L1 isolate (LRE3 for L1 retrotransposable element-3, isolated from a patient with chronic granulomatous disease (162)), and a transgene based on an active endogenous mouse L1 (G_F21, (20)). Transcripts of these L1 transgenes are readily detected by RT-PCR in donor-positive preimplantation embryos (morulae or early blastocysts) (158). Interestingly, a significant proportion of donor-negative embryos (~40%) are also positive for L1 expression. The latter finding suggests that these RNAs are expressed prior to the completion of gametogenesis and subsequently partitioned away from the donor transgene during meiosis (158). Such “carryover” RNAs are functional because retrotransposition events are observed in ~10% of donor-negative animals (158). This high prevalence of carry-over RNA in donor-negative animals is unprecedented. It may result from a higher level of expression from transgenes as compared to endogenous elements. A systematic survey of endogenous L1 expression in preimplantation embryos should provide further insights about L1 activities during this important developmental stage, as full-length endogenous L1 transcripts have been previously detected in mouse blastocysts (76).

5.1.3. Expression in somatic tissues

Currently there is little evidence that 5’UTR-driven L1 transgenes are expressed in adult somatic tissues. Inconsistent and usually faint signals can be detected by RT-PCR in lung tissue from 5’UTR-driven L1_RP lines, but their significance is not known (153, 158). Presumably, 5’UTR-driven human L1_RP transgenes are expressed in the mouse brain during fetal and adult neurogenesis because of observed de novo retrotransposition events launched from the donor transgene (63). In this regard, the detection of L1 RNA in the head portion of E10.5–E11.5 embryos provides a necessary link between L1 expression and retrotransposition in the fetal brain (158). However, to date, no L1 RNA signal has been seen in adult brain samples (153, 158), although it is unknown whether or not regions active in adult neurogenesis, such as the subventricular zone of the lateral ventricle and the subgranular zone of the hippocampus (163), were sampled in these studies. Immunofluorescence staining of wild-type brain sections indicates that a very small number of cells in the ventricular zone and the dentate gyrus of hippocampus are reactive to an antibody against mouse ORF1p (63). It is yet to be demonstrated how the expression level of human L1 transgenes compares to endogenous expression.

5.2. L1 retrotransposition in mouse models

L1 transgenes are tagged by retrotransposition indicator elements, and thus de novo retrotransposition events can be specifically tracked. Most frequently, tracking is achieved by using GFP (green fluorescent protein) expression cassettes that are specifically designed for reporting retrotransposition events (153). However, the placement of an intron alone in the 3’UTR of an L1 transgene also allows the detection of retrotransposition by PCR-based techniques (158). In this subsection, findings from transgenic mouse models are divided according to different stages of mouse development.

5.2.1. Retrotransposition in germ cells

Consistent with L1 expression in germ cells, retrotransposition events by 5’UTR-driven L1 transgenes are detectable in testes (63, 153, 158). Using RT-PCR, an earlier study detected a strand-specific spliced EGFP signal, which is indicative of retrotransposition, in elongating spermatids from one out of four 5’UTR L1_RP lines, as well as from two control lines driven by a mouse RNA polymerase II large subunit (PolII) promoter (153). The germline retrotransposition frequency is low in elongating spermatids (~0.1% as estimated by RT-PCR (153), and a very faint amplicon by PCR (158)). Different 5’UTR lines display significant variation of retrotransposition, which appears to be correlated with the transcription level. For example, the only 5’UTR line with detectable level of retrotransposition by RT-PCR also has the highest level of transgene expression (153). Retrotransposition in the testis of transgenic animals has also been demonstrated for an independent L1_RP transgene by indirect immunofluorescence detection of GFP proteins (63). Thus, it is evident that de novo retrotransposition from 5’UTR-driven human L1 transgenes can occur in germ cells. Perhaps reflecting its low frequency, so far no heritable germline insertion has been reported in mice carrying 5’UTR-driven L1 human transgenes (153, 158). In contrast, heritable germline insertions have been observed in L1 transgenic lines carrying heterologous promoters, such as PolII promoter (153) and a composite chicken beta-actin (CAG) promoter (156, 157).

5.2.2. Retrotransposition in preimplantation embryos

De novo retrotransposition can occur during early human embryogenesis, as evidenced by a case study for a specific L1 insertion in a nuclear family with choroideremia (L1_CHM) (164). Genotyping data indicate that the mother was both germline mosaic and somatic mosaic for L1_CHM insertion, and thus this retrotransposition event must have occurred during early embryogenesis, before the patient’s germ cells became a distinct lineage (164) (in mouse this happens around E6.25 (165)). Since L1_CHM is an isolated case, the magnitude of such early embryonal events has not been appreciated until recently. Several interesting observations are made when retrotransposition is measured by PCR in preimplantation embryos (morula or blastocyst stage) from transgenic rodent models of both human and mouse L1s (158). First, retrotransposition is much more frequent in preimplantation embryos than in male germ cells, although a similar, if not higher, level of L1 expression is observed in germ cells (158). This discrepancy between transcription and retrotransposition suggests a post-transcriptional block of L1 retrotransposition in developing germ cells, or a reduced fitness for germ cells that are affected by retrotransposition or simply by L1 protein products. Second, retrotransposition occurs in both donor-positive and donor-negative preimplantation embryos (158). The occurrence of retrotransposition in donor-negative preimplantation embryos is attributable to carryover RNA from germ cells (Section 5.5.1.). Interestingly, donor-positive and negative preimplantation embryos display a similarly low level of retrotransposition. The frequency of retrotransposition is estimated by semi-quantitative PCR to be approximately one in every fifty cells in late blastocysts (158). Thus, only one or two cells in a late blastocyst have an L1 insertion (the average number of cells in mouse late blastocysts is around eighty (166)). Indeed, no retrotransposition signal is detected from the majority of preimplantation embryos (158). Third, none of the insertions from donor-negative animals pass through the germ line, indicating that the majority of putative preimplantation insertion events occurred in cells not destined to become germ cells (158).

5.2.3. Retrotransposition during postimplantation and postnatal development

Several lines of evidence suggest that retrotransposition can occur in postimplantation embryos. First, GFP-positive cells, indicative of de novo retrotransposition events, are detected in the cephalic neural tube of E10.5 embryos (63). Second, retrotransposition in postimplantation embryos is supported by the comparison of the timing and relative retrotransposition frequencies between donor-positive and donor-negative samples. As discussed above (Section 5.2.2.), retrotransposition signals are present at a similarly low level in either donor-positive or donor-negative preimplantation embryos (158). Presumably, L1 RNA is continuously transcribed in donor-positive embryos post-fertilization, because all donor-positive preimplantation embryos show positive L1 RNA signals by RT-PCR. L1 RNA can also be detected in the head portion of E10.5–E11.5 donor-positive embryos (158). In contrast, L1 RNA is no longer detectable in E10.5 donor-negative embryos (158). Accordingly, retrotransposition can be readily detected by PCR in 64% of donor-positive and 9% of donor-negative weanlings. Retrotransposition signals, when measured by quantitative PCR, are roughly 10-fold higher in donor-positive animals than in donor-negative animals. These data suggest that most somatic retrotransposition events in these transgenic models occur during postimplantation and/or postnatal development. A detailed timeline of L1 expression and retrotransposition is of particular interest to developmental biologists, but must await further studies.

The physiological consequence of such somatic retrotransposition events, if any, remains to be determined. Significant variation of retrotransposition signals is observed not only among different transgenic animals but also among distinct tissues from the same animal (158). The frequency of retrotransposition, as determined by quantitative PCR, is higher in somatic cells (ranging from 0.02 to 0.002 per cell for brain, kidney, liver, lung and tail tissues), and lower in germ cells (0.001 per sperm) (158). However, in a separate study using the same L1_RP mouse line, no GFP-positive cells are detected by immunofluorescence in several adult somatic tissues (heart, kidney, intestine, liver, lung, muscle, skin, and spleen); brain is the only exception (63). Such discrepancy may be explained by the methodological difference between the two studies. It is possible that the low frequency of retrotransposition, as detected by PCR, evades immunofluorescence detection. Another fundamental difference between the methods is that immunofluorescence detection requires GFP expression from L1 insertions. Therefore, an alternative possibility is that most somatic tissues do not support GFP expression, likely as the result of transcriptional silencing of L1 insertions. The latter explanation is at least partially supported by data from in vitro experiments with rat HCN stem cells (63). HCN cells containing an L1 insertion are largely negative for GFP fluorescence unless they are differentiated into neurons (63). Inhibition of DNA methylation or histone deacetylation increases the number of GFP-positive HCN cells (63). Transcriptional silencing of L1 insertions has also been demonstrated in embryonal carcinoma cells (167).

5.2.4. Retrotransposition in the brain

Another exciting finding from transgenic L1 studies is the discovery of active L1 retrotransposition from L1 transgenes in the brain, an organ that is least suspected for L1 activity based on previous endogenous L1 expression patterns (63). In this study, retrotransposition events are detected by indirect immunofluorescence against GFP proteins, which can only be expressed if de novo retrotransposition has occurred from donor human L1_RP transgenes. Surprisingly, scattered GFP-positive cells are observed in a wide-range of adult brain regions, such as amygdala, cerebellum, cortex, hilus, hippocampus, hypothalamus, striatum, and ventricles (63). Staining for cell-specific markers indicates these EGFP-positive cells are neurons rather than oligodendrocytes or astrocytes (63). In a follow-up study, quantitative PCR analysis indicates an increased number of L1 ORF2 sequences in several regions of adult human brain as compared to other somatic tissues, such as heart and liver (168). Such results provide further support that endogenous L1s might express and retrotranspose in the human brain. However, a formal proof for brain-specific retrotransposition necessitates the mapping of new L1 insertions that are absent in control organs. A potential link between DNA methylation and L1 expression has been investigated by bisulfite sequencing of 5’UTR from fetal brain and skin tissues. Overall, two brain samples display a subtle but consistently lower level of DNA methylation when compared to matched skin samples, although it is difficult to evaluate the significance of such subtle changes of DNA methylation on L1 expression. In addition to locating GFP-positive cells in the adult brain, GFP signals are also detected in the cephalic neural tube of E10.5 embryos (63). Thus, the determination of the timing and dynamics of de novo retrotransposition events in the developing brain should provide important insights about the regulation of L1 expression in the neuronal system.

5.3. Challenges and opportunities for L1 transgenic models

So far, transgenic models of L1 have brought forth quite a few new discoveries (detailed in Sections 5.1. and 5.2.). For example, L1 transcription is abundantly detected in germ cells, but germ cells display the least amount of retrotransposition activity. Retrotransposition can also occur in donor-negative animals via RNA carryover from parental germ cells. L1 expression from transgenes spawns insertions in preimplantation embryos (albeit at relatively low frequency) and during postimplantation and postnatal development. For the first time, L1 transgenes enabled the detection of retrotransposition events in the adult brain and the developing neural tube, representing an important step toward understanding potential roles that L1 plays in generating neuronal diversification. While the advent of these exciting discoveries merits celebration, their impact will not be fully appreciated without rigorous follow-up studies. Especially, some aspects of these findings have yet to be corroborated by alternative approaches. In this section, we discuss some of the opportunities and challenges that face L1 transgenic studies.

5.3.1. Are human and mouse L1 families similarly regulated in vivo?

It is debatable whether L1 families from different mammalian species are regulated in a similar fashion in their respective cellular environment in vivo. It appears that host regulation of L1 expression hinges on L1 promoters. This control is primarily achieved at the transcriptional level, although some aspects of L1 regulation occur at the posttranscriptional phase. These regulatory mechanisms include DNA methylation, piRNAs, siRNAs, and transcription factors (Section 4.). As the result of repeated acquisition of novel 5’ sequences during evolution, different L1 families possess 5’UTR promoters with little sequence conservation (8). Most of the known sequence motifs present in the 5’UTR of currently active human Ta subfamily are only found in more recent primate-specific L1 families (17, 85). These motifs include transcription factor binding sites for p53, RUNX3 and the SOX family proteins, and the human L1 antisense promoter. The SOX-binding site is an interesting case. Despite the overall lack of sequence conservation, both mouse and rat L1s have evolved to contain at least one copy of the overlapping consensus SOX and TCF/LEF binding site (Section 4.1.). Species-specific sequence motifs and the species-specific differences of host factors, such as transcription factors and APOBEC3 proteins, may underlie differential L1 activities among mammalian lineages. DNA methylation, as a key mechanism for heritable silencing of retrotransposons, may operate similarly among humans, rats, and mice. L1 promoters from all three species are CpG rich, and likely subject to similar cycles of demethylation and remethylation during development (Section 4.2.). piRNA biogenesis is intimately coupled with de novo DNA methylation during fetal germ cell development in mice. piRNA has also been isolated from rat testes (169). Primary piRNAs are likely generated as a response to an elevated level of L1 transcription due to the initial wave of demethylation in fetal germ cells. The processing of full-length L1 transcripts into short piRNAs may also reduce the overall level of functional L1 transcripts during this critical stage of germ cell development (Sections 4.2. and 4.3.). Nevertheless, L1 proteins can be readily detected (Section 3.1.4. for human samples, and Section 3.3.2. for mouse samples). The physiological consequences, if any, of such levels of L1 expression have not been evaluated.

5.3.2. Do L1 copy number and genomic location matter?

L1 transgenes differ from their endogenous counterparts in both copy number and genomic distribution. L1 transgenes are typically present in the genome as tandem-repeated copies in a single genomic locus, a byproduct of transgenesis (170). In contrast, endogenous L1s are present largely as dispersed individual copies. Both copy number and genomic location can impact transgene expression. In particular, tandem-repeated transgenes frequently display position effects, i.e., they show substantial variation of transgene expression among different transgenic mouse lines according to their location in the genome (171). In addition, transgenes are subject to repeat-induced gene silencing (RIGS), which typically manifests as an inverse correlation between transgene copy number and expression level from a specific genomic locus (172, 173). The potential effect of RIGS on L1 transgenic studies has been demonstrated by a recent study employing an L1 transgene under the regulation of a heterologous, constitutively expressing promoter (160). Specifically, the reduction of transgene copy number to a single copy at the same genomic locus shows no decrease in retrotransposition activity. An average frequency of 0.45 insertions per cell is observed for animals carrying a single-copy donor transgene, representing a nine-fold increase of retrotransposition activity on a per-copy basis. The observed retrotransposition activity is correlated with differential CpG methylation at the heterologous promoter: the promoter region is largely methylated in animals with the high-copy transgene array but significantly hypomethylated in animals with the single-copy counterpart; the body and 3’ end of the transgene show high levels of methylation in both high-copy and single-copy animals (160).

It remains an open question whether RIGS applies to tandem-repeated L1 transgenes under the regulation of endogenous L1 promoters. At the transcriptional level, if the 5’UTR is the primary target for piRNA-guided de novo methylation and silencing of L1 retrotransposons, it is reasonable to expect that a single-copy L1 transgene containing a mouse L1 5’UTR promoter will be recognized by endogenous piRNAs, and regulated similarly as compared to endogenous mouse L1s. A human L1, when placed as a transgene in the mouse cellular environment, may elicit efficient piRNA generation if the piRNA biogenesis is purely a function of L1 expression. It remains to be determined whether the high number and dispersed status of endogenous L1s make these retrotransposons more vulnerable to DNA methylation than introduced L1 transgenes.

6. PERSPECTIVE

It is now abundantly clear that L1 is expressed during development and disease. However, in most cases, a definitive role for L1 expression has yet to be established. Currently, our knowledge about the developmental timing and cell type specificity of L1 expression and its regulation comes predominantly from studies of L1 in mice. In addition, both commonalities and contrarieties on L1 regulation have been revealed for human, rat, and mouse L1 retrotransposons. The field as a whole has just started to learn the molecular inner workings of L1 regulation. This leaves great opportunities as well as grand challenges for modeling human L1 as transgenes in mice. Especially, the similarities and differences between human and mouse L1 families, including both cis- and trans-acting factors, should always be considered. A comparative approach between humans and mice is thus recommended.

Abbreviations

3’UTR

3’ untranslated region

5’UTR

5’ untranslated region

AID

activation-induced deaminase

APOBEC3

apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like 3

DNMT3L

DNA methyltransferase 3-like

DSB

double-strand DNA break

EB

embryoid body

ECC

embryonal carcinoma cell

EGC

embryonic germ cell

ESC

embryonic stem cell

GCT

germ cell tumor

GFP

green fluorescent protein

HCN

hippocampus neural (stem cell)

L1

long interspersed element type 1

L1_CHM

an L1 insertion in a family with choroideremia

L1_RP

L1 isolated from a patient with X-linked retinitis pigmentosa

LPR

length polymorphism region

LRE3

L1 retrotransposable element-3 isolated from a patient with chronic granulomatous disease

LTR

long terminal repeat

MBD

methyl-CpG-binding domain

MeCP2

methyl-CpG-binding protein 2

miRNA

microRNA

NPC

neural progenitor cell

NSC

neural stem cell

ORF1p

open reading frame 1 protein

ORF2p

open reading frame 2 protein

PGC

primordial germ cell

piRNA

PIWI-interacting RNA

RCL

rat chloroleukemia

RIGS

repeat-induced gene silencing

RT-PCR

reverse transcriptase-polymerase chain reaction

RUNX3

runt-domain transcription factor 3

siRNA

small-interfering RNA

SOX

SRY-related HMG box containing

ssDNA

single-stranded DNA

Ta

transcriptionally active

TCF/LEF

T-cell factor/lymphoid enhancer factor

TREX1

3’ repair exonuclease 1

YY1

Yin Yang-1