Geometrisch-Morphometrische Analysen Von 3- Dimensionalen Computer-Tomogrammen Des Kopfskeletts Von Buntbarschen

Julia Maria Kalt

Geometrisch-morphometrische Analysen von 3- dimensionalen Computer-Tomogrammen des Kopfskeletts von Buntbarschen

Geometric-morphometric analyses from 3-dimensional computer-tomograms of cichlid heads.

Masterarbeit

zur Erlangung des akademischen Grades Master of Science an der Naturwissenschaftlichen Fakultät der Karl-Franzens Universität Graz

unter Betreuung von

Univ. Prof. Mag. Dr. Christian Sturmbauer, Institut für Zoologie

Graz, 2017

Danksagung (Acknowledgements)

Ich möchte an dieser Stelle allen Personen danken, die mir mein Studium sowie diese Masterarbeit ermöglicht haben:

Univ. Prof. Mag. Dr. Christian Sturmbauer für die gute Betreuung während meiner Masterarbeit, sowie die Ermöglichung dieser Arbeit. Auch noch möchte ich mich für die unvergesslichen Reisen an den Tanganyikasee bedanken.

Dr. Heather More für ihr Feedback sowie all ihr technisches Knowhow ohne dieses meine Masterarbeit nie zu einem Abschluss geführt hätte. Aufgrund von Sprachbarrieren möchte ich auch ein paar Worte in ihrer Muttersprache schreiben. Thank you Heather for your support during my master thesis. Without your technical knowledge and your programs it would not have been possible to finish my thesis. You are an awesome person.

Alles in allem möchte ich auch einigen Leuten die mich auf meinen Weg des Studiums begleitet haben namentlich erwähnen um ihnen die Aufmerksamkeit zu schenken, die sie verdienen:

Als erstes möchte ich mich bei meiner Familie bedanken, die mich in meinem Studium sowohl mental als auch finanziell unterstützt hat, ohne sie wäre meine Karriere an der Uni nicht möglich gewesen. Vielen Dank Mama, Maria Sandner, Papa, Rudolf Kalt und Marlene Kalt.

Außerdem möchte ich einem der wichtigsten Menschen in meinem Leben danke sagen. Danke Fritz Neunegger für Deine Unterstützung, sei es das Durchlesen meiner Arbeit oder auch mentale Unterstützung gewesen. Ohne Dich an meiner Seite, hätte ich das alles nicht geschafft.

Last but not least möchte ich meinen Freunden danken, die ich im Laufe meines Studiums kennengelernt habe. Danke Mädels für eure Unterstützung. Besonderer Dank gilt Alexandra Wunder, die meine Arbeit immer wieder durchgelesen hat um sie so perfekt wie möglich zu machen und danke auch für die unzähligen Gespräche die wir geführt haben. Danke auch an Jacqueline Grimm für die unzähligen Gespräche wenn es mal nicht so voran ging.

Table of Contents

Deutsche Zusammenfassung………………………………………………………………………………………………………………4 Abstract………………………………………………………………………………………………………………………………………………5

1. Introduction ...... 6 2. Material and Methods ...... 10 Species ...... 10 Processing procedure ...... 11 Landmarks ...... 12 Software and statistical background ...... 13 3. Results ...... 17 Landmarks ...... 17 Oral jaw module ...... 19 Pharyngeal jaw module ...... 27 Connective bone - Interopercle ...... 33 4. Discussion ...... 35 Landmarks ...... 35 Oral jaw module ...... 35 The pharyngeal jaw module ...... 39 Connective bone –Interopercle ...... 41 Why using three dimensional landmarks instead of two dimensional ones?...... 41 5. References ...... 43 6. Supplementary Information ...... 48

Deutsche Zusammenfassung

Die Buntbarsch-Gruppierungen der Haplochromini und Tropheini aus den drei großen Afrikanischen Seen, Tanganjika, Malawi und Viktoria, zeigen eine hohe Diversität im Zusammenhang mit Morphologie, Nischen, Färbung und Genetik. Die Familie der Buntbarsche (Cichlidae) stellt ein hervorragendes Modellsystem für die Untersuchung von evolutionären Prozessen dar. Um den Zusammenhang morphologischer Unterschiede mit dem ökologischen sowie phylogenetischen Hintergrund der Arten herauszufinden, wurden Mikro-Computer-Tomogramme des Viscerocranium von 136 Fischen aus siebzehn Arten erstellt. Mit Hilfe von dreidimensionaler geometrischer Morphometrie wurden sieben ausgewählte Knochen aus drei funktionellen Modulen des Kopfskeletts analysiert, um sie mit ökologischen oder phylogenetischen Informationen innerhalb der - und zwischen den - drei Seen in Zusammenhang zu setzen. Die Knochen des vorderen Kieferapparates (Dentale, Artikulare, Maxillare, und Premaxillare) zeigten ein klares ökologisches Signal, während die Schlundkiefer inclusive des Verbindungsknochen-Ineroperculare, ein klares phylogenetisches Signal zeigten. Die beweglichen Knochen des vorderen Kieferapparates sind klar ökologisch unterscheidbar, wohingegen die nicht beweglichen Knochen des Schlundkieferapparates inclusive des Interoperculare keine klare Auftrennung der ökologischen Nischen zeigten.

Abstract

The cichlid fish tribes Haplochomini and respectively, the Tropheini, from the Great East African Lakes Tanganyika, Malawi and Victoria, are a highly diverse group in terms of morphology, niche diversity, colour, behavior and genetic background. The cichlids as a group represent an excellent model system for analyzing evolutionary processes. To define morphological differences to find out about the underlying ecological and phylogenetic signal in the three lakes, micro–computer tomograms of the viscerocranium of 136 individuals belonging to seventeen species were produced. With the help of three- dimensional geometric morphometric methods we analyzed seven selected bones assigned to three separate functional modules, to gain information about ecological and phylogenetic relations within and between the three lakes. The defined modules act independently from each other in terms of anatomical and developmental issues. In the oral jaw module (dentary, articular, maxilla and premaxilla) a clear ecological signal was found, whereas in the pharyngeal jaw module and the independent connective interopercle bone a strong phylogenetic signal was detected. The moveable oral jaws were clearly morphologically distinguished and highly informative in the context of ecological signals, while the non- moving upper pharyngeal jaws including the interopercle did not show such a clear ecological signature.

1. Introduction

The three East African Great Lakes, Tanganyika, Malawi and Victoria, harbour cichlid fish species flocks which are endemic in each lake (Fig. 1). Within the last 6 million years more than 2000 species evolved in these three lakes alone that represent the most species-rich family within the teleost fishes (Koblmüller et al. 2008; Turner et al. 2001). The species flocks differ from each other in age, complexity, number of species and also in the degree of morphological diversity (Fryer & Iles 1972; Mayr 1984; Poll 1986; Turner 2007; Sturmbauer et al. 2011). These differences are the result of the diverse histories of the three East African Lakes involving the respective geological background and climate and the seeding species. These diverse species flocks are further influenced by lake level fluctuations, caused by variation in rainfall, temperature and evaporation. All in all, these three lakes have their own dynamic basin morphology (Tiercelin & Mondeguer 1991; Cohen et al. 1993; Cohen et al. 1997; Scholz & Rosendahl 1988; Lezzar et al. 1996; Delvaux 1995; Johnson et al. 1996; Gasse et al. 1989). Lake Tanganyika represents the oldest of the three lakes. Its central basin was formed 9-12 MYA (Scholz & Rosendahl 1988; Cohen et al. 1993), with an estimated age of a truly lacustrine environment of about 5-6 million years (Koblmüller et al. 2008; Tiercelin & Mondeguer 1991; Lezzar et al. 1996; Cohen et al. 1997). In the case of Lake Tanganyika at least three shallow and swampy proto-lakes fused to its final form of a single deep lake with a maximum water depth of 1470m. It is the second-deepest lake in the world after Lake Baikal (Scholz & Rosendahl 1988; Gasse et al. 1989; Cohen et al. 1993; Lezzar et al. 1996; Cohen et al. 1997; Tiercelin & Mondeguer 1991). The considerably younger lake Malawi rift basin developed about 4.5 million years ago and has now a maximum depth of close to 700m (Delvaux 1995). The youngest lake in this study is Lake Victoria with an estimated maximum age of around 400,000 years, so it is significantly younger than the other two representatives. It is also quite shallow with a maximum water depth of only 69m (Johnson et al. 1996). According to the age of the lakes, the age of the major diversification events differs between the lakes.

1. Introduction

Victoria

Tanganyika

Malawi

Figure 1. Map of Africa (without Madagascar) including the three East African Lakes (Tanganyika, Malawi and Victoria) that the study species belong to.

The oldest radiation was ongoing in Lake Tanganyika and peaked about 5-6 million years ago - when the truly lacustrine ecosystem emerged in the oldest lake of the East African Lakes (Cohen et al. 1997; Lezzar et al. 1996; Tiercelin & Mondeguer 1991). The Lake Malawi cichlid radiation was dated to in-between 0.57 and 1 million years – based on the onset of the last permanent truly lacustrine habitat of this lake (Delvaux 1995; Sturmbauer et al. 2001). With a maximum age of about 200,000 years for the radiation of the Lake Victoria superflock (Meyer et al. 1990; Nagl et al. 2000; Verheyen et al. 2003) it is the youngest of the three lakes.

The cichlid fish species flocks in the lakes have undergone adaptive radiation within the maturing (and fluctuating) lake ecosystem to produce multiple body shapes, trophic specializations, color morphs and behaviors, in order to diversify into almost every conceivable ecological niche (Fryer & Iles 1972; Greenwood 1984; Sturmbauer et al. 2011). 7

1. Introduction

The adaptive radiations of the two younger lakes (Malawi and Victoria) consist of one of tribes only, the Haplochromini (Poll 1986). Poll assigned the members of this tribe in Lake Tanganyika to the tribe Tropheini (Salzburger et al. 2005). The so-called modern haplochromines consist of the endemic tribe Tropheini in Lake Tanganyika on the one hand, and a complicated assemblage of riverine species plus additional flocks of lacustrine haplochromines on the other hand, such as those of Lakes Malawi, Victoria, Kivu and Turkana (Salzburger et al. 2005; Verheyen et al. 2003; Koblmüller et al. 2008). With their representatives in each of the lakes, the haplochromines (and tropheines) occupy almost every ecological niche in Lake Victoria, Malawi and Tanganyika, a result of convergent evolution (Kocher et al. 1993; Kassam et al. 2006; Sugawara et al. 2005). Probably because of the younger evolutionary age the species flocks in the younger lakes are less diverse in terms of phenotypic, behavioral and genetic divergence, in comparison with those in Lake Tanganyika (Meyer et al. 1990; Verheyen et al. 2003; Greenwood 1979; Greenwood 1980). All haplochromines from Lakes Malawi and Victoria, as well as the Tropheini from the oldest Lake Tanganyika, are maternal mouthbrooders, a phenomenon whereby the female incubates the eggs and the fry in its buccal cavity (Koblmüller et al. 2008). The study species of my thesis work all belong to these two tribes: the Malawi and Victoria species to the Haplochromini and the Tanganyika species to the Tropheini or the Haplochromini.

Concerning their trophic anatomy, cichlid fishes possess two sets of jaws - the oral jaws and the pharyngeal jaws - located in the back of the buccal cavity (Liem 1973). Each set consists of an upper and lower jaw. Because of having two sets of jaws which are functionally decoupled from each other, cichlid fishes could diversify more rapidly than other fish groups, to specialize towards almost every trophic niche. Thus, their fast diversification is considered to be the consequence of this particular "key innovation", in the form of two independent jaw modules (Hulsey et al. 2006; Liem 1973).

I chose geometric morphometry to define the differences between the species, with respect to their food niche and lake origin. It is a good tool to retrieve information about shape and size- and accordingly this method is often used in morphological studies. Over the last years the state of knowledge concerning geometric morphometric methods and landmark setting has increased substantially for cichlids (Clabaut et al. 2007). Landmarks can be positioned in

1. Introduction two or three dimensional coordinate systems, to define a kind of positional homology across the measured specimens (Bookstein 1991). The innovative aspect in this thesis was the establishment and use of 3-dimensional landmarks instead of 2-dimensional ones.

Figure 2. Overview of the seven bones we used in this study. The anterior part in blueish represents the oral jaw module, while the reddish-yellowish bones belonging to the pharyngeal jaw module. The bone in the middle represents the interopercle - the connection between the two sets (pink).The figure was modified from one by Heather More, I added the interopercle. The two independent sets of jaws are shown - the oral jaw module for getting and the pharyngeal for processing the food.

Therefore, we produced three-dimensional images by means of micro-computed tomography, and used these images to reconstruct the head, to capture its shape by setting three-dimensional landmarks precisely on the targeted bones.

Earlier two-dimensional studies showed that there are differences in the viscerocranial bones among populations of the species Tropheus moorii (Postl et al. 2008). There are also former studies dealing with the pharyngeal jaws showing that there are differences among the tribes (Muschick et al. 2012). The goal of this study was to find out if there are differences in the oral, as well as the pharyngeal jaws, with respect to species, niche or lake. We analyzed seven bones belonging to the two functional modules. We also investigated if the species specialized towards the same ecological niche in a different lake are also morphologically similar. We selected 17 species from the three East African Lakes, belonging to four trophic niches, to test if there are correlations between the different classifiers of species, niche, and lake. As an innovative aspect of this thesis, we tested for these correlations with the help of geometric morphometric methods in 3 dimensions. 9

2. Material and Methods

Species

We chose 17 species (n=136) in four different niches from all three Great East African lakes, Tanganyika, Malawi and Victoria. All specimens were bred and brought up under standard conditions in the laboratory at the University of Graz. The dietary niches were defined as carnivore (as well as shrimp), omnivore, algae browser and algae grazer. The species are listed in Table 1.

Table 1. The 17 species used in the study. The table also includes the feeding specialization and the lake each species belong to.

Feeding Species specialization Lake

Ctenochromis horei carnivore Tanganyika Gnathochromi spfefferi carnivore Ta nganyika Sciaenochromis fryeri carnivore Malawi Haplochromis thereuterion carnivore Victoria

Limnotilapia dardennii omnivore Tanganyika Cynotilapia afra omnivore Malaw i Maylandia zebra omnivore Malawi

Simochromis diagramma algae browser Tanganyika Tropheus moorii algae browser Tanganyika Labeotropheus fuelleborni algae browser Malawi Labeotropheus trewavasae algae browser Malawi Neochromis omnicaeruleus algae browser Victoria

Petrochromis trewavasae algae grazer Tanganyika Petrochromis famula algae grazer Tanganyika Petrochromis polyodon algae grazer Tanganyika Petrotilapia sp. "yellow chin Chewere " algae grazer Malawi Petrotilapia sp. "thick bars" algae grazer Malawi

2. Material and Methods

Life stages We defined different life stages; juvenile (J), young adult (A), and fully grown-up (G). The juvenile stage was defined as the moment when the yolk sack had been fully absorbed and the yolk gap was completely closed. The adult stage was defined as the time point when the males were reproductively mature and show the first sexual behavior, such as courtship, breeding color. The fully grown-up stage was defined as the time point as of which the fish had spawned multiple times or had offspring. Also, the fish which died because of old age were included in this group.

Processing procedure

The individuals were killed by a lethal dose of MS222 (Tricainemethanesulfonate). Afterwards the fish were sexed and put into 4% formalin in 10% PBS (phosphate-buffered saline) solution, with a pH of 7.4 for one week. Small fish were scanned whole, larger individuals were cut and only the head analyzed. The viscerocranium, which includes the two sets of jaws, were scanned by means of the micro-computertomograph SCANCO Medical MicroCT 40.

Scanning procedure First, the front part of the fish was put into a tube for the microCT. After the tube was placed in the sample chamber, we started the scanning procedure targeting the entire viscerocranium. The tube containing the specimen(s) was pre-scanned to define the appropriate section containing the oral and pharyngeal jaws, to be scanned in a resolution between 2µm and 18µm. Juveniles were scanned in the highest possible resolution (2µm). The resolution used for adults varied between 10-12µm, whereas the fully grown-up fishes were scanned between 10-18µm. The scan diameter varied between 11mm and 36mm.

2. Material and Methods

Data preparation The raw-data from the microCT were converted into TIFF-files. For the subsequent analysis of these pictures we cropped and transformed the files into DICOM format using a custom written MATLAB-code.

Figure 3. 3D reconstruction produced in TINA. The figure show a Ctenochomis horei male individual from the left lateral side.

Landmarks

Target structures We focused on 4 bones belonging to the oral jaw functional unit, 2 bones belonging to the pharyngeal jaws and one bone in between the two sets of jaws (connecting bones), to analyze the differences in relation to the dietary niches, lakes, and species. The two separate sets of jaws were analyzed separately – in accordance to the definition as two different functional modules. The selected oral jaw bones were the dentary, articular, maxilla and premaxilla, representing the oral jaws and the selected pharyngeal jaw bones were the upper pharyngeal jaws (pharyngobranchials 2, 3, 4) and the lower pharyngeal jaw. The last bone in this thesis is the interopercle - the connection between the two separate sets of jaws.

2. Material and Methods

The sex of juveniles was not discriminated in the analyses. Concerning adult and fully grown fish, both males and females were included in the analysis, as previous work has shown that the sex effect is relatively small in relation to interspecific differences (Herler et al. 2010 showed it in Tropheus). Many two-dimensional morphometric studies on cichlid fish have been conducted at the Institute for Zoology of Graz in the lab of Dr. Sturmbauer (Kerschbaumer et al. 2014; Wanek & Sturmbauer 2015; Postl et al. 2008; Herler et al. 2010; Wunder 2016). According to the convention of these studies, we placed landmarks on the bones of the left lateral side. However, we used anterior, posterior ventral, and lateral views of the head, to confirm the optimal positioning of the landmarks. We defined 48 landmarks in total, of which 7 were semi-landmarks. We note that the semi-landmarks were excluded in the analyses presented in this thesis. Additionally, 6 landmarks representing the tooth of the upper and the lower pharyngeal jaws were excluded because they were difficult to place consistently and were highly variable. We placed 6 landmarks on the dentary, 6 landmarks on the articular, 6 landmarks on the maxilla, and 5 landmarks on the premaxilla. For the pharyngeal jaws we set 5 landmarks on the lower pharyngeal jaw. We carried out two separate analyses for the lower pharyngeal jaw, one including the landmark(s) addressing the depth of the lower pharyngeal jaw, and one without this point- to make it more comparable with previous two-dimensional analyses. We also used the left side of the lower pharyngeal jaw only, assuming it to be symmetric. We placed 4 landmarks on the upper pharyngeal jaw. Last but not least, we set 3 landmarks on the interopercle – the connective bone between the two jaws units (oral and pharyngeal).

Software and statistical background

TINA To combine the 2D images into a 3D reconstruction we used the program TINA (Schunke et al. 2012). In TINA we loaded the 3D reconstruction, as well as the three axis views separately. We oriented the landmarks in the x-axis, the y-axis and the z-axis, to define the landmark as precisely as possible. With the help of various viewpoints it was possible to even set cryptic landmarks, for example the one in the articular cavity.

2. Material and Methods

Figure 4. Volume rendering interface in TINA, including the two sigmoid curves for the “Scalar Classification” as well as for “Gradient Classification”. This volume rendering interface was originally made for mouse-skulls (Schunke et al. 2012).

Volume rendering interface in TINA The volume rendering interface was originally produced for mouse-skulls. The “Scalar Classification” curve is shown as a graph in which the x-axis displays the grey level (intensity) in the original 3D image loaded into the sequence tool, and the y-axis displays the opacity of converted pixels in the volume rendered picture. We found the best results using the default sigmoid curve. The “Render quality” is the balance between the rendering speed and the image quality. We used the option “fast” where all voxels are rendered. This option makes calculation relatively slowly but it turned out to produce the most suitable outcome in displaying the 3D reconstruction. We used the standard volume rendering settings (provided as a .txt file) which were produced to digitalize micro-CT pictures of mouse skulls. This file is included with the program in the folder with the extra files, and is called “tmlt_surface_style.txt. Each pixel in the 3D volume represents a vector through the rendered image. The renderer runs along these vectors, the intensity as well as the intensity gradient of every voxel (3D-image) is being converted into opacity and reflectivity. All these

2. Material and Methods factors together produce the intensity of the final pixel in the rendered picture shown in Figure 4 (the settings) and Figure 3 a 3D reconstruction of a fish head (Bromiley et al. 2012).

Calculate the scale factor and separate the TPS files We first set all the 48 landmarks on the fish reconstruction, then separated the generated TPS-files into TPS- files for each bone later with the help of a Matlab program. Geometric morphometric analysis was then continued in MorphoJ (version 1.06d). The TPS files had to contain the scale factor in µm per voxel, which is unfortunately not automatically computed in TINA, so we had to calculate it by hand. The scale factor is important for carrying out the Procrustes analyses and was calculated from the raw data of the program (voxelsize and downsampling).

MorphoJ Geometric morphometric analyses were carried out in the computer program MorphoJ (version 1.06d)(Klingenberg 2011). This is one of the standard programs to carry out Procrustes superimposition analyses, which eliminates variation in position, size and orientation from the landmark coordinates (Bookstein 1996; Dryden & Mardia 1998; Goodall 1991; Zelditch et al. 2004). In MorphoJ one can look at the different views of the landmarks over all 3 axes. After Procrustes superimposition, for our data the view of axis 1 vs. 2 reflects the frontal view at the bone. Axis 1 vs. 3 characterizes the vertical view, and the view of axis 2 vs. 3 reflects the lateral view at the bone. We initially included semilandmarks, treating them as “normal” 3-dimensional landmarks, - this increased variation in all the landmarks of the bone, so we did not include semilandmarks in our final analyses.

Regression In MorphoJ we also did a regression on our data to test for allometry. Regression analysis is a technique to forecast the values of one or more dependent variables from the values of one or more independent variables. It is often used for size correction of shape data (see e.g. Klingenberg 2016). We calculated the regression on the centroid size – the dependent variable being the Procrustes coordinates (shape) and the independent variable being the centroid size. Specifically, we carried out a pooled within-group regression analysis. This is a

2. Material and Methods method which uses the size and shape variance of each specimen from the average of size and shape of the group (species in our case), not from the mean of all species, to calculate the variance and the covariance. The pooled within- group regression is a two-step process: the first step centers the data by group (species) removing the average differences among the groups. Step 2 is a standard regression using the centered data. We then carried out a standard multivariate regression of shape, working with the residuals from the multivariate regression of size (Klingenberg 2016).

CVA Canonical variate analysis This is a method to find those shape features which best distinguish among different groups of specimens. For a CVA is important that the specimen assignment is clear, as the group assignment is pre-designated (see the MorphoJ help menu for more details). CVA is a two- step procedure of orthogonal rotation. The first step is a Principal Component Analysis of the original variables/data. The next step is a PCA of the group means for the former in the first- stage produced eigenanalysis (Campbell et al. 1981).

The PCA as well as the CVA- graphs were produced in Python 3.6.

What is the difference between 2D and 3D? Why 3D? The only difference between 2D and 3D Procrustes is that it is more computation-intensive because the matrices are larger. The complicated part in all these analyses is the rotation of the different data. Therefore, Bookstein’s formula for 2D data has been expanded. The biggest problem to overcome in doing a Procrustes with three-dimensional data is to minimize the angles of distance (Zelditch et al. 2004).

3. Results

Landmarks

The landmarks were set and subsequently optimized in a step-wise process. For each bone we defined a preliminary set of landmarks and iteratively changed number and/or exact positioning until a consistent placement for all species in question and all perspectives was reached. Thereby, we considered published papers, for example Albertson (2001) and Frederich (2008) for the oral jaw module. For the second set of jaws we mainly considered the placement of Hellig, as well as other authors (Albertson & Kocher 2001; Frederich, Adriaens, et al. 2008; Hellig et al. 2010). Some of the landmarks were newly defined by us after the anatomically description of Barel (Barel et al. 1976). After setting landmarks on the first 10 specimens we carried out a pre-analysis to see if they were consistent. Then we did the same for at least 10 specimens of a second species to check intra- versus interspecific variation. If necessary, we optimized at each step. After these preliminary analyses we set the landmarks on all fish over all study species. As mentioned above we assigned the seven bones to two functional modules, the oral jaw set including four bones (dentary, articular, maxilla and premaxilla), the pharyngeal jaw module containing two bones (lower and the upper pharyngeal jaw), and the interopercle as , the connection between the two modules. The interopercle does not belong to one of the two modules we were looking at; it belongs to the suspensorium but for reason of simplicity we grouped it to the pharyngeal jaw module in the discussion section. The landmarks are defined in Table 2. For better orientation, the landmarks are depicted on the photos.

3. Results

Table 2. List of the homologous landmarks including the morphological description of the belonging bone (Barel et al. 1976).

Element Landmark Description Dentary 1 rostral tip of the dentary 2 last posterior tooth of the dentary 3 dorsal tip of the coronoid (dentary process) 4 surface dentary immediately rostral to landmark 1 of the articular 5 posterior end of the dentary 6 ventral-most point of the dentary

Articular 1 tip of the rostral process of the articular 2 dorsal tip of the primordial (articular) process 3 dorsal process of the suspensoriad articulation facet 4 postarticulation process (of the suspensoriad articulation facet) 5 retroarticular process 6 rostral process of the coulter area

Maxilla 1 medial process of the palatinad wing 2 lateral process of the palatinad wing 3 posterior end of the maxilla 4 shank process 5 lateral process of the premaxillad wing 6 medial process of the premaxillad wing

Premaxilla 1 rostral-most point of the dentigerous arm 2 dorsal process of the ascending spine 3 dorsal process of the maxillad spine(bumpy) 4 caudal process of the dentigerous arm 5 last tooth of the premaxilla

Lower 1 caudal end of the median suture pharyngeal jaw 2 caudal-dorsal tip of the left horn 3 caudal-ventral tip of the horn 4 rostral edge of dentigerous area

Upper 1 caudal edge of the neurocraniad articulation facet (neur.d art. fct) pharyngeal jaw 2 dorsal-most tip of the caudal-lateral eminence (cI II-emin.) 3 caudal ventral-tip of the upper pharyngeal jaw 4 articulation of pharyngobranchial 3 with pharyngobranchial 2

Interopercle 1 dorsal-rostral point of the interopercle 2 ventral -rostral point of the interopercle 3 caudal-ventral point of the interopercle

3. Results

Only landmarks were selected that could be consistently placed in all the study species. Moreover, we detected only minor sexual dimorphism in the study species, in agreement with earlier findings (Wanek & Sturmbauer 2015). Our overall goals were to first detect species-specific signals, then test for an ecological signal in the oral and pharyngeal jaw modules, and finally test for a lake-specific (phylogenetic) signal.

Oral jaw module

In the oral jaw module four bones were included. The dentary (Figure 5), articular (Figure 7), maxilla (Figure 9), and the premaxilla (Figure 11). All of these bones are moveable bones, and were often analyzed by 2D morphometry in earlier studies: Some with developmental background including juveniles (Fujimura & Okada 2008; Le Pabic, P., Cooper, W.J., Schilling 2016), and some for morphological interest (Kassam et al. 2004; Albertson & Kocher 2001; a Frederich, Adriaens, et al. 2008; Frederich, Pilet, et al. 2008; Otten 1983).

Dentary 3

1 4

5 Figure 5.The dentary from the lateral view with the six homologous landmarks. The figure was modified from Albertson et al., 2001.

We found a strong species-specific differentiation signal for this bone so that one could clearly separate the species from each other: specimens of the same species were always

3. Results clustered together (Fig. 6A). On the graph defining the target classifier “species” the browsers could be separated well. When analyzing the clustering with regard to “niche” as classifier, there was also a clear-cut ecological signal (Fig. 6B). Thus, the dentary is highly informative to separate the trophic niches, with most variation in PC1, but also in PC2, which is necessary to separate the different browser species from each other. The distance between the last tooth and the end of the posterior end of the dentary was narrower in the grazers (orange) than in the carnivores (green), as explained by PC1. A second observation for the first principal component was that the jaw got wider in the positive PC-scores. Towards the negative PC-scores the jaw got pointier and longer. PC2, which separated the browsers, indicated variation in the angle between landmark 1 and 2. Towards the positive values the angle got obtuse, all in all the jaw turned blunter. The angle was changing from acute in the negative PC-scores to an obtuse angle in the far positive PC-scores (Supplementary Fig. 5 and 6). To see if there was a phylogenetic signal concerning morphological differentiation, we set the target classifier to “lake”, and found out that there was no clear separation in the dentary in case of the lake, so the phylogenetic signal was low (Fig. 6C). The corresponding CVA “niche” yielded highly similar and even more clear-cut results (Supplementary Fig. 3A).

3. Results

Figure 6. The PCAs from the dentary. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes. (LM= Lake Malawi; LT= Lake Tanganyika, LV= Lake Victoria).

3. Results

2 Articular

1 3

6 5 Figure 7. The articular from the left lateral side including the 6 homologous landmarks. The figure was modified from Albertson et al., 2001.

As in the dentary we also found a strong species-specific signal for the articular. The individuals belonging to one species were all clustered together except for one individual of Petrotilapia sp. “thick bars” (Petrotilapia sp. ‘Chitimba’) which did not cluster with the other Petrotilapia species. It was placed near the omnivorous species Limnotilapia dardennii and the browser Simochromis diagramma close to the center of the graph (Fig. 8A). By setting the target classifier to “niche” a stronger distinction signal was found (Fig. 8B). For the articular most of the variation could be explained by PC 1 as it accounted for 69.547% of the whole variation. In the direction towards positive values the articular became pointier at landmark 2. Towards the opposite direction the articular got wider and blunt. The carnivorous specimens were placed at the negative PC-scores regarding PC 1 whereas the grazer species were placed on the positive side. The omnivores were positioned in the center, so the different trophic niches could be distinguished well from each other. Thus, much information could be obtained from the first principal component. The distance between the retroarticular process and the rostral process of the coulter area got smaller in the grazers. Less information could be observed in PC 2. The distance between landmark 2 and 3 got wider in the more positive values (Supplementary Fig. 7 and 8). PC 1 plotted versus PC 3 showed one line containing all specimens (not shown). When setting the target classifier to “lake”, no specific signal was observed (Fig. 8C). So, separating the specimens by lake was not possible in this case. The corresponding CVA “niche” yielded highly similar and even more clear-cut results (Supplementary Fig. 3B).

3. Results

Figure 8. The PCAs from the articular. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

3. Results

Maxilla 2

5 3

6 4

Figure 9. The maxilla from the left lateral side including the six homologous landmarks. The figure was modified from Albertson et al., 2001.

In the next bone of the oral jaw unit, the maxilla, some species were well distinguishable from the others, such as Gnathochromis pfefferi on the right hand side and the two Labeotropheus species on the lower left hand side (Fig. 10A). Setting the target classifier to “niche” revealed that the ecological signal was low. The first principal component represented only 33.259 % of the total variance (Fig. 10B). In PC 1 the width between landmark 2 and 5, thus the Maxilla, turned slightly broader in the negative PC- scores. PC 2 showed the rotation of the bone (Supplementary Fig. 9 and 10). PC3 - the distance between landmark 4 and 5 grew longer in the negative PC-scores. Towards the positive PC-scores the distance between landmark 3 and 4 became less pointy and broader on this side of the graph (Supplementary Fig. 21). That means the browsers had a wider and more rotated maxilla, whereas the carnivore specimens had a pointier and narrower maxilla. Thus, not much variation could be obtained from the first principal component and at least two principal components were needed to distinguish the different trophic niches from each other. Setting the target classifier to “lake” revealed that no clear phylogenetic signal was found for this bone. However, the Lake Victoria species were in the center, in-between the two other lakes (Fig. 10C). In this bone also the CVA with the classifier set to “niche” was needed to separate the trophic niches well from each other. As expected, the CVA showed the same outcome as the PCA, and even more clearly as shown in the graph. The carnivore

3. Results species had a narrower and pointier maxilla, whereas the browser had a wider and more rotated maxilla (Supplementary Fig. 3C).

Figure 10. The PCAs from the maxilla. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

3. Results 2

Premaxilla

1 5 4

Figure 11. The Premaxilla from the left lateral side including the 5 homologous landmarks. The figure was modified from Albertson et al. 2001.

Looking at the premaxilla with the classifier set to “species”, the last bone belonging to the oral module in our data set, some species were well distinguishable and clustered together, for example Gnathochromis pfefferi on the upper left side of the graph. Near this cluster two other carnivore species Ctenochromis horei and Haplochromis thereuterion could be found (Fig. 12A). Setting the target classifier to “niche” the niches could be separated well from each other. The first principal component accounted for 79.743% of the total variation among the premaxilla. It turned out to be a very informative bone (Fig. 12B). There was also a fractionation of the carnivore species possible: the Gnathochromis pfefferi individuals were at the far left along the x-axis, while Ctenochromis horei and Neochromis omnicaeruleus clustered together. Sciaenochromis fryeri as carnivore species was placed in the center of the omnivorous and the browser species for this bone. All in all, the carnivore species were accumulated on one side, followed by the omnivorous species, and by the algae eaters at the other side. Setting the target classifier to “lake” a lake-specific signal was detected (Fig. 12C). The Lake Malawi species clustered as separate group, whereas the Lake Victoria and the Tanganyika specimens were clustered together. The PC 1 axis revealed that the premaxilla turned squatter in the grazer than the carnivore specimens, for example G. pfefferi which had a really long ascending arm compared to the grazers which were on the positive side of the PC 1. The angle of the ascending arm changed in PC 1 from acute in the carnivore species (G. pfefferi, C. horei) to less acute in the grazer species. Also the last tooth changed from being further away from the caudal end of the premaxilla in carnivore species to being near the end of the premaxilla in the grazers. The second principal component described that the jaw became more squashed/got shorter on the ventral side (-the tooth bearing area) as PC scores became more positive (Supplementary Fig. 11 and 12).

3. Results

Figure 12. The PCAs from the premaxilla. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

Pharyngeal jaw module

The second set of bones belongs to the pharyngeal jaw functional module. The pharyngeal jaw module included two bones, i.e. the lower pharyngeal jaw (Fig. 13.) and the upper pharyngeal jaw (Fig. 15.). For convenience only we also included the last observed bone in this module the interopercle (Fig. 17.). It does not belong to one of the two modules

3. Results we defined. It is a part belonging to the suspensorium and represents one of the bones which are in between the jaw modules. The bones are immobile and the interopercle connects the two functional units. The lower pharyngeal jaw has been analyzed in several papers, in the context of object symmetry (Klingenberg et al. 2002) and food processing (e.g. Hellig et al., 2010).

Lower pharyngeal jaw 3

1 4

Figure 13. The lower pharyngeal jaw from the rostral view including the five homologous landmarks. The figure was modified from Elmer et al. 2010.

When setting the target classifier to “species” there was not a clear-cut distinction between the species possible, except for the carnivore species Gnathochromis pfefferi and Sciaenochromis fryeri, which could be separated really well from the other species (Fig. 14A). When changing the classifier to “niche” an ecological signal was found, in that the trophic niches were clearly separated from each other (Fig. 14B). However, it turned out that the degree of distinction was not so clear in the lower pharyngeal jaw compared to the oral jaws, but the niches were still distinguishable. In this bone we carried out two analyses with two sets of landmarks, to find out if there was a difference in the signal when including a specific landmark adding the three-dimensionality, or if the added signal was negligible. It became clear that there was not a big difference between the analysis without the 3D landmark and within this specific landmark number 2. By setting the target classifier to

3. Results

“lake” the following signal was found: Lake Victoria and Tanganyika were clustered together whereas the third lake showed up further away from the center of the graph (Fig. 14C). The differences in PC 1 were shown as the angle of the lower pharyngeal was changing from an obtuse angle in Gnathochromis pfefferi to an acute angle in Sciaenochromis fryeri. So, the right side of the graph (positive values) shared a shorter distance between LM 1 and LM 4. Overall the lower pharyngeal jaw became a little bit shorter in these species. In addition, the distance between LM 1 and 2 got wider in this direction of the PC 1 (analysis made without the third dimension - Supplementary Fig 2; 15 and 16). Changing the settings to include the third dimension, (including landmark 2), the most variable parameter was manifested in the width of the jaw - from a thin jaw (in Petrochromis famula) in the negative PC-scores towards a robust one at the positive values - representing for example Sciaenochromis fryeri. Looking at the changes going on in PC 2 without the third dimension - the distance between LM 2 and 3 got much wider. This means that the horn was getting more robust towards the positive x-axis. The same analysis for PC2 but including the 3D landmark yielded almost the same outcome: the variability was also manifested in the width of the horn with a wider one towards the positive values (Supplementary Fig. 13 and 14). One of the differences which could be detected by including and excluding landmark 2 was that variability in the first principal component decreased from 62% for PC 1 in the analysis without the 3D landmark to 48% of the total variance in the analysis including this specific point. The corresponding CVA setting the classifier to “Niche” yielded a separation of the carnivore and the algae eater with the omnivorous species in-between. It did not matter if landmark 2 was included or not (Supplementary Fig. 4 A and B). All in all, it was much harder to distinguish the niches compared to the oral jaws.

3. Results

Figure 14. The PCAs from the lower pharyngeal jaw. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

3. Results

Upper Pharyngeal jaw

A B 2 1 1

3 2 3

Figure 15. Left upper pharyngeal jaw without pharyngobranchial 2. A. dorsal view including three landmarks. B. caudal-medial view. The figure was modified from Barel et al. 1976.

As in the former bones, we also started the analysis by setting the target classifier to “species” (Fig. 16A). There was no species-specific signal found for the upper pharyngeal jaw, a separation of the species was not possible, as all species occupied nearly the same shape- space concerning the upper pharyngeal jaw. By changing the target classifier to “niche” there was the same outcome, also in that the niches could not be separated as well (Fig. 16B). All niches were really close to each other, and the overall variation from PC1 was really small for the upper pharyngeal jaw (38.704%). When setting the target classifier to “lake”, the following phylogenetic signal could be found: the Lake Tanganyika species clustered together, whereas the Malawi species were spread out over the first principal component axis (Fig. 16C). The biggest variation in the bone was found in the area of the upper pharyngeal jaw. This was reflected in the PC 1: in the positive PC-scores the jaw got smaller and thinner, the distance between LM1 and 2 was changing from wide in the positive PC- scores to narrower in the negative values. But overall, the jaw was getting more robust. The area of the whole jaw grew larger and got in the shape of a triangle in the negative PC- scores. The effects for PC 2 were hard to tell, the distance between landmark 2 and 3 was getting taller in the positive values (Supplementary Fig. 17 and 18). The corresponding CVA setting the classifier to “Niche” separated the different niches better (Supplementary Fig. 4C). Changing the classifier to “Lake” to produce a CVA to test for phylogenetic signal, there was a fractionation possible between the lakes- Lake Victoria on the left side of the graph, Malawi species in the middle close to the Tanganyika ones (Supplementary Fig. 22).

3. Results

Figure 16. The PCAs from the upper pharyngeal jaw .A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

3. Results

Connective bone - Interopercle

2 3 Figure 17. The representative bone from the bones between the two jaw modules the interopercle from the left lateral side including the three homologous landmarks. Modified from Barel et. al, 1976.

Concerning the interopercle, we found the following patterns: When setting the target classifier to “species”, we observed a separation between the different species, but less clearly than observed for other bones in this thesis (Fig. 18A). All of the species were spread out over the PC 1 axis, because there was high variation assigned to this PC-axis (PC 1 accounted for 75.863% of the total variance in this bone). On the other hand, the variation observed for the second principal component was low. By setting the target classifier to “niche” there was no ecological signal evident for this bone: the niches were hard to differentiate (Fig. 18B). All in all, the differences in the landmarks were not that high. The omnivorous species were placed close to the carnivore species. This phenomenon was only found for the interopercle. The variation in PC 1 showed that the interopercle got more robust and wider in the browsers and grazers (the algae eaters, symbolized in red and orange in the figures), whereas it became smaller and thinner between landmark 1 and landmark 2 in the negative PC-scores, towards the carnivores (Supplementary Fig. 19 and 20). Setting the target classifier to “lake”, there was a specific phylogenetic signal observed (Fig. 18C). The Lake Victoria species on the right hand side mixed up with the Malawi species. The Tanganyika species were widely spread out on the first principal component axis. The corresponding CVA setting the classifier to “Niche” showed a clearer separation between the carnivore species and the algae eaters including browsers and grazers (Supplementary Fig. 4D).

3. Results

Figure 18. The PCAs from the interopercle. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

4. Discussion

Landmarks

Landmark interpretation The definition and optimization of the landmarks was an iterative process, mainly by optimizing the positioning of the landmarks using a set of model individuals, to subsequently use the entire landmark-set on all study specimens. After this step we started the actual geometric morphometric analysis and firstly looked at sexual dimorphism found out that its effect was really small in the bones, in contrast to color where the males are usually more brightly colored than the females (Barel, Van Oijen, et al. 1977). In the analysis software TINA it was possible to set the landmarks precisely on the surface of the bones, the articular cavity, and the middle of the teeth. This was possible because of the ability to use three possible viewpoints, along each axis in the three-dimensional space. Some of the landmarks had to be set very close to each other, like for example landmark 4 (dentary) and landmark 1 (articular). Generally, interspecific variation exceeded intraspecific variation in the landmarks. Therefore, it turned out that there was a clear-cut ecological signal in most of the bones analyzed, with some exceptions in particular bones.

Semilandmarks are points set on the surface or edge of a structure to better trace the shape of this structure, like for example landmarks 4 and 7 on the dentary (Supplementary Table 3). They, however, do not have any functionality per se for the animal. Semilandmarks allow to evenly distribute points along smooth curves or surfaces which could otherwise not be traced adequately (Mitteroecker & Gunz 2009). In the analysis presented in this thesis we excluded the semilandmarks, because of program limitations, we could not slid them along the surface. A future analysis including additional and morphologically more divergent species might add these points to optimize the power of discrimination.

Oral jaw module

The four skeletal elements belonging to this module showed a clear-cut ecological and species-specific signal. Nearly all bones showed these two signals, with the exception of the

4. Discussion premaxilla, which only showed a clear-cut phylogenetic signal. It may be worthwhile to note that all bones in this module are moveable bones (Otten 1983).

Dentary As mentioned above, the dentary separated the species and niches really well from each other, the carnivores being positioned at the left-hand side, the grazers at the right-hand side, and in the middle the omnivores. Moreover, the browsers could be discriminated along the PC2 axis. Thus, there is a clear-cut ecological signal in this bone. In the first principal component -in the direction of the grazer - the dentary would get shorter and less pointy. Also, the distance between LM1 and 6 - the width of the jaw - turned out to be larger in the grazers. In the carnivore species one could see that the jaw was pointier. In carnivore species the position of the last tooth was further forward, more in rostral direction, whereas the grazer species had their teeth positioned more in caudal direction. The width of the jaw was smaller in the carnivore specimens. A larger and wider jaw was found in the grazer specimens. These species evolved this jaw-form because of taking big bites of algae from the rock surface. Anatomically, the lower jaw is not so elongated in the algae feeders than in the zooplankton feeders (Kassam et al. 2004).

The dentary and the articular form the lower jaw, but they do not develop as a unit (Albertson & Kocher 2001). The dentary is formed from the mid-brain neural crest cells- it is a so-called dermal bone. The articular forms from cartilage cells and later ossifies endochondrally (Köntges & Lumsden 1996). In terms of mechanical function, the dentary and the articular are parts of the so-called four-bar linkage. This is a mechanism where the lower jaw rotation is communicated to the rotating maxilla. The dentary is also rotated on the quadrate so that it can open effectively (Hulsey et al. 2006; Otten 1983; Westneat 1990; Anker 1974; Barel, Van der Meulen, et al. 1977).

Articular The articular carries significant information to separate the trophic niches and species well from each other. There is a clear ecological signal evident: all specimens belonging to the same ecological niche were clustered together. The carnivore species (green) clustered together. The omnivore fish were, as expected, in the center of the principal component graph. The single individual of Petrotilapia sp. “thick bars” analyzed, belonging to the grazer section, was not clustered as expected with the remaining grazers but was placed in the

4. Discussion middle of the browsers (colored in black). The reason could be that its mouth morphology is actually more browser-like or omnivorous, and not like a typical grazer. In fact, its mouth morphology is markedly different from the second Petrotilapia species analyzed. Another possible explanation could be that it is an outlier, as we did not have a second sample for comparison. In terms of shape space, the distance between LM 1 and 2 is getting shorter along the x-axis in the direction towards the carnivore species (negative values). In terms of anatomy, we found that the articular is getting more compressed in the carnivore species. On the other side of the bone there was a shift towards the back in landmark 2- this resulted in a pointier and bigger articular. All grazer species had a bigger articular. A shorter articular process indicates a faster jaw closing motion and greater angular rotation. A higher process means that there is a greater force transmission – a stronger biting force (Albertson & Kocher 2001; Otten 1983). The articular is originated from the mid- and hindbrain cranial neural crest - an endochondral originating structure (Köntges & Lumsden 1996; Albertson et al. 2003). Anatomically, the articular is a joint which articulates with the suspensorium (Albertson & Kocher 2001; Otten 1983). Carnivores were placed at the left hand side and the grazers on the right hand side of the PCA.

Maxilla Also in this bone, a clear-cut ecological signal was detected; the niches could be separated well from each other. All in all, the variation was not as great as observed in the other bones belonging to the oral jaw module. On the positive PC-scores the carnivore fish were placed, that indicated a maxilla which was smaller in-between LM 1 and LM 6. It turned out that the maxilla was thinner in these species compared to the browsers - they had a more robust maxilla. The omnivores were positioned, as expected, in the center between the two extreme feeding types. The grazers had a more rotated maxilla. The browser had a wider and longer palatinad wing (LM1 und 2). The maxilla is a moveable bone together with the premaxilla (Otten 1983), which evolved independently multiple times (McGee et al. 2016).

Premaxilla There was a clear ecological signal and a strong phylogenetic signal observed for this bone, in that the trophic niches, the species and the lakes could be separated. Interestingly, there was a strong phylogenetic signal in the Lake Malawi species, as they all clustered together. In terms of ecological distinction, the carnivore Sciaenochromis fryeri was placed among the

4. Discussion browsers and the omnivore fish not - as otherwise often observed - near the carnivores which were observed on the negative side of the graph. The grazers were positioned on the positive side of PC1. Overall, there was the highest amount of variation of all the bones analyzed, so it is the most informative bone. On the x-axis in direction of the positive values the premaxilla is getting smaller and more compressed (LM1 –LM2). Also, the last tooth moved further back near LM 4 in the grazer species. On the opposite side of the morphospace the last tooth of the premaxilla was moved forward - the jaw was also getting bigger. Thus, the proportion of the ascending to dentigorous arm was changing in this PC- axis, in that the premaxilla was getting pointier in the carnivores. Checking the corresponding CVA grouping by the delimiter "Niche" we saw a clear-cut separation between the different niches: right hand side the carnivores, left hand side the browsers and the grazers, and in-between the omnivorous species. Anatomically, there was a longer ascending arm observed in the carnivore species and vice versa a shorter ascending arm for the browsers. There was a correlation between the ascending arm (LM 2) and the maxillad spine (LM 3), which was also longer in the carnivore species. A shorter ascending arm means that there is higher force transmission during biting; this is accompanied with a less acute angle between dentigorous arm and ascending arm (Otten 1983).

4. Discussion

The pharyngeal jaw module

Figure 19. Principal component analysis from Muschick et. al. 2012. Representing the differences in the lower pharyngeal jaw over 71 species belonging to 14 tribes in Lake Tanganyika. The green part represents the Haplochromini ,specifically the Tropheini- the tribe we were looking at.

Comparison of our data to former studies

The paper from Muschick et al. (2012) was placed in a much wider phylogenetic context, across all the Lake Tanganyika lineages. The study included over 71 species from 14 tribes. They focused at the body shape as well as at the morphology of the lower pharyngeal jaw. They set 28 landmarks on the lower pharyngeal jaw (8 true landmarks, 20 semilandmarks). Concerning the position of the landmarks, they were very similar to ours. However, there were some differences, for example they looked at the whole lower pharyngeal jaw, whereas we only looked at the left side of this jaw because of object symmetry. Talking about Figure S1 of Muschick et al. (2012): In this graph (in this thesis Fig. 19) only the green part was important for us, as it concerned the Haplochromini, specifically the Tropheini. We analyzed 17 species belonging to this lineage, but we also included some species from Lakes Malawi and Victoria. Another difference to the Muschick paper was that we set three- dimensional landmarks, with the specific landmark 2 in our analysis. To make our analysis more comparable, we excluded this landmark for a parallel analysis. The outcome was that there was not a specific difference between the two analyses, with or without the third

4. Discussion dimension. We used 5 or 4 landmarks to define the shape of this jaw. The second bone in the pharyngeal jaw module represented the upper pharyngeal jaw which did not show a clear ecological signal in the PCA. Looking at the corresponding CVA a distinction of the species was possible. Anatomically, the pharyngeal jaws originated out of the 5th ceratobranchial arch (Schneider et al. 2014). It is formed by fusion of the right and left 5th gill arch and it is thus sutured. This bone is a good example of object symmetry (Klingenberg et al. 2002), so that we decided to set the landmarks on the left hemisphere only. The pharyngeal skeleton develops from cranial neural crest cells (Albertson et al. 2003). The upper pharyngeal jaw (pharyngobranchial 3, 4), as well as the lower pharyngeal jaw, are the two largest bony elements belonging to the branchial apparatus. The upper pharyngeal jaw is a lumpy element, unlike the lower pharyngeal jaw – this is more flat in dorso-ventral direction. The dentigerous area (tooth-bearing area) is flat in contrast to the opposite side which is more curled. The tooth-bearing face defines the relative horizontal plane in ventral direction. The lower pharyngeal jaw is typically heart-shaped, and it is the largest element in the branchial apparatus. The point of the heart lies anteriad. Some species from Lake Victoria, currently belonging to the genus Haplochromis, are ambivalent in their food preference (Barel et al. 1976). The upper pharyngeal jaw with all parts (pharyngobranchial two, three and four) articulates with the apophysis of the neurocranium – this is an unusual phenomenon in fishes. The lower pharyngeal jaw bone is a single bone (Barel, Van Oijen, et al. 1977). During food processing the type of force generation is diverse in cichlid fishes, and a strong indication of their trophic specialization. Its shape is very different, depending if the individual is specialized to crush hard-shelled prey or process small algae (Hulsey et al. 2006; Hulsey 2006).

The function of the pharyngeal jaws is to crush and process the food. They are functionally decoupled from the oral jaws and they are located immediately before the pharynx (Hellig et al. 2010; Hulsey et al. 2006; Schaeffer & Rosen 1961; Muschick et al. 2012; Liem 1973). This is why we treated them as a separate module. Cichlid fishes have pharyngeal jaws with unique morphological features allowing them to occupy different trophic niches (Gunter et al. 2013; Liem 1973). Small pharyngeal jaws are found in suction feeders, whereas robust and larger jaws are found in snail shell crushers (Barel 1983; Barel et al. 1989).

4. Discussion

It is important to note also that pharyngeal jaws can change their morphology during the lifetime of a fish, depending on the current diet. This was shown in a series of split-brood experiments (Meyer 1989; Gunter et al. 2013). Also, the tooth size can change during lifetime (Greenwood 1965). Our preliminary experiments also revealed that the pharyngeal apparatus shows differences between wild caught and aquarium-breed fish, again due to phenotypic plasticity. It was already shown by Greenwood that aquarium-raised specimens showed less massive pharyngeal jaws (Greenwood 1965). This could in part explain why we observed a less clear separation in the pharyngeal jaw module. Our fish were aquarium bred, so it could be that the pharyngeal jaw showed less variation than under natural conditions, as found by Greenwood (1965).

Connective bone –Interopercle

The interopercle was generally less informative but showed a somewhat stronger phylogenetic than an ecological signal. Most of the variation was found in PC1, the width of the interopercle was the changing factor. Browsers showed a wider interopercle in the front, carnivores a thinner one. Anatomically speaking, the interopercle is the connection between the two jaw modules and is shaped as a flat rectangular element on the medio-ventral side of the fish, belonging to the suspensorial apparatus (Barel et al. 1976). The interopercle has also a ligament connection to the lower oral jaw, the so-called primary jaw (Anker 1986). Overall, there also was great overlap among the representatives of the trophic niches in the CVA. To analyze this bone more in-depth, one might have to include a set of semilandmarks to better trace the shape of this bone with the two horns in its front.

Why using three dimensional landmarks instead of two dimensional ones? So far, most of the studies on cichlid fishes were carried out on the basis of two-dimensional geometric morphometrics. This is in part due to the difficulty to produce three-dimensional images, which became increasingly feasible by the recent advances in the field of computer tomography. In addition, a lot of computational power is needed, which also became increasingly possible with the development of faster core processors. Given the current state of the art, the main disadvantage of using 2D-landmarks is that there is a substantial loss of information. There is an unavoidable loss of information when projecting a 3D-object into a

4. Discussion

2D photograph. Some of the structures may be perpendicular to the objected plane- so this bone could be overseen (Cardini 2014; Roth 1993). Earlier studies had to analyze different 2D-views of a single bone to capture some 3D information. Another factor for the 3D landmarks is that the structures can be visualized and analyzed in the original positional context of the bones. There were already some studies conducted with the help of 3D morphometrics in different species. In humans, for example, anthropologists used this kind of morphometrics to quantify curved surfaces between the landmarks to study facial and cranial variation (e.g. Gunz & Mitteroecker 2013). In analysis on mouse skulls the use of 3D landmarks yielded a finer characterization of the mandible shape, especially when shape was compared with genetic factors (Navarro & Maga 2016). So far, only few studies on cichlid fishes have worked with 3D images produced by micro-CT based imaging, for example a study addressing bone volume and force transmission (Cooper et al. 2011; Hulsey et al. 2008). Our study is the first using 3D landmarks to examine 3D shape of the oral and the pharyngeal jaw module.

A very recent study was analyzing beak shapes in a functional evolutionary context for over 2000 bird species (over 97% of the extant genera). This study represented the full range of bill shapes in all living bird lineages. The authors calculated the evolutionary rates of all of the tribes in the avian-fauna (Cooney et al. 2017). Our study on cichlid mouth morphology had a relatively narrow phylogenetic background, going into the past for less than 10 Mio years.

Conclusion The two modules we were looking at showed different outcomes in connection with the ecological and the phylogenetic signals. On the one hand the oral jaws portrayed the ecological signal, whereas the pharyngeal jaws and the interopercle portrayed the phylogenetic signal. All of the homologous landmarks were found in all of the specimens. We feel it is quite remarkable how much ecologically and phylogenetically relevant information can be drawn from this functionally crucial section of the head morphology.

5. References

Albertson, R.C. & Kocher, T.D., 2001. Assessing morphological differences in an adaptive trait: a landmark-based morphometric approach. The Journal of experimental zoology, 289(6), pp.385– 403.

Albertson, R.C., Streelman, J.T. & Kocher, T.. D., 2003. Genetic Basis of Adaptive Shape Differences in the Cichlid Head. Journal of Heredity, 94(4), pp.291–301.

Anker, G.C., 1974. Morphology and kinetics of the head of the stickleback, Gasterosteus aculeatus. The Transactions of the Zoological Society of London, 32(5), pp.311–416.

Anker, G.C., 1986. The Morphology of joints and ligaments in a head of a generalized Haplochromis species: H. elegans Trewavas 1933 (Teleostei, Cichlidae). Netherlands Journal of Zoology, 36(4), pp.498–530.

Barel, C. et al., 1989. Constructional constraint and its ecomorphological implications. Acta morphologica Neerlando-Scandinavica, 27, pp.83–109.

Barel, C.D.., Witte, F. & Van Oijen, M.J.P., 1976. The shape of the skeletal elements in the head of a generalized Haplochromis species: H. elegans trewavas 1933 (Pisces, Cichlidae). Netherlands Journal of Zoology, 26(2), pp.163–265.

Barel, C.D.N., Van Oijen, M.J.P., et al., 1977. An introduction to the taxonomy and morphology of the haplochromine Cichlidae from Lake Victoria. Netherlands Journal of Zoology, 27(4), pp.333–389.

Barel, C.D.N., 1983. Towards a constructional morphology of cichlid fishes (Teleostei, Perciformes). Netherlands Journal of Zoology, 33(4), pp.357–424.

Barel, C.D.N., Van der Meulen, J.W. & Berkhoudt, H., 1977. Transmissionskoeffizient und Vierstangensystem als Funktionsparameter und Formmodel für mandibulare Depressionsapparate bei Teleostiern. Anat. Anz, 142, pp.21–31.

Bookstein, F.L., 1996. Biometrics, biomathematics and the morphometric synthesis. Bulletin of Mathematical biology, 58(2), pp.313–365.

Bookstein, F.L., 1991. Morphometric tools for landmark data,

Bromiley, P. a, Ragheb, H. & Thacker, N. a, 2012. The TINA Geometric Morphometrics Toolkit. Analysis, (2010).

Campbell, N.A. et al., 1981. the Geometry of Canonical Variate Analysis. Syst. Zool, 30(3), pp.268– 280.

Cardini, A., 2014. Missing the third dimension in geometric morphometrics: How to assess if 2D images really are a good proxy for 3D structures? Hystrix, 25(2), pp.73–81.

Clabaut, C. et al., 2007. Geometric morphometric analyses provide evidence for the adaptive character of the Tanganyikan cichlid fish radiations. Evolution, 61(3), pp.560–578. 43

5. References

Cohen, A.S. et al., 1997. New palaeogeographic and lake-level reconstructions of Lake Tanganyika: Implications for tectonic, climatic and biological evolution in a rift lake. Basin Research, 9(2), pp.107–132.

Cohen, A.S., Soreghan, M.J. & Scholz, C.A., 1993. Estimating the age of formation of lakes: an example from Lake Tanganyika, East African Rift system. Geology, 21(6), pp.511–514.

Cooney, C.R. et al., 2017. Mega-evolutionary dynamics of the adaptive radiation of birds. Nature Publishing Group.

Cooper, W.J. et al., 2011. Functional and Genetic Integration in the Skulls of Lake Malawi Cichlids. Evolutionary Biology, 38(3), pp.316–334.

Delvaux, D., 1995. Age of Lake Malawi (Nyasa) and water level fluctuations. Mus. roy. Afrr. centr., Tervuren (Belg.), Dept. Geol. Min., Rapp. ann. 1993 & 1994, 108, pp.99–108.

Dryden, I. & Mardia, K. V., 1998. Statistical shape Analysis. Wiley, Chichester.

Frederich, B., Pilet, A., et al., 2008. Comparative Trophic Morphology in Eight Species of Damselfishes (Pomacentridae). Journal of Morphology, 269, pp.175–188.

Frederich, B., Adriaens, D. & Vandewalle, P., 2008. Ontogenetic shape changes in Pomacentridae (Teleostei, Perciformes) and their relationships with feeding strategies: A geometric morphometric approach. Biological Journal of the Linnean Society, 95(1), pp.92–105.

Fryer, G. & Iles, T., 1972. The cichlid fishes of the Great Lakes of Africa: Their Biology and Evolution. Oliver and Boyd, Edinburgh.

Fujimura, K. & Okada, N., 2008. Shaping of the lower jaw bone during growth of Nile tilapia Oreochromis niloticus and a Lake Victoria cichlid Haplochromis chilotes: a geometric morphometric approach. Development, growth & differentiation, 50(8), pp.653–63.

Gasse, F. et al., 1989. Water-level fluctuations of Lake Tanganyika in phase with the oceanic changes during the last glaciation and deglaciation. Nature, 342, pp.57–59.

Goodall, C., 1991. Procrustes Methods in the Statistical Analysis of Shape. Journal of the Royal Statistical Society, 53(2), pp.285–339.

Greenwood, P.H., 1984. African cichlid fishes and evolutionary theories. In A. A. Echelle & I. Kornfield (Eds) Evolution of Fish Species Flocks. Orono, USA: University of Maine at Orono Press., pp.141– 154.

Greenwood, P.H., 1965. Environmental effects on the pharyngeal mill of a cichlid fish, Astatoreochromis alluandi, and their taxonomic implications. Proc.Linn.Soc.London, 176, pp.1– 10.

Greenwood, P.H., 1979. Towards a phyletic classification of the “genus” Haplochromis (Pisces, Cichlidae) and related taxa. Bull. Br. Mus.Nat. Hist. (Zool.), 35, pp.265–322.

Greenwood, P.H., 1980. Towards a phyletic classification of the “genus” Haplochromis (Pisces, Cichlidae) and related taxa. Part 2. Bull. Br. Mus.Nat. Hist. (Zool.), 39, pp.1–99.

5. References

Gunter, H.M. et al., 2013. Shaping development through mechanical strain: The transcriptional basis of diet-induced phenotypic plasticity in a cichlid fish. Molecular Ecology, 22(17), pp.4516–4531.

Gunz, P. & Mitteroecker, P., 2013. Semilandmarks: A method for quantifying curves and surfaces. Hystrix, 24(1).

Hellig, C.J. et al., 2010. Allometric shape change of the lower pharyngeal jaw correlates with a dietary shift to piscivory in a cichlid fish. Naturwissenschaften, 97(7), pp.663–672.

Herler, J. et al., 2010. Sexual dimorphism and population divergence in the Lake Tanganyika cichlid fish genus Tropheus. Frontiers in Zoology, 7(4).

Hulsey, C.D. et al., 2008. Convergence in a mechanically complex phenotype: Detecting structural adaptations for crushing in cichlid fish. Evolution, 62(7), pp.1587–1599.

Hulsey, C.D., 2006. Function of a key morphological innovation: fusion of the cichlid pharyngeal jaw. Proceedings. Biological sciences / The Royal Society, 273(1587), pp.669–675.

Hulsey, C.D., Garacía de León, F.J. & Rodiles-Hernández, R., 2006. Micro- and Macroevolutionary Decoupling of Cichlid Jaws : a Test of Liem ’s Key Innovation Hypothesis. , 60(10), pp.2096– 2109.

Johnson, T.C. et al., 1996. Late Pleistocene Desiccation of Lake Victoria and Rapid Evolution of Cichlid Fishes. Science, 273(5278), pp.1091–1093.

Kassam, D. et al., 2006. Nuclear markers reveal that inter-lake cichlids’ similar morphologies do not reflect similar genealogy. Molecular Phylogenetics and Evolution, 40(2), pp.383–388.

Kassam, D., Yamaoka, K. & Adams, D., 2004. Functional significance of variation in trophic morphology within feeding microhabitat-differentiated cichlid species in Lake Malawi. Animal Biology, 54(1), pp.77–90.

Kerschbaumer, M., Mitteroecker, P. & Sturmbauer, C., 2014. Evolution of body shape in sympatric versus non-sympatric Tropheus populations of Lake Tanganyika. Heredity, 112(2), pp.89–98.

Klingenberg, C.P., 2011. MorphoJ: An integrated software package for geometric morphometrics. Molecular Ecology Resources, 11(2), pp.353–357.

Klingenberg, C.P., 2016. Size, shape, and form: concepts of allometry in geometric morphometrics. Development Genes and Evolution, 226(3), pp.113–137.

Klingenberg, C.P., Barluenga, M. & Meyer, A., 2002. Shape Analysis of Symmetric Structures : Quantifying Variation among Individuals and Asymmetry. , 56(10), pp.1909–1920.

Koblmüller, S. et al., 2008. Age and spread of the haplochromine cichlid fishes in Africa. Molecular Phylogenetics and Evolution, 49(1), pp.153–169.

Kocher, T.D. et al., 1993. Similar Morphologies of Cichlid Fish in Lakes Tanganyika and Malawi Are Due to Convergence. Molecular Phylogenetics and Evolution, 2(2), pp.158–165.

Köntges, G. & Lumsden, A., 1996. Rhombencephalic neural crest segmentation is preserved throughout craniofacial ontogeny. Development (Cambridge, England), 122(10), pp.3229–42. 45

5. References

Lezzar, K.E. et al., 1996. New seismic stratigraphy and late tertiary history of the North Tanganyika basin, East African rift system, deduced from multichannel and high-piston core evidence. Basin Research, 8(1 996), pp.1–28.

Liem, K., 1973. Evolutionary strategies and morphological innovations: cichlid pharyngeal jaws. Syst. Zool, 22, pp.425–441.

Mayr, E., 1984. Evolution of fish species flocks: a commentary. University of Maine :Orono Press, pp.3–12.

McGee, M.D. et al., 2016. Replicated divergence in cichlid radiations mirrors a major vertebrate innovation. Proc R Soc B, 283(1822), p.20151413-.

Meyer, A., 1989. Costs and benefits of morphological specialization: feeding performance in the trophically polymorphic cichlid fish, Cichlasoma citrinellum. Oecologia, 80, pp.431–436.

Meyer, A. et al., 1990. Monophyletic origin of Lake Victoria cichlid fishes suggested by mitochondrial DNA sequences. Letters to Nature, 347, pp.550–553..

Mitteroecker, P. & Gunz, P., 2009. Advances in Geometric morphometrics. Evolutionary Biology, 36(2), pp.235–247.

Muschick, M., Indermaur, A. & Salzburger, W., 2012. Convergent evolution within an adaptive radiation of cichlid fishes. Current biology : CB, 22(24), pp.2362–8.

Nagl, S. et al., 2000. The origin and age of haplochromine fishes in Lake Victoria, East Africa. Proceedings of The Royal Society London B, 267(February), pp.1049–1061.

Navarro, N. & Maga, A.M., 2016. Does 3D Phenotyping Yield Substantial Insights in the Genetics of the Mouse Mandible Shape? G3 (Bethesda, Md.), 6(5), pp.1153–63..

Otten, E., 1983. The jaw mechanism during growth of a geralized Haplochromis species: H. elegans Trewavas 1933 (Pisces, Chichlidae). Netherlands Journal of Zoology, 33(1), pp.55–98.

Le Pabic, P., Cooper, W.J., Schilling, T.F., 2016. Developmental basis of phenotypic integration in two Lake Malawi cichlids. EvoDevo, 7(3), pp.1–26.

Poll, M., 1986. Classification des Cichlidae du lac Tanganika: tribus, genres et espéces. T. XLV – Fascicule. 2nd edition.

Postl, L. et al., 2008. Geometric morphometrics applied to viscerocranial bones in three populations of the Lake Tanganyika cichlid fish Tropheusmoorii. Journal of Zoological Systematics and Evolutionary Research, 46(3), pp.240–248.

Roth, V.L., 1993. On three-dimensional morphometrics, and on the identification of landmark points. In: Marcus LF,Bello E, Garcia-Valdecasas A, editors. Contributions to morphometrics. Madrid: Museo Nacional de CienciasNaturales, pp.41–61.

Salzburger, W. et al., 2005. Out of Tanganyika: genesis, explosive speciation, key-innovations and phylogeography of the haplochromine cichlid fishes. BMC evolutionary biology, 5(1983), p.17.

5. References

Schaeffer, B. & Rosen, D.E., 1961. Major Adaptive Levels in the Evolution of the Actinopterygian Feeding Mechanism Author. , 1(2), pp.187–204.

Schneider, R.F. et al., 2014. Regulatory gene networks that shape the development of adaptive phenotypic plasticity in a cichlid fish. Molecular Ecology, pp.4511–4526.

Scholz, C.A. & Rosendahl, B.R., 1988. Low Lake Stands in Lakes Malawi and Tanganyika , East Africa , Delineated with Multifold Seismic Data. Science, 240(4859), pp.1645–1648.

Schunke, A.C. et al., 2012. TINA manual landmarking tool: software for the precise digitization of 3D landmarks. Frontiers in zoology, 9(1), p.6..

Sturmbauer, C. et al., 2001. Lake level fluctuations synchronize genetic divergences of cichlid fishes in African lakes. Molecular Biology and Evolution, 18(2), pp.144–154.

Sturmbauer, C., Husemann, M. & Danley, P., 2011. Explosive speciation and adaptive radiation of East African cichlid fishes. In: Zachos FE, Habel JC, editors. Biodiversity Hotspots: Distribution and Protection of Conservation Priority Areas. Germany: Springer-Verlag Berlin Heidelberg, pp.333–362.

Sugawara, T. et al., 2005. Parallelism of amino acid changes at the RH1 affecting spectral sensitivity among deep-water cichlids from Lakes Tanganyika and Malawi. Proceedings of the National Academy of Sciences of the United States of America, 102(15), pp.5448–5453.

Tiercelin, J. & Mondeguer, A., 1991. The geology of the Tanganyika trough. Oxford University Press, London, (In Coulter, G. W. (ed.), Lake Tanganyika and its Life), pp.7–48.

Turner, G.F., 2007. Adaptive radiation of cichlid fish. Current Biology, 17(19), pp.827–831.

Turner, G.F. et al., 2001. How many species of cichlids are there in African lakes? Molecular Ecology, 10, pp.793–806.

Verheyen, E. et al., 2003. Origin of the superflock of cichlid fishes from Lake Victoria, east Africa. Science, 300, pp.325–330.

Wanek, K.A. & Sturmbauer, C., 2015. Form, function and phylogeny: comparative morphometrics of Lake Tanganyika’s cichlid tribe Tropheini. Zoologica Scripta, 44(4), pp.362–373..

Westneat, M.W., 1990. Feeding Mechanics of Teleost Fishes (Labridae; Perciformes): A Test of Four- Bar Linkage Models. Journal of Morphology, 205(3), pp.269–295.

Wunder, M., 2016. Geometric morphometric analysis of adaptation to the pond environment in Tropheus moorii. Master-thesis.

Zelditch, M.L. et al., 2004. Geometric Morphometrics for Biologists: A Primer. Elsevier/ Academic Press.

5. References

6. Supplementary Information

Outlook and possible optimizations for the future:

Better standardize the jaw position, and load in not all of the pictures of a stack, so that one can set the landmarks more precisely in the pharyngeal jaw. Use only the ventral part of the fish- image, which I think is really hard because the slices the micro-CT produces are slices along the dorsal-ventral axis. Another idea is to cut the TIFF-Files in a second step, and try to catch the pharyngeal jaws better. For the upper pharyngeal jaw one has to separate the ventral side to view the dorsal part of the fish. Thus, one has to find the horizontal axis of the fish and try to exclude the dorsal (or the ventral part) of the fish, to gain a better view at the pharyngeal jaw.

Better standardize the measurements, so that the plane points in TINA can be used, to get all the axes in the right plane. For my project some error might have been included if one looks at the bone surface and there is a bias because of the angle the pictures have been taken. Also, an idea would be not to look at the left side of the fish only, but also on the right side. Then one could test for bilateral symmetry, which we took for granted.

It may be better not to convert the raw data into tiff format. The dicom format has more information in the background, which gets lost by converting the pictures from one format to the other. Maybe then one does not have to calculate the scale factor by hand.

Somewhat older juvenile fish and adults should be analyzed in further studies, to find out if our findings also hold for adults. However, the oldest and fully grown fish (G) should be excluded – so that the class of adults is better standardized. The microCT should also become faster- to get more juvenile fish analyzed in a shorter time period.

6. Supplementary

Supplementary Table 1. All 136 specimens the two red underlined specimens were excluded later because of missing pictures or bad pictures. From 138 fish 136 were included in all the analyses we made. The colour shows the niches the specimens belong to. Blue = carnivore; brownish = omnivore; light green = browser and darker green = grazer. The scalefactor was included by hand into the masterfile for MorphoJ.

Species Code Sex Count Age Voxelsize [µm] Scalefactor Lake Niche Tribe Ctenochromis horei CtHo_A_F_01 F 1 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_02 F 2 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_03 F 3 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_04 F 4 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_05 F 5 A 15 30 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_06 F 6 A 15 30 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_07 F 7 A 15 30 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_08 F 8 A 15 30 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_F_09 F 9 A 15 30 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_01 M 1 G 18 54 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_01 M 1 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_02 M 2 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_03 M 3 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_04 M 4 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_05 M 5 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_06 M 6 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_07 M 7 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_08 M 8 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_09 M 9 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_10 M 10 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_A_M_11 M 11 A 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_02 M 2 G 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_03 M 3 G 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_04 M 4 G 15 45 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_05 M 5 G 18 54 Tanganyika carnivore Haplochomini

6. Supplementary

Ctenochromis horei CtHo_G_M_06 M 6 G 18 54 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_07 M 7 G 18 54 Tanganyika carnivore Haplochomini Ctenochromis horei CtHo_G_M_08 M 8 G 18 54 Tanganyika carnivore Haplochomini Cynotilapia afra CyAf_G_F_01 F 1 G 15 45 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_F_01 F 1 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_F_02 F 2 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_F_03 F 3 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_F_04 F 4 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_G_M_01 M 1 G 15 45 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_01 M 1 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_02 M 2 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_G_M_02 M 2 G 15 45 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_03 M 3 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_04 M 4 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_05 M 5 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_06 M 6 A 15 45 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_07 M 7 A 15 30 Malawi omnivore Haplochomini Cynotilapia afra CyAf_A_M_08 M 8 A 15 30 Malawi omnivore Haplochomini Gnathochromis pfefferi GnPf_G_F_01 F 1 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_F_02 F 2 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_F_03 F 3 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_F_04 F 4 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_F_05 F 5 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_01 M 1 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_02 M 2 G 18 54 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_03 M 3 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_04 M 4 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_05 M 5 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_06 M 6 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_07 M 7 G 15 45 Tanganyika carnivore Tropheini

6. Supplementary

Gnathochromis pfefferi GnPf_G_M_08 M 8 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_09 M 9 G 15 45 Tanganyika carnivore Tropheini Gnathochromis pfefferi GnPf_G_M_10 M 10 G 15 45 Tanganyika carnivore Tropheini Haplochromis thereuterion HaTh_A_F_01 F 1 A 15 45 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_02 F 2 A 15 45 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_03 F 3 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_04 F 4 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_05 F 5 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_06 F 6 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_07 F 7 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_08 F 8 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_09 F 9 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_10 F 10 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_F_11 F 11 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_M_01 M 1 A 10 30 Victoria carnivore Haplochomini Haplochromis thereuterion HaTh_A_M_02 M 2 A 10 30 Victoria carnivore Haplochomini Labeotropheus fuelleborni LaFu_G_F_01 F 1 G 15 45 Malawi browser Haplochomini Labeotropheus fuelleborni LaFu_G_F_02 F 2 G 18 54 Malawi browser Haplochomini Labeotropheus fuelleborni LaFu_G_M_01 M 1 G 18 54 Malawi browser Haplochomini Labeotropheus fuelleborni LaFu_G_M_02 M 2 G 18 54 Malawi browser Haplochomini Labeotropheus fuelleborni LaFu_G_M_03 M 3 G 15 45 Malawi browser Haplochomini Labeotropheus trewavasae LaTr_A_M_01 M 1 A 15 30 Malawi browser Haplochomini Labeotropheus trewavasae LaTr_A_M_02 M 2 A 15 30 Malawi browser Haplochomini Labeotropheus trewavasae LaTr_A_M_03 M 3 A 15 30 Malawi browser Haplochomini Limnotilapia dardenii LiDa_G_F_01 F 1 G 0 Tanganyika omnivore Tropheini Limnotilapia dardenii LiDa_G_F_02 F 2 G 18 54 Tanganyika omnivore Tropheini Limnotilapia dardenii LiDa_G_M_01 M 1 G 18 54 Tanganyika omnivore Tropheini Limnotilapia dardenii LiDa_G_M_02 M 2 G 18 54 Tanganyika omnivore Tropheini Limnotilapia dardenii LiDa_G_M_03 M 3 G 18 54 Tanganyika omnivore Tropheini Limnotilapia dardenii LiDa_G_M_04 M 4 G 18 54 Tanganyika omnivore Tropheini

6. Supplementary

Metriaclima zebra MaZe_A_F_01 F 1 A 15 45 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_01 M 1 A 15 30 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_02 M 2 A 15 30 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_03 M 3 A 15 45 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_04 M 4 A 15 45 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_05 M 5 A 15 45 Malawi omnivore Haplochomini Metriaclima zebra MaZe_A_M_06 M 6 A 15 45 Malawi omnivore Haplochomini Neochromis omnicaeruleus NeOm_A_F_01 F 1 A 15 45 Victoria browser Haplochomini Neochromis omnicaeruleus NeOm_A_F_02 F 2 A 10 30 Victoria browser Haplochomini Neochromis omnicaeruleus NeOm_A_F_03 F 3 A 10 30 Victoria browser Haplochomini Neochromis omnicaeruleus NeOm_A_M_01 M 1 A 15 45 Victoria browser Haplochomini Neochromis omnicaeruleus NeOm_A_M_02 M 2 A 15 30 Victoria browser Haplochomini Sciaenochromis fryeri ScFr_G_F_01 F 1 G 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_G_F_02 F 2 G 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_F_01 F 1 A 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_01 M 1 A 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_G_M_01 M 1 G 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_02 M 2 A 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_G_M_02 M 2 G 18 54 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_03 M 3 A 15 45 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_04 M 4 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_05 M 5 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_06 M 6 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_07 M 7 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_08 M 8 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_09 M 9 A 10 30 Malawi carnivore Haplochomini Sciaenochromis fryeri ScFr_A_M_10 M 10 A 18 54 Malawi carnivore Haplochomini Simochromis diagramma SiDi_A_F_01 F 1 A 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_01 F 1 G 15 60 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_02 F 2 G 18 54 Tanganyika browser Tropheini

6. Supplementary

Simochromis diagramma SiDi_G_F_03 F 3 G 18 54 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_04 F 4 G 18 54 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_05 F 5 G 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_06 F 6 G 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_07 F 7 G 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_08 F 8 G 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_M_01 M 1 G 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_A_M_01 M 1 A 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_A_M_02 M 2 A 15 45 Tanganyika browser Tropheini Simochromis diagramma SiDi_G_F_09 F 9 G 18 54 Tanganyika browser Tropheini Tropheus moorii TrMo_G_F_01 F 1 G 18 36 Tanganyika browser Tropheini Tropheus moorii TrMo_G_F_02 F 2 G 18 54 Tanganyika browser Tropheini Tropheus moorii TrMo_G_F_03 F 3 G 18 54 Tanganyika browser Tropheini Tropheus moorii TrMo_G_M_01 M 1 G 15 45 Tanganyika browser Tropheini Petrochromis trewavasae PeTr_A_M_01 M 1 A 18 54 Tanganyika grazer Tropheini Petrochromis trewavasae PeTr_A_F_01 F 1 A 15 45 Tanganyika grazer Tropheini Petrochromis trewavasae PeTr_G_M_01 M 1 G 15 45 Tanganyika grazer Tropheini Petrotilapia sp. "yellow chin Chewere" PeYc_A_M_01 M 1 A 18 54 Malawi grazer Haplochromini Petrotilapia sp. "thick bars" PeTb_A_F_01 F 1 A 18 54 Malawi grazer Haplochromini Petrochromis famula PeFa_G_M_01 M 1 G 18 54 Tanganyika grazer Tropheini Petrochromis famula PeFa_A_M_01 M 1 A 15 45 Tanganyika grazer Tropheini Petrochromis famula PeFa_A_F_01 F 1 A 18 54 Tanganyika grazer Tropheini Petrochromis polyodon PePo_G_M_01 M 1 G 18 54 Tanganyika grazer Tropheini

6. Supplementary

Supplementary table 3. Landmarks in the raw version-including all of the semilandmarks and the tooth- belonging points. All the italic characters were excluded for the analysis.

Element Landmark Description Dentray 1 rostral tip of the dentary 2 last tooth of the dentary 3 dorsal tip of the coronoid (dentary process) 4 semilendmark between LM 3 and 5 5 surface dentary immediately rostral to landmark 9 6 posterior end of the dentary 7 semilandmark between 6 and 8 8 ventral-most point of the dentary

Articular 9 tip of the rostral process of the articular 10 dorsal tip of the primorial (Articular) process 11 dorsal process of the suspensoriad articulation facet 12 postarticulation process (of the suspensoriad articulation facet) 13 retroarticular process 14 rostral process of the coulter area

Maxilla 15 m edial process of the palatinad wing 16 lateral process of the palatinad wing 17 posterior end of the maxilla 18 shank process 19 lateral process of the pramaxillad wing 20 medial process of the premaxillad wing 21 meet of the maxilla and the premaxilla (hole) semilandmark

Premaxilla 22 r ostral-most point of the dentigerous arm 23 dorsal process of the ascending spine 24 dorsal process of the maxillad spine(bumpy) 25 semilandmark between 24 und 26 26 semilandmark between 25 und 27 27 semilandmark between 26 und 28 28 Caudal process of the dentigerous arm 29 Last tooth of the premaxilla

Lower pharyngeal jaw 30 caudal end of the median suture 31 depth of the lower pharyngeal jaw 32 caudal-dorsal tip of the left horn 33 caudal-ventral tip of the horn 34 semilandmark between 34 und 36 35 rostral edge of dentigerous area 36 height of the first tooth of the lower pharyngeal jaw 37 left side where the tooth meet the lower pharyngeal jaw 38 right side where the tooth meet the lower pharyngeal jaw

Upper pharyngeal jaw 39 caudal edge of the neurocraniad articulation facet (neur.d art.fct)

6. Supplementary

40 dorsal-most tip of the caudal-lateral eminence (cl ll-emin.) 41 caudal ventral tip of the upper pharyngeal jaw 42 articulation of the pharyngeobranchial 3 with pharyngeobranchial 2 43 height of the first tooth on the upper pharyngeal jaw 44 left side where the tooth meet the upper pharyngeal jaw 45 right side where the tooth meet the upper pharyngeal jaw

Interopercle 46 dorsal -rostral point of the interopercle 47 ventral-rostral point of the interopercle 48 caudal- ventral point of the interopercle

6. Supplementary

Ctenochromis horei Gnathochromis pfefferi Sciaenochromis fryeri

Haplochromis thereuterion Limnotilapia dardennii Cynotilapia afra

Maylandia zebra Simmochromis diagramma Tropheus moorii

Labeotropheus fuelleborni Labeotropheus trewavasae Neochromis omnicaeruleus

6. Supplementary

Petrochromis trewavasae Petrochromis famula Petrochromis polyodon

Petrotilapia sp. “yellow chin Petrotilapia sp. “thick bars” Chewere”

Supplementary Figure 1. 3D-scans from all of the species we used for this study represented by one of the individuals per species.

6. Supplementary

Supplementary Figure 2. The PCAs from the lower pharyngeal jaw without the specific landmark 2 so without the third dimension. A. classifier "SPECIES", B. classifier "NICHE", C. classifier "LAKE". The analysis were made over all the 136 specimens belonging to 17 species in the Three East African Lakes (LM= Lake Malawi; LT= Lake Tanganyika; LV= Lake Victoria).

6. Supplementary

Supplementary Figure 3. CVA- graphs with the target classifier “NICHE” from the oral jaw module. A. dentary, B. articular, C. maxilla and D. premaxilla.

6. Supplementary

Supplementary Figure 4. CVA- graphs with the target classifier “NICHE” from the pharyngeal jaw module plus interopercle. A. lower pharyngeal jaw including specific landmark, B. lower pharyngeal jaw without LM2, C. upper pharyngeal jaw and D. the connective bone - interopercle.

6. Supplementary

6 6 5 5

1 1

4 4

2 2

3 3

PC1 PC1

5 5 3 3

2 2

4 4

6 6

1 1

PC1 PC1

5 5 3 3

2 2

4 4

6 6

1 1

PC1 PC1

Supplementary Figure 5. Wireframes of the dentary: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.3 and -0.3. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6 5

4 1

2 3

PC2 3

PC2

5 3

2 4

1 PC2

PC2

5 3

2 4

6 PC2 1

PC2

Supplementary Figure 6. Wireframes of the dentary: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.3 and -0.3. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

6 5

3 1 3 1

2 PC1 PC1

4 4

3 3 5 5

2 2

6 6

1 1

PC1 PC1

3 5 4

3 5

2 6

PC1

Supplementary Figure 7. Wireframes of the articular: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.3 and -0.3. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6 5 6

3 1 3 1

PC2 2

PC2

4 4

5 3 5 3

2 2 6 6

1 1 PC2 PC2

4 4

3 5 3 5 2 2

6 6

1 1 PC2 PC2

Supplementary Figure 8. Wireframes of the articular: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.3 and -0.3 First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

4 4

6 6

3 3 5 5

1 1 2 2

PC1 PC1

3 3

4 4

2 2

6 6 5 5

1 1

PC1 PC1

3 3

4 4

2 2

6 6 5 5

1 1

PC1 PC1

Supplementary Figure 9. Wireframes of the maxilla: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

4 4

6 6

3 3 5 5

1 1 2 2

PC2 PC2

3 3

4 4

2 2

6 6 5 5

1 1

PC2 PC2

3 3

4 4

2 2

6 6 5 5

1 1

PC2 PC2

Supplementary Figure 10. Wireframes of the maxilla: on the left side the positive side of the principal componentaxis 2 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

4 4

5 5

1 1

3 3

2 2

PC1 PC1

2 2

4 4

3 3

5 5

1 1

PC1 PC1

2 2

4 4

3 3

5 5

1 1

PC1 PC1

Supplementary Figure 11. Wireframes of the premaxilla: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.3 and -0.3. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

4 4

5 5

1 1

3 3

2 2

PC2 PC2

2 2

4 4

3 3

5 5

1 1

PC2 PC2

2 2

4 4

3 3

5 5

1 1 PC2 PC2

Supplementary Figure 12. Wireframes of the premaxilla: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

5 5

2 2

1 1

4 4 3 3

PC1 PC1

3 3 4 4

1 1

2 2

5 5

PC1 PC1

3 3 4 4

1 1

2 2

5 5

PC1 PC1 .

Supplementary Figure 13. Wireframes of the lower pharyngeal jaw including the third dimension: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.2 and -0.2 First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

5 5

2 2

1 1

4 4 3 3

PC2 PC2

3 3 4 4

1 1

2 2

5 5

PC2 PC2

3 3 4 4

1 1

2 2

5 5

PC2 PC2

Supplementary Figure 14. Wireframes of the lower pharyngeal jaw including the third dimension: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.2 and -0.2 First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

4 4

1 1

3 3 2 2

PC1 PC1

2 2 3 3

1 1

4 4

PC1 PC1

2 2 3 3

1 1

4 4

PC1 PC1

Supplementary Figure 15. Wireframes of the lower pharyngeal jaw without the third dimension: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

4 4

1 1

3 3 2 2

PC2 PC2

2 2 3 3

1 1

4 4

PC2 PC2

2 2 3 3

1 1

4 4 PC2 PC2

Supplementary Figure 16. Wireframes of the lower pharyngeal jaw without the third dimension: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.1 and -0.1. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

4 4

1 1

2 2 3 3

PC1 PC1

3 3

2 2

1 1

4 4

PC1 PC1

3 3

2 2

1 1

4 4

PC1 PC1

Supplementary Figure 17. Wireframes of the upper pharyngeal jaw: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.3 and -0.3. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

4 4

1 1

2 2 3 3

PC2 PC2

3 3

2 2

1 1

4 4

PC2 PC2

3 3

2 2

1 1

4 4

PC2 PC2

Supplementary Figure 18. Wireframes of the upper pharyngeal jaw: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

2 2

1 1

3 3

PC1 PC1

3 3

2 2

1 1

PC1 PC1

3 3

2 2

1 1

PC1 PC1

Supplementary Figure 19. Wireframes of the interopercle: on the left side the positive side of the principal component axis 1 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

2 2

1 1

3 3

PC2 PC2

3 3

2 2

1 1

PC2 PC2

3 3

2 2

1 1

PC2 PC2

Supplementary Figure 20. Wireframes of the interopercle: on the left side the positive side of the principal component axis 2 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.

6. Supplementary

Supplementary Figure 21. CVA- graph with the target classifier “LAKE” from the upper pharyngeal jaw.

4 4

6 6

3 3 5 5

1 1 2 2

PC3 PC3

3 3

4 4

2 2

6 6 5 5

1 1

PC3 PC3

3 3

4 4

2 2

6 6 5 5

1 1

PC3 PC3

Supplementary Figure 22. Wireframes of the maxilla: on the left side the positive side of the Principal component axis 3 on the right side the negative values. Values are + 0.2 and -0.2. First row represents axis 1 vs 2, second row axis 1 vs 3 and last row axis 2 vs 3.