A Posterior Centre Establishes and Maintains Polarity of the Caenorhabditis elegans Embryo by a Wnt-Dependent Relay Mechanism

Cellular polarity is a general feature of animal development. However, the mechanisms that establish and maintain polarity in a field of cells or even in the whole embryo remain elusive. Here we provide evidence that in the Caenorhabditis elegans embryo, the descendants of P1, the posterior blastomere of the 2-cell stage, constitute a polarising centre that orients the cell divisions of most of the embryo. This polarisation depends on a MOM-2/Wnt signal originating from the P1 descendants. Furthermore, we show that the MOM-2/Wnt signal is transduced from cell to cell by a relay mechanism. Our findings suggest how polarity is first established and then maintained in a field of cells. According to this model, the relay mechanism constantly orients the polarity of all cells towards the polarising centre, thus organising the whole embryo. This model may also apply to other systems such as Drosophila and vertebrates.


Introduction
Many developmental processes depend on polarised cells; cell movements and cell divisions are oriented [1][2][3] and also the specification of cell fates is influenced by polarity [4][5][6]. The polarity of individual cells has been studied in Drosophila [7,8] and Caenorhabditis elegans [9,10], but the establishment and maintenance of polarity in a field of cells is still mysterious [11,12].
It is a general notion that during development of the C. elegans embryo, cell fate specification [13,14] and division orientation [15] are coordinated strictly in the anteriorposterior (a-p) direction. The first cleavage generates an asymmetric 2-cell stage embryo with an anterior AB and a posterior P 1 blastomere [15]. Most of the cells (70%) of the hatching embryo are derived from the AB blastomere. The first quantitative 4D-microscopic analyses of cleavage directions in the C. elegans embryo showed that AB-derived cells do not cleave strictly along the a-p axis but deviate, on average, approximately 36 8 from the a-p axis until the premorphogenetic stage (256-AB cell stage) of embryogenesis [16]. During early development from the 4-AB to the 64-AB cell stage, the deviation from the a-p axis is even higher (approximately 45 8 6 20 8; see Figure S1 for examples of cleavages in a normal embryo).
Goldstein reported that in vitro contact with P 2 , the posterior daughter of P 1 , orientates the mitotic spindle in its anterior sister EMS and that P 3 , the posterior daughter of P 2 , does the same to its anterior sister E [17]. Besides the influence of P 1 descendants on EMS cleavage orientation, it was also shown that the orientation of the ABar division depends on contact with the C blastomere, the anterior daughter of P 2 [18]. Furthermore, the combination of blastomeres in vitro demonstrated that also individual ABderived cells of the 12-cell stage embryo align their mitotic spindles towards MS and C [19]; however, the directions of the next cleavages were not reported. Spindle orientation, as well as cell fate specification, was shown to depend on Wnt signalling in C. elegans [19][20][21][22].
To investigate cellular polarity in AB-derived cells of the C. elegans embryo, we performed blastomere combination experiments in vitro [23,24]. The generated embryonic fragments were analysed by 4D-microscopy, which allowed us to monitor cell positions, mitoses, movements, and division orientation of all cells during development (see Materials and Methods and [16,25]).
Here we show that P 2 and its descendants introduce a polar bias into the cleavages of AB-derived cells, which causes a shift of the average division angle with respect to the a-p axis from 62 8 to 45 8. This polarisation depends on a MOM-2/Wnt signal originating from P 2 and its descendants. Moreover, we provide evidence that P 2 and its descendants constitute a polarising centre, which signals throughout early development. Furthermore, we show that the Wnt signal is transduced from AB descendant to AB descendant by a relay mechanism.

Results/Discussion
Elongation of AB-Derived Embryonic Fragments Depends on the Presence of P 2 AB blastomeres isolated from P 1 at the 2-cell stage show an abnormal ''spiralian-like'' cleavage pattern, in which the direction of cleavages rotates approximately 90 8 after each division [26,27]. This is due to the migration pattern of the centrioles, which always move orthogonally out of the plane defined by the last two divisions [28]. The developing embryonic fragments have a spherical shape (Table 1, row A; Figures 1A and 2D). This suggests that AB descendants do not have any intrinsic polarity to orient their divisions in the normal a-p direction [15,16] and that the a-p polarity of these cells is induced after the 2-cell stage. This appears to disagree with Park and Priess, who showed that sister pairs of 16-AB and 32-AB cell stage AB blastomeres display an asymmetric expression of POP-1, although they never touched P 2 or its descendants [19]. However, we recently showed that isolated AB blastomeres display a-p asymmetries in cell fates [27]. Thus, the POP-1 asymmetry described by Park and Priess might indicate an asymmetry in cell fates rather than in the polarisation of cleavage direction.
When a P 2 blastomere is added to the two daughters of isolated AB blastomeres, the developing AB-derived embryonic fragments elongate and become twice as long as they are wide (Table 1, row B; Figure 1A). The eight regions deriving from the eight AB descendants of the 12-cell stage embryo [25] are also elongated. This finding contrasts with the embryonic fragments without P 2 , in which the regions are approximately spherical ( Figure 1A). In the elongating embryonic fragments, the cells do not divide ''spiralian-like,'' but orient towards the added P 2 blastomere (white box in Figure 1C). This suggests that the observed elongation is due to cleavage orientation.

Elongation of Embryonic Fragments Results from Oriented Divisions
We then tested whether the orientation of cell divisions was the main cause of the elongation. We used SIMI Biocell to record the 3-D coordinates of the mother cells and then those of the daughters 105 s after the initiation of the cleavage furrow, and we calculated the cleavage angles of all cells relative to the a-p axis (see Materials and Methods). A comparison of division angles between isolated AB blastomeres and AB blastomeres cultured with an added P 2 blastomere showed a significant difference (p , 0.001, Student's t-test). In isolated AB blastomeres, average divisions deviated 63 8 6 20 8 (n ¼ 3 embryonic fragments) from the a-p axis, whereas in embryonic fragments with P 2 , divisions deviated on average 45 8 6 22 8 (n ¼ 3) from the a-p axis.
To assess cleavage directions quantitatively, we projected the division angles onto a target screen (Figure 2A) (''Lö we projection'', named after the mathematician Harald Lö we, Braunschweig, who kindly suggested this method). The Lö we value is a measure of the deviation of cleavages from the a-p axis in all three dimensions. An isolated AB-derived fragment demonstrated a Lö we value of 6.3 6 9.8, which corresponds to the mean angle of 62 8 6 23 8 with respect to the a-p axis. After addition of P 2 , the Lö we value decreased to 2.9 6 4.4 and the angle to 44 8 6 22 8 ( Figure 2D and 2E).
The direction of cleavages correlates very well with the variation of shape (elongation) of the embryonic fragments; the statistical correlation between elongation, the average cleavage angle, and the Lö we value are better than 0.92 (p , 0.005; considering 2,200 cell divisions). Thus, we used the shape of embryonic fragments to assess division orientation in the large number of embryonic fragments analysed (120 fragments in 30 experiments; Table 1). For many of the experiments, the Lö we projections shown allow a more comprehensive comparison.
To assess the contribution of cell movements between cleavages, we removed these movements from our dataset using an algorithm described in [16]. In this way, only the contribution of mitoses is left and can be seen in the 3-D representations of the developed fragments ( Figure 1). Even after this procedure, embryonic fragments including P 2 were approximately two times longer than those derived from only AB blastomeres ( Figure 1A). Thus, the shape of the embryonic fragments depends almost entirely on the cleavage directions of cells. Our measurements show that the addition of P 2 to an AB blastomere causes a significant alteration of cleavage orientation of AB-derived cells of the embryonic fragment which, in turn, leads to an elongation.
The analyses also indicate that the alteration of cleavage direction is not limited to the AB descendants that touch P 2 , but occurs in all AB descendants (white box in Figure 1C). P 2 biases division orientation from a ''spiralian-like'' pattern observed in isolated AB blastomeres (corresponding to an average angle of approximately 60 8; i.e., the mean of three cleavages with angles of 0 8, 90 8, and 90 8 with respect to the ap axis) to 45 8 6 22 8. This result is very similar to the 45 8 6 20 8 (n ¼ 3) observed for the cleavages in early development of normal embryos ( Figure 2B and 2C).
At first sight, our results differ from studies in C. elegans that have shown that isolated AB-derived blastomeres of the 12cell stage embryo [19] and the EMS blastomere [17] divide directly towards an added P 2 blastomere (0 8) in their first division [22]. However, in these studies, the behaviour of cells in subsequent divisions was not discussed. When we analysed division orientation in normal embryos between the 4-AB and the 64-AB cell stage (the first two divisions of AB occurring perpendicular to the a-p axis), we observed that AB-derived cells divide rarely 0 8 but on average 45 8 6 20 8 with respect to the a-p axis (e.g., Figure S1 and Figure 2B and 2C). Also, in vivo (in contrast to in vitro [17]), EMS does not divide directly towards P 2 . In the five normal embryos we analysed, the divisions occurred on average 35 8 6 6 8 off the a-p axis. Thus, our observations in embryonic fragments are consistent with normal embryonic development.
In normal development, an average cleavage angle of 45 8 may reflect the optimal compromise between minimising the need for compensatory cell movements to less than one embryo length per cell [16] while still allowing a stringent a-p assignment of cell fates [15]. Strict a-p cleavages would displace cells 40 times the embryo length in eight cleavages (see Protocol S1 for further discussion).

Cleavage Angles Are Independent of Cell Fate
Because the addition of P 2 induces an ABp fate in the daughters of AB [29], and because ABp-derived regions are longer than those derived from ABa [25], the elongation of embryonic fragments could merely reflect the presence of ABp-derived fates; i.e., the bias in division orientation might EI is the elongation index of fragments calculated by dividing the length by the width of the fragments in the 3-D representations ( Figure S2). If P 2 was not present, the a-p axis was placed orthogonal to the first AB spindle so that it intercepts the contact point where P 2 was normally added. The genotype of blastomeres is indicated in square brackets to the right of the blastomeres. Ablated blastomeres are underlined. Significance differences performing one-way analysis of variance (ANOVA) (StatView, Cary, North Carolina, United States): * indicates p , 0.05; ** indicates p , 0.001; capital letters and superscripts in the ANOVA column identify the pairs compared in post-hoc tests. Accolades indicate summarized groups for statistics. Both elongated and nonelongated fragments in the ablation experiments are shorter than the corresponding non-ablated experiments. EIs of embryonic fragments (mean 6 SD, n ¼ sample size) are shown. Capital letters help to navigate in the table. A scheme of the performed experiment can be found in the Design column (AB, white; P 2 , black; ablated blastomeres are crossed out). The shape of the embryonic fragments compared to the nonelongated control experiment is shown in the Interpretation column (Ñ, elongated; u, nonelongated). In a few experiments in which embryonic fragments elongate, the EIs do not only differ significantly from the negative but also from the positive control (indicated by brackets in the Interpretation column). This is due to a larger variance between the single experiments rather than to a lack of elongation. a The two AB-derived parts per embryonic fragment were considered separately in the statistical test. b Summary of all three alleles used.   [25], were colour coded independently of their fate to show the topology of each region. P 1 descendants are colour coded according to their fate. The right column shows 3-D representations of the same generation that illustrate how embryonic fragments look if the contribution of cell movements is removed using an algorithm described in [16] so that only the contribution of mitoses is left (''mitoses only''). In (C) 3-D representations of the sixth generation of cells (16 AB descendants) are shown. To indicate the orientation of the cleavages of their mother cell, the two sisters are connected with a white bar shortly after mitosis.
(A) Development of isolated AB blastomeres (the two ABx daughters are shown on the left) compared to AB blastomeres to which a P 2 blastomere was only be due to the execution of specific cell fates. We therefore prevented the ABp induction by using different combinations of blastomeres in which either the responsible receptor GLP-1/Notch or its ligand APX-1/Delta were mutant [29], and we then analysed the elongation of these embryonic fragments. Below, we refer to blastomeres from embryos whose mothers were homozygous for a mutation, for example in glp-1, as glp-1 blastomeres. When glp-1 AB and P 2 blastomeres are combined, cells mainly divide away from P 2 and the embryonic fragments elongate ( Table 1, row C; Figures 1A and 2F). The same is true when other blastomere combinations are used that also prevent the induction of ABp (Table 1, row C). Thus, the induction of ABp-derived fates by P 2 is not the reason for the elongation of the fragments. Based on these results, we instead suggest that P 2 and/or its descendants influence division orientation by being the source of a polarising signal.

Orientation of Cleavages Depends on a Wnt Signal Originating from P 2
Earlier studies have shown that MOM-2/Wnt is involved in spindle orientation in EMS and in individual AB descendants [19][20][21][22]. To test if the polarisation of embryonic fragments depends on the MOM-2/Wnt ligand, we generated chimeric fragments by combining wild-type AB and mom-2 P 2 blastomeres. Three different mutant alleles interfere with the elongation of the fragments (Table 1, row D). Fragments in which P 2 was derived from mom-2(or9) do not elongate at all, because the divisions of the AB-derived blastomeres and their descendants were not aligned towards P 2 ( Figures 1B,  1C, and 2G). These results suggest that MOM-2/Wnt is the relevant signal given out by P 2 or its descendants. The polarisation also depends on the MOM-5/Frizzled receptor [20,21] in AB but not in P 2 ( Table 1, row E). That MOM-2 and MOM-5 organise division orientation can also be seen in mutant embryos. Cell divisions in AB descendants between the 4-and 64-AB cell stage deviate from the a-p axis by 55 8 6 19 8 in mom-2(or9) embryos and by 55 8 6 20 8 in mom-5(zu193) embryos; this is significantly different from wild-type embryos (40 8 6 20 8 mean deviation from the a-p axis, embryos #2 and IBþRS [16]; Student's t-tests: p , 0.001 [n ¼ 2 each]).

A Posterior Polarising Centre
By analysing the polarising signal of P 2 and its descendants in detail, we found that these cells display properties of a polarising centre. Regardless of whether P 2 is added in a 45 8 or a 90 8 angle (as compared with the normal experimental conformation), P 2 always defines the posterior pole of the fragment and induces an elongation (Table 1, row F). Also, embryonic fragments consisting of P 2 and the granddaughters of AB (instead of the daughters of AB) elongate, suggesting that polarisation does not depend on the establishment of contact between blastomeres at a particular developmental stage ( Table 1, row G). The signal of P 2 acts in all directions, because the addition of two AB blastomeres to one P 2 blastomere on opposite sides leads to an elongation of both AB-derived parts ( Table 1, row H). In an embryonic fragment consisting of EMS, AB, and P 2 , the descendants of AB and EMS orient towards the descendants of P 2 , showing that P 2 and its descendants are dominant over the P 2 sister EMS and its descendants ( Figure S3A). However, in the absence of P 2 , the EMS blastomere has a polarising activity itself (Table 1, row I) which was reported before by Park and Priess [19]. The addition of two P 2 blastomeres to the daughters of one or two AB blastomeres from opposing directions leads to two elongated structures with opposing polarities ( Figure S3B). Adding one of the daughters of P 2 (either C or P 3 ) to the two AB daughters causes an elongation of the embryonic fragment, whereas the addition of another AB daughter (which never touched P 2 ) does not (Table 1, row J). This shows that the two daughters of P 2 are able to polarise AB-derived embryonic fragments and that this effect is specific to P 2 and its descendants. This supports Park and Priess, who showed the same for single AB-derived blastomeres [19]. Taken together, these results show that in all combinations tested, P 2 and its descendants define the posterior pole of an elongated structure and orient the cell divisions of AB-derived blastomeres.
The following experiments also argue that P 2 and its descendants constitute a polarising centre. When two mutually perpendicular P 2 blastomeres were added to four AB daughters, two different patterns developed from the same starting configuration (Figure 3). Because in both experiments glp-1(e2144) blastomeres were used, any influence of different cell fates can be excluded. In one fragment, the descendants of the two P 2 blastomeres remained in their initial positions, resulting in an L-shaped structure with two a-p axes. In the other fragment, the descendants of one P 2 remained in their original position for 75 min, until after the P 3 division, and then moved and joined the descendants of the second P 2 . This resulted in an elongated structure with only one a-p axis. This finding indicates that the AB-derived blastomeres respond continuously to a polarising signal; they even reorganise their polar behaviour when the P 2 descendants move. This experiment again demonstrates that the descendants of P 2 provide the polarity signal, because not P 2 itself but only its descendants reached the posterior part of added after the two daughters were born. P 2 causes the embryonic fragment to elongate independently of whether only ABa or ABp fates are present. As becomes obvious from the colour code, the AB-derived regions stretch out along the axis defined by P 2 whereas they do not in isolated AB cells. The elongation is caused by the alignment of mitoses along the a-p axis [see white box in (C) and Figure 2D-2F]. A comparison of the original and the ''mitoses only'' representations reveals that the shape of the embryonic fragments is mainly due to mitoses. This suggests that a bias in mitosis direction is the cause of the elongation of the embryonic fragments. (B) Chimeric embryonic fragment consisting of a wild-type AB and a mom-2(or9) P 2 blastomere. This mom-2 allele prevents the elongation of the embryonic fragment completely. Like in an isolated AB blastomere (A), the eight AB-derived regions are not elongated, indicating that the mitoses of all AB-derived cells were not mainly oriented towards the P 2 blastomere ( Figure 2G). Again, the original and the ''mitoses only'' representations look very similar. (C) Orientation of cell divisions (fifth division of AB-derived blastomeres) in three embryonic fragments where wild-type AB and mom-2 P 2 blastomeres were combined [embryonic fragment 1 also shown in (B)] and a wild-type control (white box). In contrast to the wild-type control embryonic fragment, divisions are not directed towards the position of P 2 and its descendants. For a quantitative analysis of division angles of embryonic fragments, see Figure 2. Furthermore, the shown wild-type embryonic fragment exemplifies that AB-derived cells that do not touch P 2 also orient their divisions towards P 2 . DOI: 10 the embryonic fragment and caused its elongation. An experimental displacement of cells, recorded under the 4Dmicroscope while development is progressing appears to be impossible to us at this time. Thus, the coincidentally occurring movement of the P 2 descendants observed here was a unique chance to examine the consequences of the displacement of signalling cells during development.

The Polarising Signal Reaches Cells Not Touching P 2
Some of the experimental results imply that cells that are not directly touched by P 2 or its descendants are also affected by the polarising signal. To test this, we combined two pairs of daughters of AB blastomeres with one P 2 blastomere so that only one pair of AB daughters touched P 2 . To ensure that P 2 did not induce the ABp fate in the touched AB-derived blastomeres, glp-1 blastomeres were used. The descendants of both AB blastomeres form elongated structures, in contrast to the controls without P 2 ( Table 1, rows K and L). Thus, the polarising signal is transported to AB-derived blastomeres not touching P 2 .

How Is the Polarising Signal Transported?
It is not clear how signalling molecules are transported [30]. A secreted molecule could be transmitted actively, it could each dividing cell is placed on the a-p axis and a vector is projected from the centre of the mother cell through the centre of the daughter cell. The point where this vector hits a target screen is marked by a dot (the distance between mother cell and screen along the a-p axis is normalised to 1). The red semi-circles represent particular division angles (as indicated). Embryos were analysed from four to 64 AB-derived cells. For statistical analysis, the mean distance of the dots from the target centre and the standard deviation are calculated. This ''Lö we value'' and its standard deviation give a reasonable measure of the general direction of cell cleavages. This is exemplified by showing nine cleavages of a ''polarised'' normal embryo (blue) and a ''nonpolarised'' isolated AB blastomere (green). The mean is indicated by a circle and the standard deviation by a dotted circle. Note that the representation of cleavage angles is not linear, because the distance from the centre to a dot corresponds to the tangent of the cleavage angle to the a-p axis. Thus, angles close to 90 8 cannot be shown on the target screen (here, this is the case for two cleavages of the ''nonpolarised'' isolated AB blastomere). It becomes obvious that for the normal embryo, in which cells divide mainly along the a-p axis, most dots are located in close proximity to the centre and the corresponding Lö we value is low. In contrast, the dots are widely scattered for the isolated AB blastomere, in which cells do not mainly divide in the a-p direction, and the Lö we value is high. (B-G) ''Lö we projections'' of two normal embryos and the embryonic fragments shown in Figure 1A and 1B. The red circles correspond to the red semicircles in (A). In the lower left corner, Lö we values and mean division angles are shown. (B and C) Normal embryos #1 and #2 from [16,25]. Cells mainly divide in an a-p direction. The Lö we value and especially its standard deviation are low. If a third normal embryo (not shown) is also considered, normal embryos have Lö we values between 1.3 and 1.6 with standard deviations ranging from 3.0 to 4.2. The mean cleavage angles vary between 45 8 and 50 8, the standard deviations between 20 8 and 23 8. (D) Isolated AB blastomere. Compared to normal embryos (B and C) and AB blastomeres where a P 2 blastomere was added (E), the division angles scatter all over the target area; i.e., cells do not mainly divide in the a-p direction. This is reflected in the higher Lö we value and especially the high standard deviation. Five points are off the target screen due to 90 8 divisions.
(E) AB blastomere where P 2 was added. Cells mainly divide in the a-p direction, and the Lö we value and its standard deviation decrease noticeably compared to the isolated AB blastomere in (D). (F) AB blastomere where P 2 was added in a glp-1(e2144) background. Again, cells mainly divide in the a-p direction. (G) Addition of a mom-2(or9) P 2 blastomere to a wild-type AB blastomere. Cells do not mainly divide in the a-p direction. The Lö we value is higher than in embryonic fragments dividing mainly in the a-p direction and its standard deviation resembles the standard deviation of the isolated AB blastomere. DOI: 10.1371/journal.pbio.0040396.g002 spread passively by diffusion, or it could be passed on from cell to cell by relay. In a relay mechanism, it would be produced newly in each cell in response to the signal from the neighbouring cell [31,32] (Figure 4A). To discriminate among these possibilities, we put a mom-2 AB blastomere between a wild-type AB and an apx-1 P 2 blastomere. If the signal were transduced via a relay mechanism, the mom-2 blastomere would inhibit further signalling, because it could not produce a MOM-2/Wnt dependent signal. In contrast, an actively or passively transported signal should still reach the AB descendants ( Figure 4B). In these experiments, apx-1 or glp-1 blastomeres were again used to obtain only ABa-derived fates. To prevent cells of the different genotypes from intermingling at a later stage, we laser irradiated the two mom-2 AB daughters after 10 min to prevent further divisions. The cells were not killed ( Figure 4D and 4E). Control experiments showed that laser ablation did not hinder the transmission of the polarity signal (Table 1, rows M and N; Figure 4F and 4G). We found that the wild-type AB-derived parts of the fragments, which never touched P 2 , did not elongate when a mom-2 AB blastomere was placed between a wild-type AB and an apx-1 P 2 blastomere (Table 1, row O; Figure 4H). Moreover, the directions of mitoses deviated strongly from the axis defined by P 2 , indicating that the cells were not polarised. The Lö we projections of the cleavage angles for this experiment and the controls are shown in Figure 4F-4H. These experiments suggest that the polarising signal is relayed from cell to cell in C. elegans. The MOM-2/Wnt signal is not farranging but must be produced anew as each neighbouring cell receives the signal; this is not possible in the mom-2 mutant blastomere.
Placing a mom-5 AB blastomere between a wild-type AB and an apx-1 P 2 blastomere also impairs the elongation of the wild-type part of the embryonic fragment (Table 1, row P). Embryonic fragments elongate less as compared with wildtype controls; however, the effect is not as extensive as in experiments using mom-2 blastomeres. These findings confirm that MOM-5 is needed in AB blastomeres to receive the MOM-2/Wnt signal.

Polarised AB-Derived Blastomeres Can Polarise Nonpolarised AB-Derived Blastomeres
We then sought further evidence that polarised AB descendants are indeed able to produce a polarising signal themselves. To show that the polarisation of neighbours is independent of the original source (i.e., P 2 or its descendants), we conducted the following technically difficult experiment ( Figure 5A). We added the two daughters of an AB blastomere to P 2 . After 5 min, we removed P 2 and added the now polarised AB blastomeres to freshly isolated AB daughters. The cleavages of these originally nonpolarised blastomeres were not nondirected as in an isolated AB blastomere, but were in many cases directed towards the polarised AB descendants ( Figure 5B). The descendants of the AB blastomere that was transiently in contact with P 2 show elongated AB-derived regions. This experiment demonstrates that AB-derived blastomeres that transiently touch P 2 become polarised and, furthermore, are able to transduce their acquired polarity to other nonpolarised AB-derived cells.

A Transient Contact of P 2 Is Sufficient to Polarise AB-Derived Blastomeres
To further corroborate the finding that AB-derived blastomeres can be polarised by transient contact of P 2 , we tested the influence of P 2 on single AB blastomeres. The behaviour of single AB-derived blastomeres is not different from two combined AB-derived blastomeres; in both, cells do not divide mainly in an a-p direction and the fragments have rounded shapes ( Figure 1A, 2D, and 5C; Table 1, rows A and K).
First, we analysed isolated AB blastomeres that touched wild-type P 2 transiently ( Figure 5D). In five of 12 experiments, the AB-derived fragment elongated (Table 1, row Q); in six fragments, it elongated weakly (Table 1, row R), and in one fragment, no elongation was observed. Second, we analysed AB blastomeres that transiently touched mom-2 mutant P 2 blastomeres, and we found that in all four cases, the resulting embryonic fragments did not elongate ( Figure 5E; Table 1, row S). These experiments confirm the finding above that a transient signal of P 2 can polarise AB-derived blastomeres. Furthermore, the latter experiment again suggests that MOM-2 is the polarising signal. It also shows that an induction of ABp fate by P 2 is not the reason for an elongation of the embryonic fragment, because the induction of ABp occurs also in mom-2(or9) mutant embryos (unpublished data).

Establishment and Maintenance of Polarity in the C. elegans Embryo
Our investigations argue that the a-p polarity of AB descendants in C. elegans is not an autonomous feature established early in embryonic development, but rather that this polarity is induced in AB descendants by P 2 for the first time at the 4-cell stage. We show that the division orientations of most of the AB descendants are biased towards P 2 and its descendants (Figures 2 and 5) and that AB-derived cells continuously orient their divisions towards it, even when these signalling cells move during development ( Figure 3). This suggests that P 2 and its descendants constitute a polarising centre in the posterior of the embryo. We further show that MOM-2/Wnt-which has already been implicated in certain aspects of polarity of the early C. elegans embryo [19][20][21][22]-is the polarising signal (Figure 1 and 2; Table 1, row D). Moreover, we demonstrate that the MOM-2/Wnt signal is transduced from AB blastomere to AB blastomere by a saltatory relay mechanism (Figure 4).
We restricted our experiments to the AB-derived part of the embryo. An interesting question is how polarity is established and maintained in the remainder of the embryo derived from P 1 . To address this question, we compared the cleavage angles in wild-type and mom-2(or9) embryos starting at the 8-cell stage embryo for the next four cleavage rounds. The P 1 -derived cells from MS, E, C, and D cleave with an average angle of 35 8 6 21 8 in the wild-type embryos as compared to 57 8 6 21 8 in the mom-2 embryos. Thus, a lack of MOM-2 affects the cleavage orientations in the P 1 -derived cells as significantly as in the AB-derived cells (Student's ttests: p , 0.001 [n ¼ 2 each]). It may therefore be that the cleavages of all cells of the embryo are oriented by the same mechanism. Then the polarising activity of MS and C [19] would not be an autonomous feature of these blastomeres but would reflect a polarisation they acquired directly from the P 2 daughter, P 3 (in the case of C), or indirectly through a relay mechanism (in the case of MS).
Taken together, our findings might explain how polarity in the C. elegans embryo is maintained during ongoing development; after the identity of P 2 has been specified by the par genes [33]-based on the initial polarisation of the embryo during fertilisation [34]-P 2 might induce polarity in its neighbours ABp and EMS via a MOM-2/Wnt signal through direct contact [19,22]. Afterwards, polarity of all AB descendants might be continuously maintained by the posterior polarising centre whereas the saltatory relay mechanism would ensure that the polarity signal reached all AB-derived cells.
Because there is emerging evidence that the MOM-2/Wnt signalling pathway in the C. elegans embryo shows some similarity to the planar cell polarity pathway [19,22,35], our  (Table 1, rows M and N). (F) Control experiment without P 2 . No elongation of the nonablated part of the AB-derived embryonic fragment occurred. (G) Control experiment with P 2 glp-1 background to prevent induction of ABp fates. The mitosis directions of the fifth division, which produces the sixth generation of cells, in the nonablated part of the AB-derived embryonic fragment indicate a polarisation of the blastomeres. The fragment elongates. This experiment shows that laser ablation does not hinder a potentially diffusing or transported signal. (H) Two mom-2 AB-derived daughters between P 2 and wild-type AB-derived cells prevent elongation (   Figure 2A for explanation). Right column: 3-D representations from which the contribution of cell movements was removed (''mitoses only'', see legend to Figure 1). In all experiments presented, these ''mitoses only'' representations resemble the original embryonic fragment, which suggests that the shape of the embryonic fragments is mainly due to cleavage orientation. (B) Combination of an isolated AB blastomere (ABa) and a blastomere which touched P 2 transiently (ABp). Four of the eight AB-derived regions of the ABp blastomere (right side of embryonic fragment; blue, violet, dark yellow, and light yellow) elongate comparable to regions in the experiment where findings may provide a paradigm of how polarity can be organised in a field of cells in other organisms, such as planar cell polarity in Drosophila [11,12] or convergent extension [6] and gastrulation [3] in vertebrates.

Materials and Methods
Strains and genetics. Worms were grown at 25 8C as described [36]. The following strains and mutations were used: N2 Bristol [36].
In vitro culture of blastomeres. The in vitro culture of embryos was described previously [27]. Embryos were cultured in embryonic growth medium supplemented with egg yolk [23]. Unless noted otherwise, a P 2 blastomere was added to two ABx blastomeres (derived from one AB blastomere divided once) in a way that both ABx blastomeres were touched by it. Two ABx blastomeres were used to ensure that cells of the same embryonic stage were added to P 2 .
Lineage analysis. The 4D-recordings were analysed using the database SIMI Biocell (SIMI Reality Motion Systems, Unterschleissheim, Germany; http://www.simi.com) [16,25]. Cells were followed and their 3-D coordinates were saved on average every 2 min. Cell cleavages were assessed by marking the position of the mother cell immediately before the cleavage furrow ingresses and the position of its two daughters three frames (105 s) later. By following every cell in the recording, the complete cell lineage of an embryo or an embryonic fragment is built. We only show 3-D representations of the eighth generation (64 AB descendants), because the elongation can be observed best here. When necessary, fates were determined as described in [27]-e.g., to assess if an induction of the ABp fate by P 2 occurred or not.
Calculation of division angles. The software for the quantitative analysis of cleavage orientations was written in Borland Delphi (Borland Software Corporation, Cupertino, California, United States). Division angles relative to the a-p axis were calculated using the straight line defined by the 3-D coordinates of the mother cell and its posterior daughter 105 s after the cytokinesis furrow starts to ingress. Angles of all cleavages from the 4-to the 64-AB cell stage were assessed. For the definition of the a-p axis in embryonic fragments, see Figure S2. Divisions were always described using angles from 0 8 to 90 8, which allows us to use linear statistics with our data.  . The a-p axis is defined by P 2 and the AB-derived blastomere, which is placed farthest from P 2 . This axis is always assessed at the time point at which the measurement is taken (at the 64-AB cell stage), because P 2 and its descendants move during development (due to their and their neighbours' divisions) and cells orient continuously towards P 2 and its descendants. Found at DOI: 10.1371/journal.pbio.0040396.sg002 (766 KB TIF). Figure S3. Development of Embryonic Fragments in which Two P 2 Descendants Were Added to AB-Derived Blastomeres (A) P 2 overcomes the polarising activity of EMS descendants. The regions are oriented only towards P 2 ( Table 1, row I). The EMS descendants move from an anterior to a lateral position during development (bent arrow). (B) Two P 2 blastomeres, which were added to the two AB daughters from opposite sites, induce two elongated structures with opposing polarity (white arrows). If the number of cells is increased by using two AB blastomeres, then again two elongated structures form (white arrows). However, the cells which are further away from the P 2 blastomeres loop out to form one common axis (red arrow) which indicates, as discussed in the main text, that cells permanently interact to define the a-p axis. For details see legend to