Investigating Lateral Line Canal Morphogenesis as a Bone Remodeling Process in Lake Malawi Cichlids

The lateral line (LL) system is a key non-visual sensory modality founded in all fishes and larval and adult aquatic amphibians. In bony fishes, the LL system consists of series of superficial neuromasts (on the skin surface) and canal neuromasts (housed in bony canals). Differences in LL canal phenotype have consequences for LL sensory capabilities, and, as such, could contribute to speciation. The LL system develops with a process of neuromast patterning followed by the genesis of canals from dermal bone around a subset of neuromasts. This study used histochemical methods to describe canal morphogenesis by assessing osteoblast (OB, bone building cells) and osteoclast (OC, bone resorbing cells) activity through ontogeny. OB and OC activity was studied along the length and around the circumference of the mandibular LL canal in two species of Lake Malawi cichlids. Aulonocara stuartgranti (widened canals) and Tramitichromis sp. (narrow canals) are known to develop their different canal phenotypes as a result of dissociated heterochrony. In both species, OB activity is concentrated in the canal roof and floor (behind the neuromast), while OC activity is focused primarily in the canal walls. Further, both species demonstrate one pulse of bone deposition early in ontogeny and another pulse of bone resorption later in ontogeny, but in Aulonocara not Tramitichromis, the pulse of bone resorption is accompanied by high levels of OB activity. Thus, the second pulse of bone cell activity in Aulonocara is defined by bone resorption and bone deposition: active bone remodeling by the simultaneous action of OBs and OCs. These results corroborate the observation of dissociated heterochrony (Bird and Webb, 2014) at the cell level, suggesting that the nature of neuromast-centered canal morphogenesis may differ depending on LL canal phenotypes.


INTRODUCTION
The mechanosensory lateral line system senses water flow, enabling fishes to respond to hydrodynamic cues involved in prey detection, predator avoidance, communication, and navigation (Dijkgraaf 1963;Janssen 1997;Webb et al. 2008;Schwalbe et al. 2012;Yanase et al. 2014;Schwalbe and Webb 2015). Five cranial canal morphologies are found among bony fishes: narrow-simple (referred to as "narrow" from here on), narrow-branched, narrow with widened tubules, widened, and reduced (reviewed in Webb 2014b; Figure 2). The functional consequences of some aspects of this morphological variation have been investigated (Denton and Gray 1988;Janssen 1997;Van Snick Gray and Stauffer 2004;Schwalbe and Webb 2013;Webb 2014a, b;Schwalbe and Webb 2015;Klein and Bleckmann 2015;Schwalbe et al. 2016). For example, widened canals are more sensitive to low frequencies (<60Hz) than narrow canals, and this increased sensitivity plays a role in prey detection (in the water column and in the benthos). However, the increased resonance in the widened canals results in a slower reaction time than in narrow canals (reviewed in Coombs et al. 1992).
Thus, differences in canal morphology confer functional differences in the lateral line system and are therefore ecologically and evolutionarily relevant.
The diversity of fishes in the Family Cichlidae presents a unique opportunity to compare different lateral line canal phenotypes in closely related taxa. The rapid evolutionary diversification of cichlids is a well-known example of explosive speciation; over 1000 species of cichlids have arisen during the last million years in Lake Malawi and Lake Victoria together (Allender et al. 2003). Many of these species occupy partially overlapping physical ranges and ecological niches, but they still sort into discrete species in nature, despite their ability to produce viable hybrid offspring in lab settings (Allender et al. 2003;Seehausen et al. 2008). Attempts to understand the biological mechanisms underlying cichlid adaptive speciation have been a major focus of evolutionary biologists and geneticists (e.g., Allender et al. 2003;Kocher 2004;Van Snick Gray and Stauffer 2004;Albertson et al. 2005;Seehausen et al. 2008;Cooper et al. 2011;Parsons et al. 2012;Powder et al. 2015). Variation in sensory ability among species (e.g. vision) has been shown to underlie differences in behavior, which provides a basis for speciation along environmental gradients (Allender et al. 2003;Schleuter and Eckmann 2006;Seehausen et al. 2008). In Lake Malawi, one such environmental gradient is the decrease in light intensity with depth. Deeper-living cichlid species need to rely more on nonvisual sensory modalities such as the mechanosensory lateral line system to find prey in deeper, darker waters. Related shallow-water cichlids live in a brighter environment, so the selective pressures associated with light limitation is relaxed, leaving the shallow-water species free to rely less on their lateral line and more on visual prey detection.
Taking this example, phenotypic differences in lateral line canal morphology with functional consequences for sensory capability could have contributed to the rapid adaptive speciation of this group. However, any consideration of that possibility would need to be grounded in a robust understanding of the link between lateral line structure and function, particularly the differences between canal phenotypes.
The two Lake Malawi cichlids Aulonocara stuartgranti and Tramitichromis sp.
(referred to by genus name throughout) are an ideal pair of species that can be used to investigate the developmental and functional differences between lateral line phenotypes (e.g. Schwalbe et al., 2012;Schwalbe and Webb 2013;Bird and Webb, 2014;Schwalbe and Webb 2015;Schwalbe, et al. 2016;Becker et al. 2016). Both species locate prey in sandy-bottomed environments, but Aulonocara has widened canals and members of the genus live at a greater range of depths (5-120 m), while Tramitichromis has narrow canals and tends to live in relatively shallow waters (>15 m; Fryer and Iles 1972;Konings 1990Konings , 2007; Figure 3). Recent studies have demonstrated that while both Aulonocara and Tramitichromis employ their lateral line systems in combination with vision to detect prey in light environments, only Aulonocara uses its lateral line to detect prey and feed in dark environments below a critical light level (Schwalbe et al. 2012;Schwalbe and Webb 2013;Schwalbe and Webb 2015;Schwalbe et al. 2016). Further, the generation of interspecific differences in canal phenotype are a result of dissociated heterochrony during canal morphogenesis .

Lateral Line Development
In bony fishes, lateral line development occurs in three phases, beginning in the embryo and continuing until long after metamorphosis from the larval stage to the juvenile stage. In the embryo, neuromast primordia migrate (on the trunk) or elongate (on the head) from cranial ectodermal lateral line placodes, establishing the distribution of neuromasts (Northcutt andGans 1983, reviewed in Webb 2014b). The neuromasts then grow and mature, and some superficial neuromasts start to proliferate. Finally, a subset of neuromasts is enclosed in bony canals through the process of neuromast-centered canal morphogenesis (reviewed in Webb 2014b).
Each neuromast is composed of a central group of hair cells surrounded by nonsensory support and mantle cells (Figure 4a). A ciliary bundle extends from the apical surface of each hair cell, consisting of one long kinocilium adjacent to several, graded stereocilia. The location of the kinocilium in relation to the stereocilia defines the polarity, or axis of best physiological sensitivity, of each hair cell, and the presence of hair cells with like and opposing polarities (180° to each other) defines the axis of best physiological sensitivity of a neuromast (discussed in Webb 2014b). In juvenile and adult fishes, the neuromast surface occupied by hair cells is restricted to a small round or oval area known as the sensory strip (Coombs et al. 1988; Figure 4b). Basally, the hair cells are innervated by sensory neurons, and apically, the ciliary bundles project into a protective gelatinous cupula, which is thought to be secreted by the support or mantle cells (Münz 1979;Blaxter 1987;Webb 2014b Chitnis et al. 2011;Cruz et al. 2015).
The posterior (trunk) lateral line of zebrafish (Danio rerio), in particular, has been the focus of many studies aiming to tease apart the genetic control of lateral line patterning and hair cell regeneration (Raible and Kruse 2000;Gompel et al. 2001;Sapède et al. 2002;López-Schier et al. 2004;Ma and Raible 2009;Romero-Carvajal et al. 2015).

Canal Development
The pattern and timing of canal morphogenesis have been described in a number of teleost taxa (reviewed in Webb 2014b). Canal development occurs in four stages, a model that has been used in several studies (Tarby and Webb 2003;Webb and Shirey 2003;Moore 2008;Bird and Webb 2014; Carter 2014; Figure 5). It has been noted by several authors that canals start to form in the vicinity of neuromasts, a phenomenon termed "neuromast-centered canal morphogenesis" (reviewed in Webb 2014b). Canal morphogenesis begins with a presumptive canal neuromast sitting on the skin surface (Stage I). The neuromast appears to sink into a depression (Stage IIa) and canal bone starts to ossify within the dermis, forming ridges, which grow upward to form canal walls around the neuromast (Stage IIb). Soft tissue (epidermis, dermis) fuses over the neuromast (Stage III). The bony walls continue to grow within the dermis and over the neuromast, finally fusing to form the ossified roof of the canal segment (Stage IV). From above, the canal walls appear to form "scallops" of soft tissue, with bone within (described by Webb 2014a, 2014b). In histological cross sections, two growing arms of soft tissue containing bone reach around the canal circumference as growing bony fronts, which then fuse over the neuromast (Figures 5, 6). As individual canal segments form, the walls of adjacent canal segments fuse, leaving a pore in the canal roof between them. This process forms a continuous canal in which each neuromast sits in an ossified canal segment, between which there are pores that connect the canal with the outside environment. As the fish grows, canal diameter increases and the canals remain integrated within the dermal bones , but this process has not been investigated at the cell level.

Dermal Bone Development and Remodeling
All of the bones associated with the cranial lateral line canals in teleosts are dermal bones, meaning they are formed directly within the dermis without a cartilage model (Cubbage and Mabee 1996;Weigele and Franz-Odendaal 2016). In addition (with a few notable exceptions) all "higher" teleosts, including cichlids, have acellular bone (Huysseune 2000). Thus, the bones that house the lateral line canals in cichlid fishes are dermal, acellular bones (Huysseune 2000). Acellular dermal bone has several characteristic features including mode of ossification, absence of osteocytes embedded in the bone matrix, and the manner in which the bone is resorbed or remodeled. The osteogenesis of endochondral bone involves a preliminary cartilage template composed of chondrocytes, which is secreted by mesenchymal cells. In contrast, dermal bones do not form from a cartilaginous template, but form directly within the dermis. Secondly, in cellular endochondral bone, osteoblasts (bone building cells) begin mineralizing the template from the inside and become trapped in the bone matrix (Weigele and Franz-Odendaal 2016). Once they are surrounded by bone matrix, the osteoblasts are termed osteocytes, and their presence in the bone identifies the bone as cellular (Moss 1960;Weiss and Watabe 1979;Weigele and Franz-Odendaal 2016). In acellular bones, including acellular dermal bones, the osteoblasts form mineralized bone in a polarized manner, retreating from the ossification front so that they do not become entrapped in bone matrix (Ekanayake and Hall 1988).
A third feature of acellular dermal bone is the manner in which it is usually resorbed or remodeled. There are two types of osteoclasts (bone resorbing cells)mononucleated osteoclasts, which are associated with shallow sites of resorption (without obvious lacunae), and multinucleated osteoclasts, which are associated with deep lacunar resorption sites (Witten 1997;Witten and Huysseune 2009). Unlike resorption of cellular bone (e.g., in mammals), which is characterized by multinucleated osteoclasts and deep lacunae, resorption of dermal bone in teleost acellular bone is likely to be the site of shallow bone resorption carried out by mononucleated osteoclasts (Witten 1997).
However, in times of intense bone remodeling, these mononucleated osteoclasts may group together to resemble large multinucleated osteoclasts. Under other conditions, however, they may appear flat, similar to the bone-lining cells in mammals (Witten 1997).
In prior investigations, the presence of any type of osteoclast has been considered to be a hallmark of such remodeling in acellular bone (Witten and Villwock 1997;Nemoto et al. 2007). Osteoclast activity has been found primarily in bony elements undergoing growth, such as those in the craniofacial skeleton and vertebral column (Witten et al. 2001). The growth of the dentary bone (the dermal bone which houses the mandibular lateral line canal, and which composes the majority of the lower jaw of fishes) was examined for bone resorption and remodeling by Witten, Hansen and Hall (2001). It was found to be the site of resorption by both mono-and multinucleated osteoclasts, with an ontogenetic progression from mononucleated osteoclasts in 20-day old zebrafish to multinucleated osteoclasts in 40-day old fish (Witten et al. 2001). The mandibular lateral line canal, though not expressly assayed for bone remodeling in that study, was noted as an area for further study and probable bone remodeling.
The processes of acellular dermal bone ossification and remodeling (particularly in growing skeletal elements), the ontogenetic increases in lateral line canal diameter, and the observations of osteoclasts in the vicinity of lateral line canals suggest that the cranial lateral line canals are active sites of bone remodeling. The presence of osteoclasts was noted in cranial lateral line canals in a cichlid fish (Witten 1997), and their activity has been suggested by an observed increase in canal diameter that began even before the completion of canal morphogenesis (Tarby and Webb, 2003;Bird and Webb, 2014). The  (Wada, et al. 2014). However, no study has yet investigated the activity of osteoblasts in the bone remodeling process in teleost fishes despite their role as "builders" of the canals.
Thus, it is hypothesized that the combination of the activity of osteoblasts and of osteoclasts is responsible for the initial morphogenesis and subsequent growth (increase in canal diameter) of the lateral line canals, and further, that differences in the dynamics of these two cell types during larval and post-larval (juvenile) development can explain variation in adult canal phenotype among species. Also, based on the concept of neuromast-centered canal morphogenesis, it is predicted that osteoblast and osteoclast activity will differ along the canal length in relation to the location of neuromasts. In addition, it has been established that heterochronic shifts in the process of canal enclosure explain the morphological differences between narrow and widened canals . A delay in the commencement of canal morphogenesis coupled with an acceleration in the increases in canal diameter and neuromast size resulted in the development of a widened rather than a narrow canal phenotype in cichlids . However, the dynamics of osteoblast and osteoclast activity responsible for the construction of divergent adult canal phenotypes (e.g., narrow vs. widened) has not been examined.
Thus, the goal of this work was to determine: (1) how the activity of osteoblasts and osteoclasts could account for the initial enclosure of the mandibular lateral line canals (including exploration of neuromast-centered canal morphogenesis) and the increase in mandibular canal diameter with fish growth, and (2) how the activity of osteoblasts and osteoclasts contributes to divergent adult lateral line canal phenotypes (widened, narrow) in two species of cichlids: Aulonocara stuartgranti and Tramitichromis sp. A comparison of spatial patterns of bone deposition and resorption during canal morphogenesis in species with different adult canal morphologies provides a key link between activity at the cell-level, adult phenotype, and sensory function.
Understanding the dynamics of canal formation and growth will lay the foundation for future studies of the genetic mechanisms underlying canal morphogenesis. More generally, the analysis of osteoblast and osteoclast activity and distribution during development will contribute to our understanding of the contribution of acellular bone skeletogenesis and its contribution to post-embryonic craniofacial development in fishes.

Study Animals
Breeding groups of the maternal mouth-brooding cichlids, Aulonocara

Fluorescent Vital Staining for Bone and Neuromasts
An ontogenetic series of each study species (Aulonocara stuartgranti and Tramitichromis sp.) was vitally stained and analyzed to visualize canal ossification and canal neuromasts in order to identify the size classes of larvae and juveniles that would subsequently be used for histochemical assays of osteoblast and osteoclast activity.
Live fish were immersed in 0.0024% 4-di-2-ASP (Sigma-Aldrich D0815) for 5 minutes in the dark, then moved to 0.05% calcein (Sigma-Aldrich C0875) for 5 minutes in the dark to simultaneously visualize neuromasts and newly calcified bone, respectively. Fish were rinsed in tank water, and anaesthetized in 0.02% MS-222 in tank water (based on Fujimura and Okada, 2008)

Analysis of Osteoblast and Osteoclast Activity
Osteoblast and osteoclast activity was analyzed in the mandibular (MD) canal on the right side of the head only. The MD canal runs rostro-caudally, thereby allowing for accurate measurements of canal and neuromast dimensions in serial transverse sections (as in prior studies: Bird and Webb 2014;Webb et al. 2014;Becker et al. 2016).
The MD canal in both Aulonocara and Tramitichromis contains five neuromasts (MD1-5). MD1 was excluded because it is the last to enclose and because in very young larvae it can be difficult to distinguish from the nearby superficial neuromasts, which are similar in size. MD5 is located in the anguloarticular bone whereas MD1-4 are in the dentary bone, so MD5 was eliminated from the analysis to avoid any confounding factors due to its association with a different bone. Thus, the portion of the canal in the dentary bone containing neuromasts MD2, MD3, and MD4 was used for analysis. Osteoblast and osteoclast activity was analyzed at the location of each neuromast as well as in the regions between neuromasts (e.g., "mid 2-3", "mid 3-4"). In addition, a more detailed assessment of osteoblast and osteoclast dynamics in association with a single neuromast (MD3) was carried out in an ontogenetic series of each of the two study species.
An respectively. As such, the area of interest for this study is defined by five Regions along the MD canal -the MD2 Region, the mid 2-3 Region, the MD3 Region, the mid 3-4 Region, and the MD4 Region (see Figure 7). The MD2, MD3, and MD4 Regions are referred to collectively as the "neuromast-associated" Regions, and the mid 2-3 and mid 3-4 Regions as the "inter-neuromast" Regions.
Further, histological sections corresponding to each of the 29 positions along the length of the canal were used for the analysis of the spatial distribution of osteoblast and osteoclast activity around the canal circumference. To accomplish this, the canal circumference was divided into 8 radial segments ( Figure 8) and a transparency with a radial grid divided into 8 equal segments was placed over a live image of a histological section on a computer screen with the central axis aligned with the center of the neuromast. In sections where the canal was not completely ossified, the location of the advancing bone fronts (growing canal walls) was also recorded, as was the developmental stage of the canal segment (Stage 1, 2a, 2b, 3, 4; as per Tarby and Webb, 2003). After preliminary analysis, it was determined that data from pairs of the eight segments could be combined to form four quadrants, with Quadrant 1 representing the canal floor/neuromast, Quadrants 2 and 3 representing the left and right canal walls, respectively, and Quadrant 4 representing the canal roof (Figure 8c,d).

Osteoblast and Osteoclast Assays
To visualize osteoblast and osteoclast activity during canal morphogenesis, fish were stained either for alkaline phosphatase (AP), an enzyme expressed by active osteoblasts and thereby found in areas of bone ossifiation, or for tartrate-resistant acid phosphatase (TRAP), an enzyme expressed by active osteoclasts and thus found in areas undergoing bone resorption.
In each species, three fish in each of five size classes (8, 12, 16, 20, 24 mm SL ±1 mm) were stained for AP and three fish in each size class were stained for TRAP. Fish size was used to define size classes because canal development tracks with size, rather than age (Münz 1979, Ledent 2002. Fish were anaesthetized in 0.02% MS-222, then fixed in 4% paraformaldehyde in PBS for 12 -24 hours. If the staining assay (either AP or TRAP) could not be performed immediately after fixation, samples were dehydrated in a graded ethanol series to 70% ethanol for storage until they could be stained (Edsall and Franz-Odendaal, 2011). This was done in an effort to avoid the potential negative effects of a longer fixation time on staining. Dehydrated specimens were subsequently rehydrated in a reverse graded ethanol series prior to staining. A comparison of specimens that were assayed immediately to those that were dehydrated and rehydrated before being stained revealed that this additional step had no discernable effect on staining (data not shown). Roche). BCIP binds to AP, and NBT reacts with the BCIP forming a dark blue insoluble azo dye (Boenisch 2001). Negative controls were done in which DMF (without BCIP) was added in place of the BCIP solution. To determine the ideal staining duration, a doseresponse experiment was conducted. Given that AP could stain just the outermost tissue layers or may stain them more strongly than more interior tissues, a 16 mm SL Tramitichromis specimen was bisected prior to staining and sectioned, which confirmed that the stain sufficiently penetrated the tisuse. Whole stained specimens were imaged and then worked up into 80% glycerol in water for storage prior to histological preparation for detailed analysis in transverse sections. subbed slides (10% albumin in 0.9% NaCl). Slides were counterstained with 4% methyl green in distilled water for 10 minutes, then rapidly dehydrated in absolute ethanol (modified from Presnell and Schreibman 1997, N. Bird, pers. comm.), and coverslipped with Entellan (Electron Microscopy Sciences).

Quantification and Visualization of Osteoblast and Osteoclast Activity Data
Alkaline phosphatase (AP) staining for osteoblast activity appeared as a diffuse smear of color ranging from light lavender to dark indigo, which was interpreted as corresponding to the intensity of osteoblast activity. In contrast, tartrate-resistant acid phosphatase (TRAP) staining for osteoclasts usually appeared as discrete, red-stained osteoclast cells with darkly stained nuclei, and occasionally as a light wash of red localized to surface of a bone ( Figure 9). Thus, different approaches were needed to quantify AP and TRAP staining. For AP staining, a scoring system was devised where the lightest staining (low osteoblast activity) was scored as a 1 and the darkest staining (highest osteoblast activity) was scored as a 4. Although subjective, it was possible to illustrate relative levels of activity, at positions along the canal and around the circumference of the canal. For TRAP staining, a more traditional cell counting method was used since discrete osteoclast cell nuclei were darkly stained and easily identified.
Preliminary observations revealed what were apparently multinucleated TRAP-stained osteoclasts, despite the fact that acellular bone in young teleosts is usually associated with mononucleated osteoclasts (Witten 1997, Witten andHuysseune 2009). In these cases, it was unclear whether the cells were multinucleated osteoclasts or closely spaced mononuclear osteoclasts as described by Witten (1997). Since these two alternatives were not reliably distinguishable, TRAP staining intensity (osteoclast activity) was assessed using the total number of TRAP-stained nuclei.
Several data visualization tools were used to visualize AP and TRAP distribution and intensity or abundance around the circumference of the canal, at different positions along the canal length, in fishes of different size classes, in the two study species. After much trial and error, it was determined that the most effective way to represent these data would be to apply a colorimetric scale to a series of pie charts (eg. see Figure 8) where the colors correspond to the relative amount of osteoblast or osteoclast activity as shown by AP and TRAP staining, respectively. In these graphs, a light color represents a low activity level, and a darker color represents a higher activity level. Thus, the pie charts serve as diagrammatic representations of canal cross-sections. For AP data, the mean of the AP scores in each of the quadrants are represented by the AP score in each of the resulting quadrants. For TRAP data, the total number of osteoclast nuclei in each of the quadrants was recorded. The conditional formatting tool in Excel (Microsoft Office v. 15.11.2) was used to generate color gradient scales corresponding to the values of AP (the score of 0-4) and TRAP (the number of osteoclast nuclei counted), with darker colors representing higher scores.
For numerical graphs plotting osteoblast and osteoclast activity across ontogeny, the raw scores (AP) or number of osteoclast nuclei (TRAP) in all four quadrants at all 29 positions in each replicate fish were summed to yield one value that represented the total osteoblast or osteoclast activity in that individual fish. In order for osteoblast and osteoclast data to be comparable despite vastly different scales (scores vs. osteoclast nuclei), the maximum total osteoblast or osteoclast activity was identified, and the total activity values were reinterpreted as percentages of that maximum.

Canal Enclosure
The stage of canal enclosure (Tarby and Webb 2003, Webb and Shirey 2003 was also recorded for each histological section analyzed. The resulting data was used to explore patterns in the sequence and timing of canal enclosure in association with each neuromast along the length of the canal.

Statistical Analysis
Chi-squared (x 2 ) tests were computed using R (version 3.3.2) to test the hypothesis that the osteoclasts (TRAP staining) are uniformly distributed, either along the length of the canal (MD2 to MD4), or among the four quadrants around the circumference of the canal. The subjective nature of the AP data and the lack of an angular value for each TRAP nucleus relative to the axes of the data collection grid precluded other statistical analysis.

RESULTS
In both Aulonocara and Tramitichromis, vital staining with calcein and 4-di-2-ASP revealed five MD canal neuromasts within the developing MD canal (Figure 10).
Canal segments formed asynchronously in a given individual. For instance, in a 16 mm SL fish, the minimum developmental stage observed for MD1 was stage IIb (bony walls, but not yet enclosed by epithelium), while the maximum developmental stage for MD3 was stage IV (fully enclosed and ossified).
Aulonocara at 16 mm SL can be used to illustrate the combination of visualization methods used ( Figure 11). MD3 was fully ossified (Stage IV), but MD2 and

Bone Remodeling through Ontogeny in Aulonocara
The examination of whole fish showed AP staining localized to the edges of the epithelial canal pores in smaller fish and to the edges of the bone fronts (canal walls) within dermal tissue, which were starting to enclose neuromasts in larger fish (e.g., 12 mm SL; Figure 12). In the smallest size class examined (8 mm SL fish), background AP staining was present in the surface epithelium, but not in association with the mandibular canal neuromasts. AP staining appeared at the edges of the canal pores in the 12 mm SL fish, as well as in the canal walls, especially at the articulation between the dentary and anguloarticular bones. In the larger, 16 mm SL fish, AP staining was visible at the edges of the epithelial canal pores, but also on the edges of the canal walls (bone fronts). In 20 mm SL fish, AP staining was only associated with the bone fronts, not the epithelial canal pores. In 24 mm SL fish, the tympanums (where bony pores were bigger than epithelial pores, so epithelium overlaying the wider openings of the bony pores) were visible.
TRAP staining showed a different pattern. In the smallest fish (8 mm SL), TRAP marked two of the five mandibular neuromasts and some tooth buds in the lower jaw ( Figure 12). In fish > 12 mm SL, TRAP was visible in the canal walls (bone fronts). In 16 mm SL fish, TRAP was also visible in the bony canal roof enclosing the neuromasts. In some cases, there appeared to be more TRAP staining in the medial canal wall than in the lateral canal wall enclosing a neuromast. Unfortunately, TRAP-stained fish showed a high degree of background staining (eg. Figure 12H), which limited further observations of TRAP activity in whole fish.
An analysis of histological material for AP and TRAP staining revealed two peaks in osteoblast activity during ontogeny -one in 12 mm SL fish and another in 24 mm SL fish ( Figure 14). Osteoblast activity was found in the canal roof (Quadrant 4) of all Regions along the canal in all five size classes except the smallest one (8 mm SL; Figure   15). The MD2 Region was the only Region showing osteoblast activity in the canal walls   Figure 16).
However, the neuromast-associated Regions had slightly more osteoclast activity in the canal floor (behind the neuromast) than the inter-neuromast Regions.

Bone Remodeling through Ontogeny in Tramitichromis
As in Aulonocara, AP staining was present in all five size classes of fish, and was associated with the epithelial canal pores in smaller animals (<16 mm SL). In 12 mm SL fish, AP staining was visible in the bone fronts representing the growing canal walls (Quadrants 2, 3) over each neuromast. In both the 16 and 20 mm SL fish, AP stained the edges of epithelial pores, but in the largest fish (24 mm SL), staining around the epithelial pore edges was lower, approximately equal to the level of the background staining.
Whole fish stained with TRAP also had high levels of background staining ( Figure 12H), but some staining was visible in the tooth buds and some mandibular neuromasts in the smallest fish (8 mm SL; Figure 12F). In the 16 and 20 mm SL fish, TRAP staining was also apparent in the canal walls and around the epithelial pores of the canals. There was very little canal-associated TRAP staining visible in the 12 mm SL and in the largest (24 mm) fish.
In Tramitichromis, all Regions along the canal (MD2, mid 2-3, MD3, mid 3-4, and MD4) showed a peak in osteoblast activity in 12 mm SL fish, with no second peak in the 24 mm SL fish, which was observed in Aulonocara ( Figure 17). The MD3 and MD4 Regions had nearly identical patterns of osteoblast activity (Figure 18). There was activity in all quadrants around the canal circumference, but especially in the canal roof (Quadrant 4), with a peak at 12 mm SL and a decrease in larger animals. MD2, however, showed a slightly different pattern with no osteoblast activity in the canal roof (Quadrant 4) but activity behind the neuromast (Quadrant 1) before the staining peak (12 mm SL).
In larger animals, there was no activity behind the neuromast (Quadrant 1) and activity in the canal roof (Quadrant 4) decreased with size. In the inter-neuromast Regions, the canal roof (Quadrant 4) also showed the most intense osteoblast activity. With one exception (MD2 Region in 12 mm SL fish), all Regions showed the same overall pattern of osteoblast activity along canal length in each size class. The smallest fish (8 mm SL) had the strongest activity in the canal floor (Quadrant 1 in neuromast-associated Regions), but in the larger fish, the canal roof (Quadrant 4) was the most darkly stained, with a staining peak in 12 mm SL fish that decreased in larger fish. As in Aulonocara, osteoblast activity was not observed on the external surface of the canal.
Osteoclast activity peaked in 16 and 20 mm SL fish and diminished somewhat in the 24 mm SL fish. During the osteoclast activity peak, several Regions along the canal had non-uniform distributions of osteoclast nuclei around the canal ( Figure 19, Table 2).
The distribution of osteoclast activity was non-uniform in the MD2 Region in fish >16 mm SL fish, in both inter-neuromast Regions in the 16 and 24 mm SL fish, and in MD4 in the 20 mm SL fish (Table 2). Generally, the neuromast-associated Regions in fishes and canal roof (Quadrant 4) in fish >16 mm SL.

"Ring" Staining
In Aulonocara, A ring-like pattern of AP staining was observed around superficial neuromasts, including presumptive canal neuromasts, prior to enclosure in a canal ( Figure 13). In fish < 12 mm SL, AP was visible in the epithelium and in rings around both superficial and canal neuromasts. Though the rings around the surrounding superficial neuromasts were very dark, the rings around the canal neuromasts were faded in 12 mm SL fish, and were not visible in fish >16 mm SL. AP staining in a dark ring around superficial neuromasts was still evident in the largest fishes examined (24 mm SL). This ring of AP seemed to disappear as presumptive canal neuromasts were enclosed in the canal, but the staining persisted through ontogeny in neuromasts that remained superficial.
In Tramitichromis, AP staining was present in a ring pattern around superficial neuromasts in all size classes and presumptive canal neuromasts in fish <16 mm SL. In 8 mm SL fish, all neuromasts were marked by a bulls-eye pattern of AP staining, featuring a darkly stained center and a discrete ring of staining that was darker in superficial neuromasts than presumptive canal neuromasts. Despite heavy background staining, superficial neuromasts were circled by a ring of AP in fishes in all size classes.

An Analysis of MD3 in Detail
In order to resolve more detailed patterns, a finer scale analysis of the MD3 region was carried out in both species (Figure 20).
In Aulonocara, peaks in osteoblast activity in 12 and 24 mm SL fish were evident in the vicinity of neuromast MD3 (Positions 17-21) and there was little variation in osteoblast activity among positions along the length of the neuromast (Figure 21). A sizerelated pattern in radial distribution of osteoblast activity was evident. In 12 mm SL fish there was strong osteoblast activity in the canal roof (Quadrant 4) and behind (under) the neuromast (Quadrant 1). Osteoblast activity appeared to decrease overall in 16 mm SL fish, which showed consistent osteoblast activity in the canal roof (Quadrant 4), but inconsistent activity behind the neuromast (Quadrant 1). In the 20 mm SL fish the only osteoblast activity was in the canal roof (Quadrant 4), but was weaker than in smaller fish (12 and 16 mm SL). In the largest fish (24 mm SL), osteoblast activity was once again very high in the canal roof (Quadrant 4), with little to no activity around the rest of the canal circumference.
In contrast, a general trend of increasing osteoclast activity with increasing fish size was noted ( Figure 22). In 8 mm SL fish, little TRAP staining was apparent, while in 12 mm SL fish there was consistently low osteoclast activity, but the radial distribution of In Tramitichromis, in each size class, osteoblast activity was consistent among positions along the neuromast ( Figure 23). Generally, there was one peak in osteoblast activity in 12 mm SL fish that then decreased in larger fish. The 8 mm SL fish showed moderate staining below the neuromast (Quadrant 1) along its length (Positions 17 -21).
In 12 mm SL fish there was strong staining in the canal roof (Quadrant 4) and moderate staining around the rest of the canal circumference, but the 16 mm fish showed little staining in the canal roof. The 20 mm SL fish showed even less staining in the canal roof, and Position 18 had no staining in any quadrant. The 24 mm SL fish also showed no AP staining in any of the quadrants around the canal circumference.
The pattern of osteoclast activity was more variable than that of osteoblasts.
However, osteoclast activity showed a more consistent pattern in Tramitichromis than it did in Aulonocara ( Figure 24). As in Aulonocara, there appeared to be a In both Aulonocara and Tramitichromis, there did not appear to be any spatial pattern of canal enclosure, for example, rostral positions enclosing first, within a given neuromast (Figures 25 and 26).

DISCUSSION
This study used histochemical methods to determine the roles of bone-forming (osteoblasts) and bone-resorbing (osteoclasts) cells in the initial morphogenesis of the lateral line canals, the growth of the canals (increases in diameter), and the generation of narrow and widened canal phenotypes during ontogeny. In both study species, Aulonocara stuartgranti (widened canals) and Tramitichromis sp. (narrow canals), osteoblast activity was concentrated in the roof and floor of the canal, while osteoclast activity was highest in the lateral canal walls that extend to enclose the neuromasts.
Another similarity between the two species was the presence of two pulses of bone cell activity during the time period sampled: one early pulse defined by bone deposition in both species, and later in development, a second pulse of either bone resorption (Tramitichromis) or bone remodeling (Aulonocara). The patterns and timing of osteoblast and osteoclast activity revealed by this study corroborate the observations by Bird and Webb (2014), that dissociated heterochrony, that is, differences in developmental timing, is a major factor in the morphogenesis of divergent lateral line canal phenotypes. Further, in the context of neuromast-centered canal morphogenesis, this study suggests that the influence of neuromast proximity on the spatial pattern of canal morphogenesis at the cell level may vary depending on ultimate adult canal phenotype.

Osteoblast and Osteoclast Activity during Initial Canal Morphogenesis
This study showed that osteoblasts and osteoclasts are present in mid-to latestage larvae and early juveniles (8 -24 mm SL) in both Aulonocara and Tramitichromis, but that the intensity and pattern of the activity of these two cell types vary in space and While both osteoblasts and osteoclasts were present in the larvae and early juveniles examined (8 -24 mm SL), there were two distinct time periods of marked high activity in both species (Figure 27). An early peak in osteoblast activity occurred at 8 -12 mm SL (with low levels of osteoclast activity), followed by a peak in osteoclast activity (20-24 mm SL). Between these two peaks (e.g., in fishes of 16 mm SL), both osteoblast and osteoclast activity were relatively low.
The first peak in osteoblast activity (8-12 mm SL) suggests intense bone deposition, but some differences were observed between species. Tramitichromis had higher levels of osteoblast activity than Aulonocara, and the first peak in osteoblast activity in Tramitichromis occurred in smaller fishes (8 mm SL) than in Aulonocara (12 mm SL). Similarly, there were differences in timing of the second peak in osteoclast activity between species. Additionally, In Tramitichromis, the second peak was defined by high osteoclast activity, but little to no osteoblast activity. Thus, the MD canal in Tramitichromis appears to develop as the result of one pulse of bone deposition and a subsequent pulse of bone resorption. By contrast, after the first peak of bone deposition in Aulonocara, the second peak is defined by high levels of both osteoclast and osteoblast activity. Thus, the process of canal morphogenesis in Aulonocara appears to occur as a result of a pulse of bone deposition followed by a pulse of simultaneous osteoblast and osteoclast activity (active bone remodeling). It is concluded that the generation of the narrow and widened canal phenotypes is the result of these differences in the spatial pattern and timing of osteoblast and osteoclast activity.

Osteoblasts and Osteoclast Activity in Canal Growth
The spatial and ontogenetic distributions of osteoblasts and osteoclasts can also be used to understand how the canal diameter increases, a process that continues after initial morphogenesis in concert with fish growth . In both species, the early pulse of osteoblast activity is followed by high levels of osteoclast activity. This suggests a general theme of canal growth by alternating periods of bone deposition and resorption or remodeling resulting in the continuous expansion of canal diameter. This was noted in Aulonocara and Tramitichromis by Bird and Webb (2014), in zebrafish by Moore (2008), and in cichlids by Tarby and Webb (2003), and thus appears to be a generalized pattern of canal development. In both Aulonocara and Tramitichromis, osteoclast activity is low in early larval ontogeny (<16 mm SL) but is higher in larger larval and juvenile fish (16-24 mm SL). This can account for the increase in canal diameter in these species reported by Bird and Webb (2014).
In Tramitichromis, despite low levels of osteoclast activity throughout ontogeny relative to Aulonocara, there are more active osteoclasts in larger animals (16 -20 mm SL) than in smaller animals (<16 mm SL). These osteoclasts presumably increase the diameter of the canal lumen by removing bony matrix on the internal surface of the canal.
In Aulonocara, the level of osteoclast activity (indicated by the numbers of osteoclasts present) is also increased in larger individuals (compared to smaller ones), and these high levels of osteoclast activity on the internal surface coincide with a second pulse of osteoblast activity on the internal surface of the canals.
These observations support the idea that in Tramitichromis, canals are formed by a simple pattern: a pulse of bone deposition (at 8-12 mm SL) followed by bone resorption which begins at a low levels and increases as the fish grows, peaking in larger fish (20-24 mm SL). In contrast, in Aulonocara, there is more osteoclast activity overall through time, perhaps because bone resorption is necessary to continuously enlarge the (widened) canal lumen while actively remodeling the new bone that osteoblasts are laying down.

Thus, increased osteoclast activity in Aulonocara shows that canal morphogenesis in
Aulonocara is a sequence of bone deposition and bone remodeling (bone deposition and resorption, as opposed to only resorption, as in Tramitichromis). The fish studied only represent a snapshot of canal enlargement during a limited size interval (8 -24 mm SL) in larvae and young juveniles, so it is possible that patterns of alternating bone deposition by osteoblasts and bone resorption by osteoclasts (in Tramitichromis), or bone deposition by osteoblasts and bone remodeling by simultaneous activity of osteoblasts and osteoclasts (in Aulonocara), are repeated in even larger fishes (juveniles and adults). In the case of Aulonocara, another possibility is that the high osteoblast and osteoclast activity levels in the largest fish studied (20 -24 mm SL) persist, constantly remodeling the canal as the fish continues to grow.

Evidence of Neuromast-Centered Canal Morphogenesis at the Cell Level
Neuromast-centered canal morphogenesis is the idea, supported by the observations of several authors (reviewed in Webb 2014b) that canals tend to start to form in the vicinity of individual neuromasts. This suggests that neuromasts are somehow involved in the process of canal morphogenesis, potentially in some instructive or inductive capacity (J. Webb, pers. comm.;discussed in Hall 2015). It had been predicted that differences in either the intensity, ontogenetic timing, or radial distribution of osteoblasts or osteoclasts would differ in the vicinity of canal neuromasts (neuromastassociated regions) versus locations between canal neuromasts (inter-neuromast regions).
However, the results of this study showed that osteoblast and osteoclast activity in neuromast-associated versus inter-neuromast regions were more similar than expected.
In Tramitichromis, in particular, both neuromast-associated and inter-neuromast regions had similar osteoblast and osteoclast activities (Figure 28b). By contrast, in Aulonocara, the first pulse of osteoblast activity (12 mm SL) had a different radial distribution in neuromast-associated regions versus inter-neuromast regions along the canal (Figure 28a). Specifically, all neuromast-associated regions had osteoblast activity in the canal floor (Quadrant 1), while inter-neuromast regions had none ( Figure 28a).
Thus, the consistency in the distribution of osteoblast activity between neuromastassociated and inter-neuromast regions along the canal in Tramitichromis was in stark contrast to the variability in the distribution of osteoblast activity in Aulonocara ( Figure   29). These data are consistent with the observation that widened canals (Aulonocara) form as a series of scallops, or oscillations in the height of the canal walls along the canal during canal morphogenesis, with a maximum height in the vicinity of individual neuromasts. Thus, the results of this study suggest that narrow canals in Tramitichromis form as more continuous trenches with more uniform canal walls in contrast to the "scallops" observed along the widened canals of Aulonocara, where the canals wall are more prominent at neuromast positions.

CONCLUSIONS
As the first study to assay not just osteoclast activity, but both osteoblast and osteoclast activity in the cranial lateral line canals of fishes, this investigation revealed key variation in osteoblast and osteoclast activity through ontogeny. In both species, osteoblast activity (bone deposition) was concentrated in the canal floor and roof, while osteoclast activity (bone resorption) was focused in the canal walls. There was a period of bone deposition early in development, followed by a period of bone resorption (in Tramitichromis) or bone remodeling (bone deposition and resorption, in Aulonocara).
The development of Tramitichromis canals is marked by a strong, early pulse of bone deposition and a subsequent moderate pulse of bone resorption, while Aulonocara canals are formed by a less intense pulse of bone deposition, slightly delayed relative to Tramitichromis, and followed by a period of active bone remodeling. These observations provide cell-level evidence of bone development through ontogeny that fits the pattern of dissociated heterochrony (described by Bird and Webb 2014) that underlies the divergence of canal phenotype in these two species.
Considering the phylogenetic distribution of cranial lateral line canal phenotypes among bony fishes (Webb, 2014b) Future studies should look more closely at early ontogeny using in situ hybridization techniques in combination with histochemical methods to more precisely define the spatial distributions of osteoblasts and osteoclasts and their gene products.
Additional experiments involving pulses of calcein or other stains throughout ontogeny and during initial canal enclosure would provide insight regarding differences in patterns of canal ossification (calcium deposition) and how these patterns contribute to phenotypic differences in adults.