SEARCH

SEARCH BY CITATION

Keywords:

  • cell migration;
  • proneural genes;
  • mitotic quiescence;
  • BrdU;
  • neuromasts;
  • primordium

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

The sensory organs of the embryonic lateral line system are deposited by migrating primordia that originate in the otic region. Here, we examine the pattern of cell proliferation in the posterior lateral line system. We conclude that three phases of cell proliferation are involved in the generation of this system, separated by two phases of mitotic quiescence. The first phase corresponds to generalized proliferation during gastrulation, followed by a first period of quiescence that may be related to the determination of the lateral line precursor cells. A second phase of proliferation takes place in the placode and migrating primordium. This region is organized in annuli that correspond to the expression of proneural/neurogenic genes. A second period of quiescence follows, corresponding to deposition and differentiation of the sensory organs. The third period of proliferation corresponds to continued renewal of hair cells by division of support cells within each sensory organ. Developmental Dynamics 233:466–472, 2005. © 2005 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Cell migration is an important component of organogenesis in vertebrates. Many populations of cells undergo controlled migration during normal development, from neural crest and germ cells to specific neuronal populations. Conversely, uncontrolled migration is associated with various pathological conditions, most notably the establishment of metastases. Another major component of organogenesis is controlled cell proliferation, and alterations in this control also have dramatic pathological effects. Here, we investigate the relation between cell migration and cell proliferation in the particular case of the development of the zebrafish lateral line.

The lateral line system comprises a set of sense organs, the neuromasts, arranged under the skin in species-specific patterns. The organs on the body and tail form the posterior lateral line system (PLL), which at the end of embryogenesis, comprises five neuromasts evenly spread from head to tail along the horizontal myoseptum and two to three terminal neuromasts at the tip of the tail (Metcalfe, 1985). This pattern, which is remarkably constant among euteleosts (Pichon and Ghysen, 2004), is laid down by a primordium that migrates all the way from the ear region to the tip of the tail. The migrating primordium deposits in its wake five groups of cells, the proneuromasts, each of which will differentiate into a neuromast (Metcalfe, 1985), and fragments in two to three additional proneuromasts upon reaching the tip of the tail. The neuromasts are innervated by sensory neurons whose cell bodies are clustered in a cranial ganglion. Both the sensory neurons and the migrating primordium originate from a placode, which forms just posterior to the otic placode.

Time-lapse analysis has revealed that cells of the primordium occasionally divide (Gompel et al., 2001). It is also known that, within mature neuromasts, the hair cells are constantly renewed and that support cells divide to generate new hair cells (Balak et al., 1990; Williams and Holder, 2000). The control of proliferation within neuromasts may involve a feedback mechanism whereby the death of hair cells triggers new rounds of mitoses, as suggested by the observation that altering the rate of hair cell death correspondingly modifies the rate of cell division in the support cells (Corwin and Warchol, 1991; Williams and Holder, 2000).

To gain a more comprehensive view of the control of cell proliferation throughout the establishment of the lateral line system, we used bromodeoxyuridine (BrdU) incorporation to examine the incidence of DNA replication at different stages of the formation of the PLL. We examine first the pattern of BrdU incorporation during primordium migration and neuromast deposition, and second we address the question of cell proliferation in the PLL progenitor cells.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Cell Proliferation Within the Migrating Primordium

To visualize cell proliferation, we exposed the embryos to BrdU for 1 hr and assessed the incorporation of BrdU by immunolabeling. Labeling is mostly observed in the central nervous system, the developing fin, the eye, the peridermal cells, and the migrating primordium (Fig. 1A). Although labeled peridermal cells may overlie the primordium, they are easily distinguished from primordium cells based both on their more superficial location and on the shape and size of their nucleus, which is larger and more irregular than the nucleus of primordium cells (Fig. 1B, arrowheads).

thumbnail image

Figure 1. Bromodeoxyuridine (BrdU) incorporation during primordium migration. A: Overall view of a 36 hours after fertilization (haf) embryo. prim, primordium. B: The migrating primordium of the embryo shown in A. C: A migrating primordium showing a reduced number of DNA replicating cells in its trailing region relative to the average distribution. In all panels, the arrow indicates the direction of migration and anterior is to the left.

Download figure to PowerPoint

Because the size of the primordium varies between 92 and 124 cells in our sample, the number of cells that incorporate BrdU is expressed as a percentage of the total number of primordium cells as determined by counting the cells under Nomarski optics (Fig. 2). This proportion varies between 2.8 and 25%, with a mean of 14.5 ± 4.5% (N = 73). The distribution is Gaussian, suggesting that the sample is homogeneous.

thumbnail image

Figure 2. Distribution of 73 primordia as a function of the percentage of cells that have incorporated bromodeoxyuridine during a 1-hr pulse.

Download figure to PowerPoint

We estimated the time course of the proliferation cycle by fixing the embryos 2, 4, and 6 hr after the end of the BrdU pulse. Any replicating cell that has undergone mitosis will give rise to a pair of labeled cells. Very few doublets are observed after 4 hr, but the majority of labeled cells are in pairs after 6 hr, suggesting that cells undergo mitosis approximately 5–6 hr after replicating their DNA.

Spatial Heterogeneity in the Distribution of Replicating Cells

Many primordia seem to exhibit a heterogeneous pattern of BrdU labeling, with the trailing edge comprising fewer labeled cells than the bulk of the primordium (Fig. 1C). To quantify this effect, we examined separately the leading two thirds and the trailing one third of the cells in our sample of 73 primordia. The proportion of labeled cells turned out to be 18.1 ± 5.6% in the leading (caudal) region but only 7.2 ± 6.2% in the trailing (rostral) region. This difference is statistically significant (P < 0.0001, Welch's corrected t-test).

The trailing edge of the primordium comprises the cells that will be deposited next. We wondered whether the incidence of DNA replication within this region varies according to the time left before the trailing cells will be deposited. This variation can be roughly evaluated knowing the distance between the primordium and the last neuromast. On average, a primordium deposits a proneuromast every five somites; therefore, primordia that are four somites away from the last neuromast are due to deposit a new proneuromast soon, whereas primordia that are closer to the last neuromast will not deposit for a while. The results are presented in Figure 3. The rate of replication reaches a minimum in the trailing edge of primordia that are about to deposit a new neuromast (less than a fifth of the rate of replication in the leading two thirds, P < 0.001), whereas it is not significantly different in primordia that are halfway between two depositions (two somites away from last neuromast, Fig. 3). The replication does not vary significantly within the leading region at any stages. We conclude that there is a significant depression of replication in cells that are about to be deposited.

thumbnail image

Figure 3. Proportion of cells that have incorporated bromodeoxyuridine in the trailing third (lighter bars) and in the leading two thirds (darker bars) of primordia that are, respectively, one, two, three, and four somites away from the last-deposited neuromast; the numbers of embryos in each group were, respectively, 22, 19, 13, and 19. Trailing cells are much less likely to replicate when they are about to be deposited (rightmost bars).

Download figure to PowerPoint

Patterned Cell Proliferation

The examination of BrdU incorporation suggests that the pattern of labeled cells in the leading region of the primordia is not random but forms small garlands of labeled cells surrounding clusters of two to four unlabeled cells (Fig. 4A). To materialize this impression, we relied on an image analysis procedure that skeletizes the labeling to reveal the basic spatial arrangement of the labeled cells (Fig. 4B). Because the labeled nuclei are not all in one focal plane, images taken at different focal planes were projected on a single plane before skeletization. The results reveal that, in 38 of 73 primordia, a circular organization of replicating cells can indeed be readily observed (e.g., Fig. 4B). In 27 cases, one or two circles are present, and in another 9 cases, three circles can be detected.

thumbnail image

Figure 4. Spatial patterning of proliferation. A: A representative primordium that is four somites away from the last-deposited neuromast. B: The result of the skeletization procedure. Note that the procedure does not take into account the outline of the primordium and considers peridermal cells that abut the primordium (arrows) as linked. This strategy is because we introduced as few constraints as possible in order not to bias the analysis. The pattern defines two full annuli and an incomplete one. Note also the complete depression of replication in the trailing cells, which are about to be deposited (the narrowing marked by the arrows in A indicates that the slowing down of the leftmost cells has already begun). C,D: Two examples of primordia doubly labeled for bromodeoxyuridine incorporation (brown) and for zath1 expression (blue). In both cases, the zath1-expressing cells are surrounded by replicating cells but are themselves nonreplicating.

Download figure to PowerPoint

The annular organization of replicating cells is strongly reminiscent of the patterns of proneural and neurogenic gene expression reported by Itoh and Chitnis (2001). The proneural gene zath1 and the neurogenic gene Delta are both expressed in two to three clusters of cells within the primordium, whereas the receptor of Delta, Notch, is expressed in a complementary pattern (Itoh and Chitnis, 2001). It seemed possible, therefore, that the mitotically inactive cells that form the core of each annulus correspond to the zath1-expressing cells that will eventually become hair cells. We tested this possibility by performing double labeling for BrdU incorporation and for zath1 expression. The results (Fig. 4C,D) show that zath1 expression is restricted to mitotically inactive cells and that it is often surrounded by a garland of replicating cells. In a few cases (10 of 240 zath1-expressing cells examined), an overlying peridermal cell was heavily labeled with BrdU, making it impossible to ascertain whether the zath1-expressing cell was also labeled.

Cell Proliferation After Deposition

We examined the incidence of DNA replication after deposition. Nearly half of the newly deposited proneuromasts (one to three somites from the primordium) showed no BrdU incorporation at all (33 of 70), whereas in proneuromasts that had been deposited for approximately 4–5 hr, as judged from their distance of seven to nine somites to the primordium, three fourths (22 of 30) showed labeling. In most cases (17 of 22), this labeling was confined to peripheral cells. We conclude that the proneuromasts undergo a phase of mitotic quiescence after their deposition and that proliferation resumes in the peripheral cells at approximately the time the neuromasts differentiate.

We also examined the pattern of proliferation in 29 differentiated neuromasts in 48 hours after fertilization (haf) embryos. We observed an average of 2.7 ± 1.9 labeled cells per neuromast. The average number of cells is approximately 25–30 in a young neuromast, but we chose not to express our results as a percentage, because it is more difficult to achieve an accurate count of unlabeled cells in neuromasts than in the primordium.

At this stage, hair cells are easily distinguished from support cells based on their refringence and shape. A total of 97% of the dividing cells are clearly support cells, and 77% of these are in direct contact with the central hair cells.

Mitotic Behavior of Lateral Line Progenitor Cells

The PLL placode can first be detected around 18–20haf (Kimmel et al., 1995), shortly before the primordium begins to migrate (20haf). To gain some insight in the proliferation events that precede the formation of the lateral line placode, we exposed developing embryos to 2-hr-long pulses of BrdU starting at 6, 8, 10, 12, 14, 16, and 18haf, and fixed all the embryos at 24haf when the primordium was well on its way and easily visualized. A summary of the results is presented in Table 1. It appears that the proportion of replicating cells decreases from >98% to 10% between 8 and 10 and 14 and 16haf, suggesting a transition from generalized proliferation to generalized quiescence. Proliferation then resumes until 22haf, the time when the primordium begins its migration.

Table 1. Cell Proliferation Prior to Primordium Formationa
 Age (haf)
6–88–1010–1212–1414–1616–1818–2020–22
  • a

    Embryos at various ages from 6 to 20 haf were exposed to BrdU during two hours, and the presence of BrdU-labeled nuclei was assessed at 24haf, when the migrating primordium is easily delineated with Nomarski optics. The next two lines indicate, respectively, the proportion of labeled cells in the migrating primordium and the size of the clusters of labeled cells or their shape (ring, complete or partial ring of labeled cells). nd, not determined; BrdU, bromodeoxyvridine; haf, hours after fertilization.

% of labeled cells>9895591510nd5036
Clusters  4–5 cells2–4 cells1–2 cellsndRingRing

The interpretation of the data is not straightforward, however, because the presence of BrdU in a cell at 24haf does not necessarily mean that replication took place within the pulse interval (because BrdU takes a while to be washed out after the pulse), nor does the absence of labeling suggest a lack of replication (because a high level of proliferation may dilute out the BrdU that was effectively incorporated during the pulse). Thus, we refined the interpretation of the data by looking at changes in the pattern of proliferation between consecutive time intervals.

When the exposure to BrdU is given during the 6- to 8haf interval, virtually all cells of the primordium are labeled to some extent (≥98%, Fig. 5A). After an 8- to 10haf exposure, a majority of primordia (9 of 14) are labeled as after the 6- to 8haf exposure, but the remaining primordia comprise 5–10% unlabeled cells. This tendency increases between 10 and 12haf, where the proportion of unlabeled cells varies between 25 and 40% (2 of 15 primordia), 40 and 60% (5 of 15), and 60 and 85% (8 of 15). In the latter group, the labeled cells are mostly arranged in small clusters of up to four to five cells (Fig. 5B). After the 12- to 14haf exposure, we observe that less than 15% of the cells are labeled and are arranged either by four or by two (Fig. 5C, arrows), indicating that the first round of mitosis has already occurred in two-thirds of the dividing cells. After the 14- to 16haf pulse, we observe that 10% of the cells are labeled and those are either single or arranged in pairs, suggesting that the second round of mitosis occurs during this interval. Altogether, the data indicate that, although some cells of the primordium enter a phase of quiescence before 8haf, the majority of primordium progenitor cells become quiescent during the 10- to 12haf interval, and that those cells that still divide will undergo two rounds of mitosis. We have no data for the 16- to 18haf pulse. After an 18- to 20haf exposure, we observe that the proportion of labeled cells has increased to 40–60%, and the labeled cells tend to be organized in circles. Finally, a 20- to 22haf pulse results in labeling of 28–45% of the cells (32.5% average; N = 4; Fig. 5D). The larger proportion of labeled cells in the 18- to 20haf sample relative to the 20- to 22haf sample is presumably due to the fact that the cells that replicate their DNA during the 18- to 20haf interval are likely to have divided at the time the embryo was fixed (24haf approximately), whereas cells labeled during the 20- to 22haf interval are less likely to have divided at 24haf.

thumbnail image

Figure 5. Proliferation in lateral line precursor cells at various stages of development. A: The cells whose progeny will eventually form the primordium are actively proliferating between 6 and 8 hours after fertilization (haf). B: During the 10- to 12haf interval, the vast majority of precursor cells have become quiescent; the few cells that are still dividing will produce small clusters of four progeny on average. C: Between 12 and 14haf, the few cells that remain dividing will only produce pairs (arrows). D: Between 18 and 20haf, many cells have resumed replication.

Download figure to PowerPoint

The data allow us to distinguish three phases in the proliferation behavior of lateral line precursors. The first phase lasts until 6–8 hr and consists of generalized proliferation. From 8 to 12haf, more and more precursor cells enter a phase of quiescence that ends shortly before the emergence of the primordium. This quiescence appears specific for the lateral line precursor cells, as myoblast nuclei, for example, are readily labeled during this period. A small proportion of precursor cells keep on dividing, however, to produce small clones of approximately four cells on average. Finally, a new period of proliferation begins shortly before the emergence of the placode.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

The development of the posterior lateral line system involves cell proliferation, migration, and differentiation within a well-defined, easily visualized population of cells. It may be a good place, therefore, to look for interactions between these processes. Our results confirm that cell divisions take place within the migrating primordium (Gompel et al., 2001). We show here that this proliferation is highly patterned both in time and in space. We will discuss successively two aspects of the pattern: first, the annular organization of proliferating cells, and, second, the repression of proliferation in cells that are about to be deposited.

We observed that the arrangement of replicating cells is not random. Rather, they tend to form annuli around a core of a few quiescent cells. The presence of such annuli can be evidenced by a simple skeletization procedure, as shown Figure 4B. Annuli of cell proliferation could be fortuitous, as even with a random proliferation pattern, rings of adjacent cells would be likely to occur with some frequency by chance. On the other hand, the presence of a core of nonreplicating cells at the center of each annulus could be correlated to the localized expression of the proneural gene zath1. We assessed whether DNA replication and zath1 expression are inversely correlated by combining BrdU incorporation and in situ hybridization. The results show that zath1 expression is indeed restricted to nonreplicating cells.

The expression of zath1 and subsequent determination of hair cells within the central, mitotically quiescent regions of each preproneuromast presents a striking similarity with the determination of sensory organs in Drosophila, which also takes place within small clusters of mitotically quiescent cells (Usui and Kimura, 1992; Nègre et al., 2003). This similarity might reflect some shared mechanism or even shared ancestry between the two types of mechanosensory organs, as already proposed by others (Lewis, 1991; Fritzsch et al., 2000).

The depression of proliferation that is observed at the trailing edge of the primordium indicates that cells preparing for deposition enter a phase of quiescence. This phase will last as long as the proneuromast takes to differentiate. It may be that proliferation and differentiation are not compatible and that quiescence is required during the differentiative phase. Alternatively, it may be that the proliferation program is not the same in primordium and in neuromasts and that a period of quiescence is required to shift from one phase to the next. These two explanations are obviously not mutually exclusive.

When the phase of quiescence of the proneuromasts is over, proliferation resumes in the support cells but not in the hair cells. We observed that BrdU incorporation occurs mostly in support cells that are adjacent to the hair cells, consistent with the idea that the mitosis of support cells may somehow be triggered by the death of hair cells, as suggested by Williams and Holder (2000). On the other hand, these authors observed that DNA replication mostly takes place within the mantle cells that are at the periphery of the neuromasts, contrary to our observations. It may be that this difference reflects a further specialization of the neuromasts, whereby proliferation would become restricted to the mantle cells at some stage between 48haf (our observations) and 10 days (Williams and Holder, 2000).

If we now turn to the origins of the system, it appears that the progenitor cells that will form the lateral line placode enter a phase of mitotic quiescence between 10 and 12haf, and resume proliferation between 16 and 18haf. The exception to this generalized quiescence is a small population of cells that will undergo two consecutive mitoses between 10 and 16haf, and generate four daughter cells. We do not know whether this subpopulation differs in any other way from the bulk of the presumptive placodal cells.

Unfortunately, the time when the lateral line system is set aside from other tissues is not known. Fate map studies indicate that the cells that will generate the different placodes are already roughly aligned at 6haf (Kozlowski et al., 1997), but single cell lineage analysis turns out to be very difficult at later gastrula stages (unpublished experiments). Several genes are expressed in a horseshoe-shaped domain that corresponds to placodal territory at the end of gastrulation (10haf); some of them will form distinct patches of expression that correspond to lateral line placodes at 17haf (Akimenko et al., 1994; Sahly et al., 1999; Kobayashi et al., 2000). Thus, it seems likely that the determination of lateral line precursor cells takes place between 10 and 17haf and that the period of quiescence that we observed corresponds to this event. Interestingly, the progenitors for the lateral line sensory neurons are also proliferating at 8haf but have become postmitotic at 13haf (Metcalfe, 1983). We observed that exposure to BrdU between 8 and 10haf results in the labeling of most neurons, whereas exposure during the 10- to 12haf period leads to the labeling of only a few neurons (unpublished results), suggesting that the onset of quiescence may coincide for the neurons and for the other placodal derivatives.

The phase of quiescence is followed by a new phase of replication that begins around 18haf and shows an annular organization, perhaps related to the expression of zath1. The annular phase presumably begins in the placode, because this pattern is first observed after an 18- to 20haf pulse, shortly before the primordium can be seen. The larger proportion of cells labeled during the 18- to 20haf and the 20- to 22haf exposures (50 and 32.5%, respectively), relative to the proportion documented in the migrating primordium (15%) is due to two factors: first, the cells that replicate during the 18- to 20haf period (and to a much lesser extent during the 20- to 22haf period) are likely to have already divided by the time the embryos are fixed (24haf); second, the exposure to BrdU lasted for 2 hr in the 18- to 20haf and 20- to 22haf samples, although it lasted only 1 hr when monitoring the replication in the migrating primordium.

In summary, then, we can distinguish three phases of proliferation, each of which has its own properties. The first phase corresponds to a generalized proliferation that occurs during the late gastrula stage. The second phase is the annular proliferation that is observed in the placode and migrating primordium. The third phase corresponds to the proliferation of the support cells that generates new hair cells all along larval life. The three phases are separated by two phases of quiescence: one that accompanies the determination of the posterior lateral line placode, and one that accompanies proneuromast deposition and differentiation.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Eggs

Zygotes were collected soon after the onset of the light cycle and transferred to Petri dishes in 1 mM NaCl in tank water. Methylene blue was routinely added as antiseptic (0.5 mg/L final).

BrdU Incorporation

For the analysis of proliferation in the migrating primordium, 20- to 40haf embryos were manually dechorionated and transferred to a solution of BrdU (Sigma) 10 mM, 15% dimethyl sulfoxide in tank water. They were left at 4°C for 30 min, then at 28°C for 60 min, and immediately fixed in a solution of 4% paraformaldehyde in 0.1% Tween20 in phosphate buffered saline (TwPBS). They were left in the fixative either overnight at 4°C or for 2 hr at room temperature, rinsed three times 5 min in 100% methanol, and kept at −20°C.

For the analysis of progenitor cells, embryos were dechorionated in a Petri dish containing a layer of 2% agar and screened 6 hr after the eggs were laid to make sure they had reached the 6haf (shield) developmental stage. They were then screened again at the onset of the exposure time, just before being transferred into the BrdU solution for 2 hr at 28°C, and then transferred to tank water and screened again to make sure their development had proceeded normally during the exposure. They were rinsed three times 5 min with tank water and left at 28°C until the primordium had reached somite 3–6 (24haf approximately).

Immunolabeling

Immunolabeling was performed as described in the zebrafish book, http://zfin.org/zf_info/zfbook/cont.html. The primary antibody (mouse anti-BrdU G3G4, Hybridoma Bank) was diluted 1/100 and the peroxidase-coupled secondary antibody (Jackson Laboratories) was diluted 1/200. The labeled embryos were washed three times 5 min in TwPBS, transferred for 1 hr in glycerol/TwPBS (1/1) then 1 hr in glycerol 100%, and mounted in glycerol for microscopic examination.

In Situ Hybridization and Immunolabeling

The 25- to 35haf embryos were manually dechorionated, exposed to BrdU and fixed as described above, and stored at −20°C. They were then processed for in situ hybridization as described by Hauptman and Gerster (1994). They were finally immunolabeled with anti-BrdUI antibody as described above.

Imaging

Photomicrographs were taken on a Zeiss Axioplan microscope equipped with a ×40 objective and fitted with a Canon Powershot G5 digital camera. They were subsequently handled and assembled with Adobe Photoshop 7.0 for Macintosh.

Image Analysis

The cells of the primordia form a monolayer but may be located at slightly different depth levels so that some of them are out of focus in a single photomicrograph. We, therefore, took a z-series of digital images and used maximal-density projection to form a single image where all BrdU-labeled nuclei are clearly visible. This image was thresholded to get black labeled nuclei on a white background. The resulting picture was eroded to remove small specks and then dilated until nuclei of neighboring cells became confluent. Skeletonization was then applied to the dilated picture, resulting in the appearance of circles whenever the labeled cells were organized in annuli. All procedures were performed with the public-domain software Image-J (NIH, Bethesda, MD).

Statistics

The data were treated with the Graphpad Instat 3 software. The distributions of our data were not significantly different from normality. We, therefore, used parametric tests: Welch's corrected t-test for comparison between two groups, and one-way analysis of variance, followed by Tukey's post hoc test for comparison between more than two groups. Differences were considered statistically significant for P values < 0.05.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES