Loss of ciliary transition zone protein TMEM107 leads to heterotaxy in mice

Abstract

Cilia in most vertebrate left-right organizers are involved in the original break in left-right (L-R) symmetry, however, less is known about their roles in subsequent steps of the cascade – relaying the signaling and maintaining the established asymmetry. Here we describe the L-R patterning cascades in two mutants of a ciliary transition zone protein TMEM107, revealing that near-complete loss of cilia in Tmm107^nul1 leads to left pulmonary isomerism due to the failure of the midline barrier. Contrary, partially retained cilia in the node and the midline of a hypomorphic Tmem107^schlei mutant appear sufficient for the formation of the midline barrier and establishment and maintenance of the L-R asymmetry. Despite misregulation of Shh signaling in both mutants, the presence of normal Lefty1 expression and midline barrier formation in Tmem107^schlei mutants, suggests a requirement for cilia, but not necessarily Shh signaling for Lefty1 expression and midline barrier formation.

1. Introduction

Heterotaxy, an abnormal, often randomized, formation and arrangement of internal organs along the left-right (L-R) axis, typically involves severe cardiac abnormalities, associated with a high mortality rate (Kim, 2011; Lin et al., 2014). Mouse models have been essential to discern the function of genes mutated in patients and identify new heterotaxy loci (Li et al., 2015; Norris and Grimes, 2012). Among known heterotaxy-associated genes many are involved in cilia formation and function (Li et al., 2015; Pennekamp et al., 2015). Motile cilia in the left-right organizer (node in mouse, gastrocoel roof plate in frog, Kupffer’s vesicle in zebrafish) create a leftward flow of extraembryonic fluid that breaks bilateral L-R symmetry. Mutations in ciliary dyneins that lead to paralyzed cilia, as in Dnah11^iv mutant mice, or as seen in patients with primary ciliary dyskinesia, result in randomized L-R patterning, including left and right isomerism – left or right bilaterally symmetric organs, respectively (Dougherty et al., 2016; Layton, 1976; Piedra et al., 1998; Seo et al., 1992; Suppet al., 1997). Sensation of flow by cilia on the crown cells of the node is necessary for degradation of Dand5 mRNA in the crown cells at the left of the node, which relieves DAND5 inhibition of Nodal expression in these cells and ultimately leads to the induction of the left-side determinant Nodal in the left lateral plate mesoderm (LPM) (Marques et al., 2004; Yoshiba et al., 2012). Mice harboring mutations in the ciliary-localized putative calcium channel Polycystin2 (PC2, encoded by Pkd2) fail to induce Nodal in the left LPM due to their inability to sense flow (Yoshiba et al., 2012), and thus develop two right sides (Pennekamp et al., 2002).

Mounting evidence suggests that cilia have roles in L-R patterning beyond inducing and sensing the flow at the node. For example, loss of genes critical for ciliary assembly, like Kif3a, Ift172, Ift88 and Ift38, results in bilateral expression of leftness markers in the LPM, while current models would instead predict right isomerism due to a failure to generate and sense the flow, and induce the Nodal cascade (Botilde et al., 2013; Grimes et al., 2016; Huangfu et al., 2003; Murcia et al., 2000; Takeda et al., 1999). Furthermore, loss of cilia is epistatic to the loss of Pkd2, as evidenced by left isomerism in Kif3a^null; Pkd2^null embryos (Grimes et al., 2016), indicating a role for cilia downstream of flow sensation by PC2.

The ciliary transition zone (TZ) is a region at the base of the cilium, responsible for the mediation of ciliary protein composition. Defects in TZ components result in abnormal ciliary protein localization, and typically, an overall reduced number of cilia (Garcia-Gonzalo et al., 2011; Shylo et al., 2016). Disruptions of L-R patterning have been observed in mice harboring mutations in TZ protein-encoding genes, including Rpgrip1L (Vierkotten et al., 2007), Tctn1, Tctn2 and Tmem67 (Garcia--Gonzalo et al., 2011), Mks1 (Weatherbee et al., 2009) as well as B9d1 (Dowdle et al., 2011). However, none of these reports offer a thorough examination of the molecular basis for the L-R phenotypes occurring in these transition zone protein mutants.

Previously described Tmem107^schlei mouse mutants, and human patients with mutations in TMEM107, which encodes a TZ protein, display fewer and abnormally shaped cilia, but no L-R patterning defects have been reported (Christopher et al., 2012; Lambacher et al., 2016; Shaheen et al., 2015; Shylo et al., 2016; Weatherbee et al., 2009). In this study we report that Tmem107^tm1Lex mice that completely lack the Tmem107 gene product (hereafter referred to as Tmem107^null) (Tang et al., 2010), have bilateral Nodal expression in the LPM, which underlies left pulmonary isomerism, a molecular phenotype that largely resembles mutants with a global loss of cilia. We took advantage of the Tmem107^schlei and Tmem107^null mutants and their differences in L-R patterning phenotype to examine additional roles for cilia in establishing the L-R axis. We show that although both Tmem107^null and Tmem107^schlei mutants lack cilia in the endoderm at e8.0, presence of a robust population of cilia in the nodes and midlines of Tmem107^schlei animals is sufficient for normal establishment of the Nodal cascade. These results expand our understanding of limited observations in several other TZ mutants that develop left pulmonary isomerism and/or bilateral expression of markers in the LPM (Abdelhamed et al., 2015; Garcia-Gonzalo et al., 2011; Vierkotten et al., 2007), but also advance our knowledge toward deeper understanding of cilia roles in L-R patterning through establishment of the midline barrier.

2. Results

2.1. Tmem107^null mutants display multiple left-right defects

Tmem107 encodes a ciliary transition zone protein, required for the formation of a functional transition zone in primary cilia (Lambacher et al., 2016; Shylo et al., 2016). Given the ubiquitous nature of primary cilia, we were surprised to see an enrichment of Tmem107 mRNA in the embryonic node at E7.5–8.5, compared to all other embryonic tissues (Fig. 1 A, B and Fig. S1), suggesting a possible role for TMEM107 in the node at this early stage. This expression pattern also appeared at odds with the lack of L-R abnormalities in a previously described ENU-induced Tmem107^schlei mouse mutant (Christopher et al., 2012). However, close examination of the Tmem107^null mutant (Tang et al., 2010), revealed multiple heterotaxy characteristics, including randomized heart looping and stomach positioning, as well as left isomerism of the lungs (Fig. 1 C–L and Table S3). These data indicate that TMEM107 is critical for establishing the L-R axis and that Tmem107^schlei is a hypomorphic allele. To determine the extent to which TMEM107 regulates L-R patterning we further examined the heterotaxy phenotype in Tmem107^null mutants.

Fig. 1. Tmem107^null mutants display multiple left-right patterning defects.

(A, B) Tmem107 mRNA is strongly enriched in the nodes of E8.0 embryos, as shown in (A) ventral view and (B) lateral view. (C) Heart positioning (based on the orientation of the apex to the left or right) and stomach positioning are randomized in Tmem107^null embryos. Mutant lungs predominantly have left isomerism, not observed in control lungs. Numbers represent individual animals. (D–F) Ventral view of E15.5 hearts. Control hearts have their apices on left (D, red arrow), while Tmem107^null hearts have randomized looping with the apex on left (E, arrow) or right (F, arrow). (G–L) Ventral view of E13.5 (G–I) and E15.5 (J–L) lungs. Control mouse lungs have 4 right lobes and a single left lung lobe, while Tmem107^null lungs display single lobes on both sides (H, I, K, L), suggesting left pulmonary isomerism. Pulmonary hypoplasia in Tmem107^null mutants becomes pronounced by E15.5 (K, L). Cr - cranial lobe, Me - medial lobe, Ca - caudal lobe, Ac - accessory lobe, L - left lobe. Scale bars = 1 mm.

The position of the heart apex was used to determine directionality of heart looping and reveal its randomization. About a quarter of Tmem107^null mutants (n = 6/26) showed clear, abnormal right-side looping of the heart, with another quarter exhibiting ambiguous looping, without clear apex positioning towards left or right (n = 6/26) (Fig. 1 C–F, red arrows in D–F). Stomach positioning appeared to be independent of heart directionality and showed perfect randomization in Tmem107^null embryos compared to the normal left-sidedness (Fig. 1 C, Table S3). We also examined the developing lungs, which, in control mice, form one lobe on the left side, and four lobes on the right side (Fig. 1 C, G, J). The lungs in Tmm107^null embryos were bilaterally single-lobed (Fig. 1 C, H, I, K, L). Careful examination of E13.5 lungs revealed that both single lobes had equivalent numbers of branching points, consistent with normal branching pattern for the left lobe at that stage (Metzger et al., 2008) (Fig. 1 C, G–I). This could potentially result from a developmental delay in the lung lobation of the right lung. However, at later stages, we observed that Tmem107^null lungs maintain single lobes on each side, in addition to becoming severely hypoplastic, (Fig. 1 C, J–L). Overall, we conclude that Tmem107^null embryos develop left pulmonary isomerism (Fig. 1 C, G–L).

2.2. L-R perinodal signaling is disrupted in Tmem107^null mutants

Left-right patterning of the organs begins at the node, and the earliest marker of the L-R symmetry break is degradation of Dand5 mRNA on the left side of the node in response to cilia-generated flow (Inacio et al., 2013; Kawasumi et al., 2011; Marques et al., 2004; Nakamura et al., 2012; Schweickert et al., 2010; Shinohara et al., 2012). Left-side Dand5 degradation is apparent by 2–3 somites stage in controls and becomes more pronounced at the 4–5 somites stage (Fig. 2 A, B, E). In contrast, Dand5 expression in Tmem107^null embryos is bilateral and equal at the 2–3 somites stage, and either maintains equal bilateral expression, or becomes degraded in a random fashion by 4–5 somites stage (Fig. 2 C, D, F). Normally, left-sided Dand5 degradation alleviates its inhibition of Nodal, and leads to subsequent Nodal upregulation on the left side of the node (Fig. 2 G, H, K). We expected that the equal bilateral Dand5 expression would lead to bilateral down-regulation of Nodal (Marques et al., 2004), but we observed randomized Nodal expression levels around Tmem107^null nodes (Fig. 2I–L). These results suggest a disconnect between the expression of Dand5 and Nodal in the nodes of Tmem107^null embryos, but appear to be in line with some other similar reports in the literature (Ishimura et al., 2008; Marques et al., 2004; Norris et al., 2002).

Fig. 2. Flow-induced signaling in Tmem107^null nodes is compromised.

(A–D) in situ hybridization for Dand5, at early (2–3 somites) and late (4–5 somites) stages. Control embryos (A, B) show progressive degradation of Dand5 mRNA on the left side, while Tmem107^null embryos (C, D) maintain bilateral Dand5 mRNA, with some randomization of expression at later stages. (G–J) in situ hybridization for Nodal, at early (2–3 somites) and late (4–5 somites) stages. Control embryos (G, H) show reduced Nodal expression on the right side at late stages (H), while Tmem107^null embryos (I, J) maintain bilateral Nodal mRNA for longer, and ultimately have a randomized pattern of expression (L). (E, F, K, L) Graphs accompanying each probe represent the quantification of the embryos with a given expression pattern, broken down by somite number. Scale bars are 100 μm.

2.3. Nodes of Tmem107^null embryos have cilia with motile identity

Dand5 degradation is dependent on the ability to sense nodal flow (Nakamura et al., 2012; Schweickert et al., 2010), so the lack of early left-sided Dand5 degradation points to abnormal motile cilia in the pit of the node or flow sensing cilia surrounding the node. We used ARL13B as a marker of cilia, and discovered that the nodes of E8.0 (2–5 somites) Tmem107^null embryos were the only region where we detected ARL13B + cilia, as compared to extensive ARL13B + cilia throughout control embryos (Fig. 3 A–F – high magnification of representative nodes (4 somites), S2 A–D, G–J – low magnification). However, Tmem107^null mutant nodes had significantly (p < 0.0001) fewer cilia (mean = 75), as compared to controls (mean = 226) (Fig. 3 I), which is consistent with overall cilia number reductions reported in Tmem107^schlei mouse mutants, and in human patients with mutations in TMEM107. (Christopher et al., 2012; Lambacher et al., 2016; Shaheen et al., 2013; Shylo et al., 2016). We also discovered that cilia in the nodes of Tmem107^null mutants were significantly shorter than in control nodes (null mean = 1.811 μm, control mean = 2.568 μm, p < 0.0001) (Fig. 3 J). This result presents a deviation from previously published observations of longer primary cilia in cell cultures, derived from patients with mutations in Tmem107 (Lambacher et al., 2016; Shaheen et al., 2013; Shylo et al., 2016).

Fig. 3. Motile and sensory cilia are present in Tmem107^null nodes, but in reduced numbers.

(A-H, K–P) Maximum intensity projected Z-stacks of confocal immunofluorescent images of E8.0 nodes. (A, B) Low magnification images of control (A) and Tmem107^null (B) nodes show normal cilia numbers in the node and endoderm (ED) in control, but low cilia numbers in Tmem107^null nodes, and none in ED. (C, D) High magnification images of inserts in A and B, respectively. (C, E) in control, ARL13B (red) is present in all cilia in the node. (D, F) In Tmem107^null embryos ARL13B is only present in a small population of nodal cilia. (C, G) Polycystin2 (PC2, green) is enriched in sensory cilia on crown cells, around the edge of the node. The distribution of PC2 positive cilia appears normal in Tmem107^null nodes (B, D, H). (I) Quantification of the cilia in the control (n = 6) and Tmem107^null (n = 7) nodes show significant loss of cilia in the Tmem107^null (P < 0.0001) (J) Cilia are significantly shorter in Tmem107^null embryos (n = 285 cilia), compared to control (n = 824 cilia) (P < 0.001). (K–P) High magnification views of the nodal pits of control and Tmem107^null embryos that are also homozygous for the Dnah11^GFP transgene. ARL13B (M, N, and red in K, L) marks the nodal cilia showing that the nodal pit of Tmem107^null embryos has only a small population of ARL13B positive cilia. Control embryos have DNAH11-GFP positive motile cilia (K, green and O) throughout the pit of the node. Cilia in the pit of Tmem107^null nodes also have motile identity as marked by DNAH11-GFP (L, green and P). (Q–T) Four classes of lung phenotypes observed in this study. (Q) Normal – single lobe on left, four on right. (R) Left isomerism was the dominant phenotype observed in Tmem107^null; Pkd2^null animals. (S) many Pkd2^null animals had right isomerism. (T) Ambiguous lungs phenotype encompasses a range of lungs that have some lobation on either side, but neither left nor right isomerism. Most frequently observed in Tmem107^het; Pkd22^null animals. (U) Graphic representation for the quantification of the embryos with a given lung phenotype. Numbers represent individual embryos. ED - endoderm. Scale bars in (A–H) are 25 μm, in (K–P) are 50 μm, and 1 mm in (Q–T). Unpaired t-test used for statistical analysis. Error bars represent standard deviation.

The loss of ARL13B + cilia in the endoderm, surrounding the node (Fig. 3 A, B, S2 A–D, G–J) stands in contract with the specific enrichment of Tmem107 mRNA in the node. These data indicate that TMEM107 is required throughout the embryo and, in addition to its specific enrichment at the node (Fig. 1 A, B, S1); Tmem107 must be expressed in other tissues, but below the detection levels of in situ hybridization.

To test whether Tmem107^null mutants lack cilia with motile identity specifically, we introgressed a transgenic reporter for an axonemal dynein required for ciliary motility, Dnah11, fused to GFP (Dnah11-GFP) (McGrath et al., 2003). All cilia in the pit of control nodes are DNAH11-GFP + (Fig. 3 K, M, O). We found that most cilia in the pit of Tmem107^null nodes are also DNAH11- GFP + (Fig. 3 L, N, P) indicating that cilia with motile identity are specified in the absence of TMEM107 function. These results show that there is a capacity to make motile cilia. However, these cilia are shorter than normal (Fig. 3 J), and shorter motile cilia in the zebrafish Kupffer’s vesicle have been linked to reduced ciliary velocity, slower flow, and L-R patterning defects, a likely scenario in Tmem107^null nodes (Lopes et al., 2010).

The positioning of motile cilia in the node can also affect the efficiency of flow, and the resultant break in L-R symmetry. Planar cell polarity pathway (PCP) drives the positioning of the cilium to the posterior end of the cells in the node by late headfold stage (Hashimoto et al., 2010), which tilts them, and ultimately creates leftward-directed flow (Nonaka et al., 2005). Disrupted ciliary positioning in the nodal cells leads to randomization of L-R situs (Mahaffey et al., 2013; May-Simera et al., 2010; Song et al., 2010). We evaluated the positioning of cilia in Tmem107^null nodal pit cells along the anterior-posterior axes, and found that they are, indeed, located posterior to the center of the cell, similar to control (Fig. S2 N). Therefore, it seems that TMEM107 is not involved in PCP, and does not regulate ciliary positioning in the cell.

2.4. Nodes of Tmem107^null embryos retain cilia with sensory identity

To examine whether the Tmem107^null nodes have sensory cilia, we next examined the localization of PC2, which is enriched in crown cells surrounding the node and is required for flow sensing (Yoshiba etal., 2012)We observed many PC2 + cilia in Tmem107^null crown cells, suggesting that ciliated cells with flow-sensing identity are preserved in the mutants (Fig. 3 G, H – high magnification, S2E, F, K, L – low magnification of representative nodes). The absolute numbers of PC2 + cilia in Tmem107^null nodes was lower than in control nodes (null mean = 66.4 cilia vs control mean = 136.5 cilia, p ≤ 0.0125), however, proportionately to the overall numbers of cilia counted, roughly 60% of control cilia were both Arl13b + and PC2+, compared to 80% of Tmem107^nul cilia. Thus, nodes of Tmem107^null embryos maintain both cilia with motile and sensory identities, albeit at significantly decreased numbers and with altered morphology.

We were interested to test whether the remaining PC2+ cilia in Tmem107^null embryos are functional. In a genetic epistasis experiment, we analyzed the patterning and the numbers of lobes in the lungs of animals with a loss of Tmem107 and Pkd2 as a readout of the downstream L-R signaling that occurs in these animals. The four classes of phenotypes we observed were (1) normal (Fig. 3 Q), (2) left isomerism (Fig. 3 R), (3) right isomerism (Fig. 3 S), and (4) ambiguous (Fig. 3 T). Lungs deemed ambiguous included a range of phenotypes, most having some partial lobation on both sides, without being a clear right or left isomerism.

Pkd2^null mutant animals are typically reported to display right pulmonary isomerism due to their inability to sense nodal flow and relay that information to the LPM, thus failing to express Nodal and its downstream targets in the LPM (Grimes et al., 2016; Pennekamp et al., 2002). In our crosses, we used Pkd2^tm2Som mice, herein referred to as Pkd2^null (Wu et al., 1998). Most Tmem107 ^wt; Pkd2^null animals analyzed displayed right isomerism, consistent with other observations in the literature (Fig. 3 U, Table S4). However, a small subset of animals had more ambiguous lungs. Furthermore, a loss of a single copy of Tmem107 appeared to exacerbate this phenotype. Most Tmem107^het; Pkd2^null animals had ambiguous lungs, and some even presented with left isomerism (Fig. 3 U, Table S4). These data were more consistent with randomized and bilateral Pitx2 observed in Pkd2^tm1Dwo mutants, also presumed to be null (Pennekamp et al., 2002; Yoshiba et al., 2012). Overall, these observations point to susceptibility of the right isomerism phenotype in Pkd2^null animals to background differences and modifier genes, a phenomenon previously reported in the human population (Harris and Rossetti, 2010). Most importantly, these early observations suggested a genetic interaction between Tmem107 and Pkd2. We further found that Tmem107^null; Pkd2^null animals have left isomerism, most similar to Tmem107^null; Pkd2^wt animals (Fig 1 C, G–L, Fig. 3 U, Table S4).

These observations rise some key points. First, this phenotype is reminiscent of the bilateral Pitx2 phenotype, previously observed in Kif3a^null; Pkd2^null mutant that lacks cilia (Grimes et al., 2016), suggesting that Tmem107^null mutant largely behaves as a loss of cilia mutant. Lungs cannot be evaluated in Kif3a^null; Pkd2^null mutant animals due to their early lethality. Secondly, our original question of whether the remaining PC2+ cilia in Tmem107^null embryos are functional remains unanswered due to the epistatic relationship between Tmem107 and Pkd2. Together, our data suggest that TMEM107 has significant roles in L-R patterning downstream of the flow sensing event.

It is probable that short motile cilia in Tmem107^null nodes generate only weak leftward flow, and may even have randomized flow. However, the roles for TMEM107 downstream of ciliary involvement at the node override any potential phenotype that arises at the node. Given these observations, the remainder of these studies focus on the roles of cilia and Tmem107 in the late stages of establishing and maintaining L-R asymmetry.

2.5. Nodal and Pitx2 are expressed bilaterally in the lateral plate mesoderm of Tmem107^null embryos

Asymmetric Nodal at the node is thought to signal to the lateral plate mesoderm (LPM) and autoactivate robust Nodal expression specifically in the left LPM (Fig. 4. A, C), (Kawasumi et al., 2011; Yamamoto et al., 2003). Nodal expression in the left LPM is transient, starting around the 2–3 somites stage, and becomes extinguished a few hours later (6–7 somites stage). To test if there were defects in the transmission of the NODAL signal from the node to the LPM, we examined Nodal LPM expression between 0–5 somites (Fig. 4 A–C). Consistent with the lung phenotype, the vast majority of Tmem107^null embryos displayed bilateral Nodal transcripts in the LPM (Fig. 4 B, C) by 4–5 somites stage. Interestingly, a closer inspection of Tmem107^null embryos with bilateral nodal expression revealed that 7 of 8 embryos showed unequal expression between the two sides, with 6 having higher expression on the left side, and lower on the right (Fig. 4 B, C). These data and our observations of randomized Nodal at the node (Fig. 2 G–L) are consistent with previous reports of dissociation between Nodal expression at the node and its expression in the LPM in other mutants, although the mechanism by which this occurs is unclear (Ishimura et al., 2008; Marques et al., 2004; Norris et al., 2002).

Fig. 4. The lateral plate mesoderm in Tmem107^null embryos acquires bilateral left identity.

(A, B) in situ hybridization for Nodal in representative control (A) and Tmem107^null (B) E8.0 embryos showing left side restriction in controls and bilateral expression in mutants. (C) Quantification of the expression patterns for Nodal at the LPM, broken down by somite stage. (D, E) in situ hybridization for Pitx2 in representative control (D, left LPM, arrow) and Tmem107^null (E, bilateral, arrows) E8.5 embryos. Pitx2 is normally expressed bilaterally in the developing heads of control and mutant embryos. (F) Quantification of the Pitx2 expression localization. Numbers in all graphs indicate individual animals.

In the left LPM, NODAL signaling induces the expression of Pitx2, a transcription factor, whose left-side expression persists in various LPM-derived tissues, long after Nodal expression is lost (Fig. 4 D–F) (Lin et al., 1999; Meno et al., 1998; Ryan et al., 1998; Yoshioka et al., 1998). We observed bilateral expression of Pitx2 in the LPM of Tmem107^null embryos, indicating that once left-side identity is established in the LPM by NODAL, it is transduced accordingly (Fig. 4 E, F). Together, the Nodal and Pitx2 expression patterns are consistent with the left pulmonary isomerism that we observe in older Tmem107^null embryos (Fig. 1 G–L).

2.6. Tmem107^null embryos have a defective barrier and disrupted SHH signaling in the midline

Consistently bilateral Nodal and Pitx2 expression in the LPM of Tmem107^null embryos led us to examine additional factors that regulate L-R identity. The midline barrier model postulates that Lefty1, a TGFβ family member expressed in the left prospective floorplate (Meno et al., 1997), is required to inhibit the passage of NODAL to the right side of the embryo, preventing NODAL from autoactivating its expression in the right LPM (Meno et al., 1998; Yamamoto et al., 2003). Thus, bilateral Nodal expression in the LPM is a frequent indicator of the failure of the midline barrier. Additionally, broader Nodal expression in the left LPM than in the right may suggest that the signal from the node is originally propagated correctly, with the induction of Nodal in the left LPM, but due to the lack of the midline barrier, NODAL is able to auto activate in the right LPM as well (Nakamura et al., 2012).

In contrast to control embryos, none of the Tmem107^null mutants expressed Lefty1 in the midline (Fig. 5 B, C, arrowheads). The absence of Lefty1 expression, in conjunction with bilateral, frequently asymmetric (L > R) (Fig. 4 A–C), Nodal expression in the LPM indicates the loss of a functional midline barrier. As Lefty1 expression has been shown to be dependent on SHH signaling (Tsiairis and McMahon, 2009; Tsukui et al., 1999), we examined SHH signaling in the midline of Tmem107^null embryos.

Fig. 5. Tmem107^null embryos show Shh-dependent loss of the midline barrier.

(A, B) Double in situ hybridization for Lefty1 (midline, arrowheads) and Pitx2 (LPM, arrows) in representative control (A) and Tmem107^null (B) embryos. Pitx2 riboprobe was used as a positive control for in situ hybridization reaction conditions, and is expressed in left LPM in control embryos (A, arrow) but is bilateral in Tmem107^null embryos (B, arrows). In contrast, Lefty1 signal is lost in the midlines of Tmem107^null embryos (B, arrowheads). (D, E) in situ hybridization revealed unchanged midline expression of Shh in Tmem107^null (E) embryos compared to controls (D). (G, H) However, in situ hybridization for FoxA2, a direct target of SHH signaling, uncovered a strong reduction of midline FoxA2 expression in Tmem107^null (H) embryos compared to controls (G). (C, F, I) Graphs accompanying each probe represent the quantification of the embryos with a given expression pattern.

Hedgehog signaling is tightly linked to cilia in vertebrates, and our previous studies showed reduced SHH signaling in the neural tubes of Tmem107^schlei mutants (Christopher et al., 2012). At E8.0, Shh is expressed in the node and the notochordal plate (Fig. S3 A–D) (Echelard et al., 1993; Zhang et al., 2001). and SHH signaling is necessary to specify the floor plate and establish the midline (Chiang et al., 1996). We did not observe any difference in midline Shh expression Tmem107^null embryos (Fig. 5 D–F, Fig. S3 A, B). Both notochord and prospective floor plate are marked by robust Shh signal (Fig. S3 A, B). We then decided to look downstream of the ligand at the expression of two direct targets of SHH signaling, FoxA2 (Fig. 5 G–I) and Ptch1 (Fig. S4 A–C), which are both enriched in control midlines. Despite the normal expression of Shh, the midline enrichment of both FoxA2 and Ptch1 was reduced in all Tmem107^null embryos examined (Fig. 5 H, I, Fig. S4 B, C, arrowheads). The effect was more dramatic for FoxA2 reduction, but was consistently observed for both probes. These results are consistent with the requirement for TMEM107 for the transduction of the midline SHH signal, and to establish Lefty1 expression and midline barrier function.

2.7. Analysis of hypomorphic Tmem107^schlei embryos with normal L-R patterning reveals SHH-independent expression of Lefty 1

We previously reported that the hypomorphic Tmem107^schlei mutants did not show any disruptions in L-R asymmetry, but displayed abnormal SHH signaling in limbs and neural tubes (Christopher et al., 2012). Consistent with this, the most downstream marker of left-side patterning, Pitx2, is restricted to the left LPM of Tmem107^schlei mutants, similar to control embryos (Fig. 6 A–C, J, K). Shh was also expressed normally in the node and midline region of all Tmem107^schlei embryos examined, including both the notochord and the prospective floor plate (Fig. 6 D–F, S3 C, D). In contrast, FoxA2 was lost from Tmem107^schlei mutant midlines, similar to Tmem107^null embryos, and consistent with abnormal SHH signaling observed in other Tmem107^schlei tissues (Fig. 6 G–I) (Christopher et al., 2012). Notably, we detected consistent Lefty1 expression in the midline region of Tmem107^schlei embryos (Fig. 6 J–L). These data indicate that the mutant TMEM107 protein produced by the Tmem107^schlei hypomorphic allele is sufficient for the expression of Lefty1, but not FoxA2, and is capable of establishing a midline barrier and preserving L-R identity. More specifically, they point to potential differences in SHH sensitivity between FoxA2 and Lefty1.

Fig. 6. Tmem107^schlei embryos maintain left-right asymmetry despite reduced SHH signaling in the midline.

(A, B) in situ hybridizations for Pitx2 in control (A) and Tmem107^schlei (B) embryos show normal expression in the left LPM (D, E) in situ hybridization for Shh in control (D) and Tmm107^schlei (E) embryos, showing normal midline expression in both cases. (G, H) FoxA2 in situ hybridization is normal in controls (G) but is strongly reduced in the midlines of Tmm107^schlei (H) embryos. (J, K) Double in situ hybridization for Lefty1 (midline, arrowheads) and Pitx2 (LPM, arrows) revealed Lefty1 expression in the midlines of Tmem107^schlei (K, arrowheads) embryos, albeit faint and/or patchy. (C, F, I, L) Graphs accompanying each probe represent the quantification of the embryos with a given expression pattern.

2.8. Presence of midline cilia correlates with a functional midline barrier

TMEM107 is required for normal cilia number and ciliary protein localization (Christopher et al., 2012; Cui et al., 2011; Lambacher et al., 2016; Shaheen et al., 2013; Shylo et al., 2016; Weatherbee et al., 2009). The abnormal midline gene expression patterns in Tmem107^null mutant suggest TMEM107 may be required for cilia formation or function in the midline. Prior studies by scanning electron microscopy (SEM) and our analysis via immunofluorescence show normal cilia numbers and length in the nodes of Tmem107^schlei embryos (Fig. S5 and (Christopher et al., 2012)). However, Tmem107 mutant midlines have not been analyzed at E8.0 for the presence of cilia.

Examination of the ARL13B cilia marker in Tmem107^null embryos revealed very few midline cilia (Fig. 7 B, B′, E, E′) compared to controls (Fig. 7 A, A′, D, D′), consistent with the low numbers of cilia in the Tmem107^null nodes, and the reports that node cells migrate into the midline and contribute to the notochordal plate (Brennan et al., 2002; Lee and Anderson, 2008; Sulik et al., 1994). Although we observed very few cilia in the nodes of Tmem107^null embryos (Fig. 3 I), their almost complete absence in the midline was surprising. Unlike Tmem107^null mutants, Tmem107^schlei embryos appear to have a normal numbers of cilia in their nodes (Fig. S5 H and (Christopher et al., 2012)) and the midlines of these animals contain high numbers of ciliated cells (Fig. 7 C, C′, F, F’). Notably, cilia are still absent from the definitive endoderm in Tmem107^schlei embryos (Fig. 7 C), suggesting a differential requirement for TMEM107 function in primary cilia of distinct embryonic tissues, and that despite the enrichment for Tmem107 transcript at the node, the definitive endoderm is more sensitive to the Tmem107^schlei mutation. Together our findings suggest that loss of midline cilia may drive left isomerism through the loss of midline barrier function in Tmem107^null.

Fig. 7. Absence of cilia in the midline correlates with the laterality defects.

(A–C) Maximum intensity projected Z-stacks of confocal immunofluorescent images of E8.0 midlines using ARL13B as a cilia marker. Dashed lines outline the midline area. (A) Control embryos have numerous cilia in the midline (MD) and the surrounding endoderm (ED) (A′) Projected optical section of the boxed region in (A) shows widespread cilia.. (B) Only occasional cilia can be observed in the midlines of Tmem107^null embryos. (B′) Projected optical section of the boxed region in (B) shows few cilia. (C) A large population of ARL13 B positive cilia is consistently observed in the midline of Tmem107^schlei embryos. (C′) Projected optical section of the boxed region in (C) shows presence of cilia. (D–F) Confocal immunofluorescent images of sectioned E8.0 midlines, using ARL13B as a cilia marker. (D′-F′) Regions boxed in (D–F) focus on the prospective notochord that shows cilia in the control (D′, arrow) and Tmem107^schlei (F′, arrow), but not in Tmem107^null embryos (E′). Scale bar is 50 μm. (G) Summary diagram of key differences between the control and the two Tmem107 mutants.

3. Discussion

3.1. Transition zone protein TMEM107 exhibits functional specificity in different cilia types

The transition zone is a structurally complex and protein-rich domain of the cilium, and has been the focus of significant investigation in the ciliary field (Garcia-Gonzalo and Reiter, 2017; Lambacher et al., 2016; Szymanska and Johnson, 2012; Yang et al., 2015). Since cilia are ubiquitous organelles, and a transition zone is a mandatory structural component of any cilium, it would be expected for associated ciliary proteins to be expressed ubiquitously. As such, TMEM107 is expressed in all tissues examined (Christopher et al., 2012; Tang et al., 2010). Thus, it came as a surprise to see a strong Tmem107 enrichment in the nodes of developing embryos. At later stages, Tmem107 also shows strong enrichment in other tissues, including the choroid plexus, as well as olfactory and respiratory epithelia (Diez-Roux et al., 2011). All these tissues have motile cilia, suggesting a specific role for TMEM107, associated with ciliary motility.

Despite Tmem107 enrichment in the node, motile cilia appear to be more tolerant of mutations in Tmem107. Tmem107^schlei nodes and midlines have high numbers of cilia and no L-R patterning defects (Fig. 7 G). Tmem107^null mutants have a nearly complete loss of cilia, but do retain some in the node (Fig. 7 G). In contrast, primary cilia are more susceptible to disruptions in Tmem107, and are lost in both Tmm107^null and Tmm107^schlei embryos (Fig. 7 G). Overall, it has now been shown that TMEM107 is important for ciliary formation both in humans (Lambacher et al., 2016; Shaheen et al., 2015; Shylo et al., 2016) and mice (Christopher et al., 2012). Longer primary cilia have been reported in cultures of primary human fibroblasts, derived from patients with mutations in TMEM107 (Lambacher et al., 2016; Shaheen et al., 2015; Shylo et al., 2016). However, the remaining cilia in Tmem107^null nodes, positive for motility markers, are significantly shorter than in control embryos. Thus, it appears that TMEM107 is important for ciliary formation both in motile and primary cilia, likely through its requirement for ciliary length regulation.

3.2. Loss of Tmem107 results in disruption in early L-R patterning events at the node

The shorter cilia in the nodes of Tmem107^null embryos maintain their motile and sensory identities, but the pattern of Dand5 and Nodal expression at the node hint that the leftward flow, or its detection, is compromised. Work in zebrafish has shown that shorter cilia in the Kupffer’s vesicle generate slower flow, which results in a failure to degrade Dand5, and leads to randomized Nodal (Lopes et al., 2010; Sampaio et al., 2014). This phenotype is nearly identical to the phenotype in Tmem107^null nodes, leading us to believe that although cilia with motile identity are present in Tmem107^null embryos, the leftward flow is likely not strong enough to result in degradation of Dand5 on the left side of the node. We do note that among Tmem107^null embryos with bilateral Nodal expression in the LPM, 6/8 have broader expression on the left side. This phenotype may indicate that the flow in Tmem107^null embryos is predominantly leftward, albeit weak, and Nodal becomes initially activated in the left LPM, with its later induction in the right LPM, where it propagates posteriorly (Nakamura et al., 2012). In comparison, cilia numbers in Tmem107^schlei embryo nodes are similar to control, with overall normal cilia length, presumably resulting in the normal leftward flow, as evidenced by proper L-R patterning in these mutants.

After initiating the L-R patterning process, cells from the node contribute to the notochordal plate (Brennan et al., 2002; Lee and Anderson, 2008; Sulik et al., 1994). Discrepancy in cilia numbers in the nodes of the two Tmem107 mutants demonstrates another key difference between them - the absence of cilia in the Tmem107^null midline, and their presence in Tmem107^schlei (Fig. 7 G). Thus, Tmem107^null and Tmem107^schlei mutants represent an allelic series, and provide us with a unique opportunity to address the requirement for cilia in the node and the midline for the establishment and maintenance of L-R asymmetry.

3.3. Roles for cilia in establishing and maintaining the midline barrier

Cilia are hubs for SHH signaling in mammals and SHH is necessary for the proper formation of the notochord and the floor plate (Chiang et al., 1996). Since Lefty1 is expressed in the prospective floor plate, we expected its expression to be dependent on presence of cilia and SHH signaling (Tsiairis and McMahon, 2009; Tsukui et al., 1999). Both Tmem107 mutants show a strong reduction of FoxA2 in their midlines, suggesting defective SHH signaling, but Lefty1 is present in the midlines of Tmm107^schle (Fig. 7 G). Lack of Lefty1 in the Shh and FoxA2 mutant mice is likely due to structural defects and lack of the floor plate, rather than the failure to induce its expression through the SHH signaling cascade (Ang and Rossant, 1994; Dufort et al., 1998; Tsukui et al., 1999; Weinstein et al., 1994). Since the floor plate is formed in both Tmem107^null and Tmem107^schlei mutants, an alternative explanation is that low residual SHH signaling in Tmem107^schlei mutant is sufficient for Lefty1 expression (Christopher et al., 2012).

Loss of normal SHH signaling in the Tmem107^schlei midline still allows for the formation of the midline barrier. Therefore, we propose a novel SHH-independent function for cilia in coordinating Lefty1 expression and establishing the midline barrier. Furthermore, while cilia in the node are required for the establishment of the asymmetry, the midline cilia may be required for the establishment of the midline barrier and maintenance of L-R asymmetry. Thus, our interpretation of the experiments that reintroduce cilia in the crown cells of the node is that full recovery is achieved due to the migration of ciliated cells from the node into the midline, as previously observed in wild-type embryos (Botilde et al., 2013; Brennan et al., 2002; Sulik et al., 1994; Yoshiba et al., 2012).

3.4. Cilia in L-R patterning beyond the node and the midline

Hedgehog signaling has also been implicated in relaying the Nodal signal to the LPM (Tsiairis and McMahon, 2009). The model proposes that in Smo^null animals Nodal is lost from the LPM due to an overactive GLI3 repressor (GLI3R). Consistently, Smo^null; Gli3^null animals have normal L-R patterning. Furthermore, by introducing Smo back into the LPM of Smo^null animals, authors show that given proper signaling in the LPM, midline SHH signaling may be dispensable (Tsiairis and McMahon, 2009). Our results are largely in agreement with this model. In Tmem107^null animals, the phenotype is akin to loss of cilia, likely inhibiting Gli3R formation, thus making the LPM hypersensitive to NODAL. With the loss of cilia, SHH signaling and Lefty1 in the midline, NODAL is free to autoactivate in the LPM on both sides. Likewise, Tmem107^schlei embryos lack cilia in the LPM, disrupting SHH signaling in that region. However, the presence of cilia, Lefty1 and a functional midline barrier overcome the potential hypersensitivity to NODAL in Tmem107^schlei embryos.

Our genetic interaction experiment with Tmem107 and Pkd2 did not yield a clear answer of whether the remaining crown cilia in Tmem107^null embryos are functional. However, it does contribute insight towards the investigation of ciliary functions beyond the node. Pkd2^null; Tmem107^null animals present with a left pulmonary isomerism phenotype, like the single Tmem107^null animals. This phenotype is also similar to Pkd2^null; Kif3a^null animals, which lack all cilia, and have been reported to have bilateral Pitx2 expression, a molecular precursor to left pulmonary isomerism (Grimes et al., 2016). Furthermore, bilateral expression of Nodal is also the prevailing phenotype observed in mutants with a complete loss of cilia (Murcia et al., 2000; Takeda et al., 1999). Given all of this evidence, Tmem107^null appears to behave as a loss of cilia mutant. However, some ciliary functionality is maintained in the embryo, since the mutants in our study survive until at least E16.5, while mutants with a complete loss of cilia, like Kif3a and IFT88 die before E11.5 (Murcia et al., 2000; Takeda et al., 1999).

The experiment with the Pkd2^null; Tmem107^null animals advances our understanding for the roles of cilia in the LPM and the midline. Models predict that LPM on both sides may be receiving the NODAL signal at the onset of leftward nodal flow (Nakamura et al., 2006). It has been proposed that cilia may provide an inhibitory signal (like GLI3R) to the propagation of NODAL in the LPM, and the role for PC2 may be to induce signaling that would alleviate that inhibition (Grimes et al., 2016; Tsiairis and McMahon, 2009). Our results agree with this interpretation of the mutant phenotypes, with one caveat. Crown cilia were proposed as the most important contributors in this inhibition/release of inhibition signaling (Grimes et al., 2016), but our data go against this notion, since we found no abnormalities in specification of crown cilia and their ciliary localization of PC2. Instead, it may be a combinatorial mechanism, with the crown cells travelling into the midline, and the midline cilia and the midline barrier playing indispensable role in limiting hyperactivation of NODAL sensitivity in the LPM.

4. Summary

Several mouse mutants of transition zone proteins have been reported to have various L-R patterning defects, including left isomerism or bilateral expression of “leftness” markers in mouse mutants of Tmem67, Mks1 and Rpgrip1L (Abdelhamed et al., 2015; Christopher et al., 2012; Vierkotten et al., 2007; Weatherbee et al., 2009). While some investigation has been carried out into uncovering the mechanism behind left isomerism in mutants lacking cilia, less is known about mutants with a partial loss of cilia. Mouse mutants of transition zone proteins typically have some cilia, albeit with defective ciliary protein composition (Christopher et al., 2012; Dowdle et al., 2011; Roberson et al., 2015; Shylo et al., 2016; Weatherbee et al., 2009). Our finding here that Tmem107^null animals have left pulmonary isomerism is consistent with the fact that cilia are ubiquitous organelles, and a transition zone is required for the formation of a functional cilium. Although at early stages Nodal is randomized at the node in Tmem107^null, the failure of the midline barrier allows NODAL to cross the midline and induce its expression in the opposite LPM, always leading to its bilateral activation at later stages, and subsequent left isomerism. Tmem107^schlei animals, on the other hand, provide us with a natural conditional-like mutant, where complete loss of cilia occurs only in the LPM, and the cilia in the nodes and the midline look normal, resulting in normal L-R patterning.

The two mouse mutants for ciliary transition zone protein TMEM107 presented here can be invaluable assets in studying the detailed roles of cilia in establishing L-R asymmetry, and may help address questions about differential requirements for ciliary proteins in different tissues, despite the ubiquitous nature of this organelle. Our data also hint at the potential roles for cilia in establishing and maintaining L-R asymmetry outside the node, consistent with observations made by others (Grimes et al., 2016; Tsiairis and McMahon, 2009).

5. Materials and methods

5.1. Mouse strains

Mouse experiments were performed in accordance with Yale Institutional Animal Care and Use Committee guidelines. The Tmem107^tm1Lex mouse line, herein referred to as Tmem107^null, has been previously described (Christopher et al., 2012; Tang et al., 2010). The mouse line was initially generated on the C57BL/6N background, and the analysis in this paper has been performed on a mixed C57BL/6J; C57BL/6N background. Tmem107^schlei (Christopher et al., 2012) mutant line was crossed for at least 7 generations onto the C57BL/6J background prior to characterization. Dnah11^GFP (Lrd^GFP) (McGrath et al., 2003) mice were a gift of Martina Brueckner, and are maintained on a mixed genetic background of unknown composition. Pkd2^tm2Som mice (Wu et al., 1998) are maintained on a mixed genetic background of unknown composition. Genotyping primers and strategies are available in Supplementary Material, Table S1.

Timed matings were set up by pairing up male and female mice in the evenings. The following morning females were checked for the presence of vaginal plugs. If a plug was detected, noon on that day was designated embryonic day E0.5. Females were euthanized, and embryos were collected at E8.0-E16.5. Embryos were staged morphologically as E8.0 starting at the late head fold, and as E8.5 once they had 6 somites, and the heart had begun to loop.

5.2. Expression analysis

In situ hybridization analyses were performed using standard methods (Nagy, 2003). Embryos were collected in sterile PBS, and yolk sac was used for genotyping. Mutant and control embryos of appropriate stages were pooled from multiple litters into corresponding groups. The two groups were treated identically throughout the protocol. The timing to stop the alkaline phosphatase reaction was empirically determined, but was identical for mutant and control embryos in the same experiment.

Antisense probes were generated from the following cDNA plasmids: Nodal (Elizabeth Robertson), Dand5 (Martina Brueckner), Lefty1 (Hiroshi Hamada), Pitx2 (Axel Schweickart), Ptch1, Shh and FoxA2 (Andrew McMahon).

The Tmem107 riboprobe plasmid was generated by RT-PCR from C57BL/6J RNA with subsequent cloning into pCRII-TOPO and contains full-length mouse Tmem107 (NM_025838.2) coding sequence. In situ hybridization with Tmem107 sense probe showed no visible staining (not shown).

5.3. Wholemount immunostaining

Embryos were collected at E8.0-E8.5 and fixed in 4% paraformaldehyde (PFA) for 20 min at room temperature. Embryos were washed 3 times (15 min each) in phosphate buffered saline (PBS), then blocked with 1% goat serum and 0.1% Triton-X in PBS (PBGT) for 1 h at room temperature. Primary antibodies in PBGT were added and incubated overnight at 4 °C. After 315 min washes in PBGT, secondary antibodies in PBGT were added for 2 h at room temperature. Then, samples were washed 3 times (15 min each) in PBS and 0.1% Triton-X, and mounted on slides (Denville) using ProLong Diamond Antifade Mountant (Life Technologies). Antibodies used are available in Supplementary Material, Table S2.

Alternatively, embryos previously dehydrated in methanol series were rehydrated, blocked in PBGT, and the protocol was followed as described above.

To acquire sections through the midline, embryos were stained as wholemount and imaged, then rinsed 3 times in PBS and 0.1% Triton-X, dehydrated in 30% sucrose, embedded in OCT and cryosectioned. Coverslips were applied to sections with ProLong Diamond Antifade Mountant.

5.4. Imaging and image analysis

Images of organs and post-in situ hybridization embryos were taken on a Zeiss SteREO Discovery microscope. Wholemount immunofluorescence images were captured using Leica SP5 inverted confocal microscope using Leica Application Software. Immunofluorescent images of sectioned embryos were acquired on Carl Zeiss LSM 700 confocal microscope.

Image analysis was performed in ImageJ (NIH). Specifically, cilia were counted using the point tool, and their length was determined using the line tool and known image scale. Optical sections through midlines were acquired using Reslice function, and then a Z-stack projection was created from the region of interest. Resulting images were slightly stretched out in dorsal-ventral direction for easier visualization.

Graphs created and statistical analysis regarding cilia numbers and length were carried out using Prism software.

5.5.

Key resources table

Reagent or resource	Source	Identifier
Antibodies
Rabbit polyclonal anti-PC2	Stefan Somlo (Cai et al., 1999)
Rabbit polyclonal anti-ARL13B	Tamara Caspary (Caspary et al., 2007)
Mouse monoclonal anti-ARL13B	UC Davis/NIH NeuroMab	ab136648
Chicken polyclonal anti-GFP Secondary antibodies: See Supplementary Methods Bacterial and Virus Strains	Abcam	ab13970
Biological Samples Mouse embryos	This paper, see Experimental Models
Chemicals, Peptides, and Recombinant Proteins Critical Commercial Assays
Deposited Data
Experimental Models: Cell Lines
Experimental Models: Organisms/Strains
B6; 129S5(Cg)-Tmem107tm1Lex/Mmucd	Lexicon (Tang et al., 2010)	MGI:5433209
Pkd2tm2Som Genetic background: Involves: 129/Sv*129S2/ SvPas*C56BL/6*SJL	Stefan Somlo (Wu et al., 1998)	MGI:1860866
B6; C3Fe.B6-Tmem107schlei	This study, (Christopher et al., 2012)	MGI:5433051
Dnah11tm11Bknr Genetic Background: Involves 129S1/Sv*FVB/N*C56BL/6 Oligonucleotides See Supplementary Methods Recombinant DNA	Martina Brueckner (McGrath et al., 2003)	MGI:3623 683
In situ construct: Tmem107-pCRII- TOPO	This paper
Full-length mouse Tmem107 (NM_025838.2) Software and Algorithms
Carl Zeiss AxioVision Product Suite	Zeiss
Other