Differences in the Blood Cell Composition between two Ecotypes of Astyanax Mexicanus (Characid)

Table of Contents

1 Abstract

2 Introduction
2.1 La Cueva del Pachón - Abiotic and biotic factors of the Astyanax mexicanus cavefish's habitat
2.2 Niche conformance of the cave-dwelling ecotype of A. mexicanus
2.3 The fish's immune system
2.4 Adaptations in A. mexicanus blood cell composition due to selective pressures - Hypotheses and research questions

3 Materials and Methods
3.1 Fish origin
3.2 Animal husbandry
3.3 Sample collection and blood cell fixation
3.4 May-Grünwald/Giemsa stain
3.5 Light microscopical imaging and image processing
3.6 Image segmentation
3.7 Statistical analysis

4 Results
4.1 Improvement of the staining
4.2 Staining results of the samples
4.3 Relative abundances of different blood cell types
4.4 Different counting methods - manual counting versus counting by Biodock
4.5 Differences between Pachón cavefish and surface fish concerning the erythrocytes area

5 Discussion
5.1 Discussion of the fixation and blood sampling
5.2 Discussion of the May-Grünwald-Giemsa stain
5.3 Appearance of blood cells
5.4 AI quantification and cell classifying
5.5 Hypothesis conclusions
5.5.1 Immune cell abundance
5.5.2 Erythrocytes’ size

6 Conclusion

7 Outlook

8 Appendix

9 Acknowledges

10 References

1 Abstract

The characid Astyanax mexicanus is a coveted model organism for evolutionary ecology development and medical research.

Perpetual darkness, low nutrition availability and low-oxygenized waters characterize cave environment. That puts selective pressure on the cave-dwelling form towards extensive morphological and physiological adaptations, while still being able to generate fertile offsprings with the surface form. Under low parasite abundance wild forms of Astyanax mexicanus cavefish shifted their immune systems strategy towards a more sensitive adapted immune response compared to the surface form.

In this thesis, the blood cell composition between the cave-dwelling and the surface-dwelling form is analyzed. On top the erythrocyte size between these ecotypes is compared.

Blood smears from lab raised individuals of the Pachón-cave ecotype (N =9) and the surface ecotype (N = 9) were stained following the Pappenheim method. Light microscopy was used for generating images, which were analyzed with a trained deep-learning AI-model by Biodock.

No significant differences in blood cell composition were measured between the two ecotypes or between sexes and therefore no shift in the immune systems strategy could be proved. Cavefish's erythrocytes were on average significantly larger (22 %) compared to the surface form, indicating that hypoxia seems to be a strong selective pressure for fish.

Future investigations could reveal the abundance of immune cells in the peripheral blood of infected lab-raised fish or of wild fish inhabiting a different cave.

2 Introduction

2.1 La Cueva del Pachón - Abiotic and biotic factors of the Astyanax mexicanus cavefish's habitat

One of the Astyanax mexicanus cavefish's original habitat is a natural subterranean freshwater network in the Pachón cave. It is located on the western ridge of the northern part of the Sierra El Abra region in Northeast Mexico, which is a hilly, karstic, limestone terrain with a length 200 km and a width of 60 km as shown in Figure 1. (Espinasa, Pech, 2023) (Wilson et al., 2021)

Illustrations are not included in the reading sample

Figure 1: Localization of the Pachón cave. It is inhabited by the cave-dwelling form (CF) of Astyanax mexicanus. It is locoated in the northern part of the Sierra El Abra region in Tamaulipas, Mexico. Snapshot from Google Earth within the coordinates 22°36‘24.0‘‘N 99°02'21.0"W (Krishnan et al., 2020), (available at https://earth.google.com/web/ , accessed on January 22, 2024)

The basic, constant energy source in such an environment is chemoautotrophy with chemosynthesis, since, due to the lack of light, photosynthesis is not possible. There are no freshwater streams entering the cave. Percolating water dripping through layers of soil can provide some energy for the web of food by carrying particulate organic matter (POM) as well as dissolved organic carbon (DOC). It also carries lots of different small-sized species from the surface and the vadose zone. (Keene et al., 2016)

The amount of allochthone organic matter, therefore, fluctuates seasonally with the rainy season starting from June (monthly average precipitation = 213 mm) and ending in October (144 mm). (nomadseason, 2023) Decomposing organic materials provided by dead animals, brought in by bats and biofilms contribute to basic energy sources in the cave, shaping an oligotrophic environment. (Culver, Pipan, 2019) (Poulson, 2012). Ultimately, the absence of photosynthesis and the lack of constant water inflow also results in a low water oxygenation of the subterranean cave water in comparison with surface water. Only 2.97 mg/L of dissolved oxygen were found in the subterranean waters of the Pachón cave whereas 8.20 mg/L were measured in surface habitats. (Boggs et al., 2022)

The A. mexicanus surface form (SF) on the contrary, inhabits rivers and lakes in the same region with constant and rich nutrition income through the river. (Maldonado et al., 2020)

There is a diverse microbiotic community in caves accounting for the base of the food chain. Auto- trophic-chemolithotrophic sulfur-based as well as heterotrophic-chemoautotrophic microorganisms can settle in caves. (Pacioglu et al., 2023) , (Moldovan et al., 2018). In the Pachón cave there is also a multicellular fauna present building a complex web of food, with the Astyanax mexicanus being the top predator in its aquatic habitat. Cladocera water fleas, harpacticoid, copepod, ostracod, isopod and other unidentified arthropods had been described as preys in the digestive system of Astyanax fry in the Pachón cave. The adult fish's digestive system, however, indicated only mostly guano of bats, partly soil detritus and occasional arthropods as food sources for the A. mexicanus cavefish. There was a high abundance of empty stomachs among adults. (Espinasa et al., 2017)

Besides prey-predator relations, also host-parasites interactions are part of the A. mexicanus ' biotic habitat. Only two macroparasites, that were classified as monogeneans, were found at the cavefish: Cacatuocotyle cf. chajuli infecting the skin were found in 40 % of all examined cavefish, Characithe- cium cf. costaricensis infecting the gills were found in all examined cavefish. On A. mexicanus SF on the contrary, the more microparasite species were abundant (Santacruz et al., 2023). Moreover, there are theoretical pattern suggesting a strong correlation between low biodiversity and low parasite diversity (Lafferty, 2012).

In summary, the cavefish's abiotic habitat of the Pachón cave are oligotrophic, aphotic, and low-oxygenized aquatic pools with a fluctuating input of DOC or POM and low parasite abundance. Although being at the top of the food chain, adults of A. mexicanus cavefish must show physiological and morphological adaptions to the extreme low energy availability and develop niche conformance. The Asty- anax mexicanus SF habitat is characterized by a high nutrition availably, and higher parasite abundance than in caves.

2.2 Niche conformance of the cave-dwelling ecotype of A. mexicanus

The Pachón cave is one of 34 caves in the northeast of Mexico inheriting cave-dwelling, troglobitic ecotypes of Astyanax mexicanus (Miranda-Gamboa et al., 2023). The exact time, when the characid teleost Astyanax mexicanus invaded the caves is still controversial, although it was revealed that two phylogenetic lineages exist. Depending on the method of analysis the timeframe ranges between 20.000-100.000 ya for the old lineages and 20.000-30.000 ya for the newer lineages (sequencing single nucleotide polymorphisms) (Fumey et al., 2018) and 7.8-8.1 mya (sequencing of mitochondrial DNA) (Gross et al., 2013). Although being separated for a long time period, the CF and SF are still able to generate fertile offsprings and are defined as one species, despite strong phenotypical characteristics as seen in (Maldonado et al., 2020)

Illustrations are not included in the reading sample

Figure 2: Appearance of two ecotypes of Astyanax mexicanus. Highly remarkable are the eye regression and loss of pigmentation of the cave-dwelling (CF) form as morphological differences to the surface-dwelling form (SF). (Rohner, 2023)

Morphological examples of regressive evolution such as the loss of eyesight are evident in the cavefish and must be seen in the light of natural selection favoring genetic compositions of low energetic costs. (Protas et al., 2007) (Jeffery, MARTASIAN, 1998)

Natural selection also drives physiological adaptation, equipping the cavefish with insulin receptors that show lowered affinity to bind glucose and enable therefore a thriftier consumption of glucose (Kling, 2018). Through mutations in the melanocortin 4 receptor cavefish populations developed hyperphagia, enabling them to consume more nutrition in rare times of high nutrition availability. (Aspiras et al., 2015). This comes along with an increased fat storage capacity in the Pachón cavefish thanks to hypertrophic visceral adipose tissue. (Xiong et al., 2018)

Adaptions in related metabolism pathways enable a healthy form of hyperglycemia and adiposities and inhibit the expected consequences of a high blood sugar and saturated fatty acids level such as diabetes or cardiovascular diseases. Counteracting the metabolic stress triggered by strongly fluctuating nutrition availability including the pressure to consume all nutrients if available and to endure long-term starvation, seems to be an ability favored by natural selection. (Medley et al., 2022).

Also, hypoxia made cavefish develop a more efficient oxygen transport. As cells containing the oxygenbinding hemoglobin, erythrocytes were objects of investigations concerning possible adaptations. Two perspectives on cellular level are currently discussed considering a higher erythrocytes production or larger erythrocytes. Erythrocytes from cavefish in the Pachón cave were significantly larger (16 %) than those from surface fish. Larger erythrocytes are positively correlated with an elevated hematocrit and hemoglobin level and optimize the oxygen/carbon dioxide diffusion area due to a more efficient area/volume ratio (Boggs et al., 2022). An elevated hematocrit level could also be the consequence of a higher erythrocytes count induced by an overexpression of primitive and definitive waves of hematopoiesis regulator genes such as gfi1aa during embryogenesis. (van der Weele, Jeffery, 2022)

In summary, A. mexicanus cavefish shows multiple examples of natural selection pressured adaptions. One of them includes its immune system, which is in the focus of the next section.

2.3 The fish's immune system

For all vertebrates the immune system is divided into innate and adapted immune response. The innate immune response of vertebrates consists of outer physical barriers like scales or mucosa as well as cellular defenses in tissues of infection (Lieschke, Trede, 2009). Here, phagocytosis is a main factor to firstly counteract pathogens, releasing molecules to combat them or to support the own immune response. This is done by granulocytes such as eosinophiles, neutrophiles and heterophiles as well as monocytes differentiating to macrophages in place and time of infection. In healthy state, they circulate in the blood acting at the inflammatory site within minutes. Their recognition of pathogenic structures (antigens) on phylogenetically old bacteria and parasites appearing with high frequency in populations is based on inheritance. (Urry et al., 2019). This innate system comes with high constitutive costs because of high production of reactive inflammatory molecules like cytokines, triggering severe consequences for the entire organism like fever, which reduces the hosts fitness. (Weber et al., 2022)

Illustrations are not included in the reading sample

Figure 3: Hematopoiesis of immune and blood cells. Beginning from a Hematopoietic Stem Cell (HSC) the diversification is firstly divided into two branches: the lymphoid and the myeloid strain. Several progenitors are developed and further specfied resulting in Red Blood Cells (RBC): Erythrocytes, thrombocytes and White Blood Cells (WBC): Eosinophiles, neutrophiles, monocytes and lymphocytes (here: B-Cells). Natural Killer (NK) as well as basophiles cells are not present in cavefish's blood. T- cells would be present in cavefish's blood in case of infection. (Traver et al., 2003 a)

If the antigens of pathogens cannot be bound and pathogens are not disassembled by the hosts phago- cytotic cells, then the adapted immune system is recruited. It begins with lymphocytes presenting antigens which then enables a cascade of cell recruiting mature differentiated B- and T-lymphocytes. They are responsible for binding pathogens and releasing their cytotoxic material (Urry et al., 2019). This immune reaction has far less consequences for the entire organism and therefore does not reduce the host's fitness in case of infection. (Weber et al., 2022)

The production of immune cells originates in the hematopoiesis as shown in Figure 3. During the adult phase of hematopoiesis in teleost, the Hematopoietic Stem Cell (HSC) develops and splits into the Common Lymphoid Progenitor (CLP) and the Common Myeloid Progenitor (CMP) in the head kidney (pronephros) sectioning off the lymphoid branch and the myeloid branch of blood cell genesis. The lymphoid branch of hematopoiesis is related to the adapted immune response, generating lymphocytes such as B cells, T cells and NK-cells. The myeloid branch of hematopoiesis is related to the innate immune response generating erythrocytes, thrombocytes, monocytes, and granulocytes (Kondera, 2019).

2.4 Adaptations in A. mexicanus blood cell composition due to selective pressures - Hypotheses and research questions

When comparing the immune cell composition in the head kidney of wild ecotypes of A. mexicanus cavefish in the El Pachón cave and surface fish, the A. mexicanus cavefish showed a higher proportion of cells belonging to the adapted immune response. This was shown by calculating the ratio of immune cells of the myeloid strain and the cells of the lymphoid strain of hematopoiesis after triggering immune cell proliferation by bacterial endotoxins. (PeuB et al., 2020)

As cavefish cannot afford a temporarily reduced fitness in a harsh cave environment, the question is whether healthy individuals, born and raised under controlled lab conditions, also show this shift of immune response. Other adaptations seem to fade away, when exposed to a controlled lab condition. For instance, genes in Pachón cave fish, if raised in lab conditions, show a downregulated expression concerning their glucose metabolism, circadian rhythm, and oxidative stress in their liver transcriptome. (Krishnan et al., 2020)

Nevertheless, the expression of the immune systems shift seems to be genetically determined, since a genetical regulation of HSCs proliferation in Pachón cavefish towards the lymphoid strain of hematopoiesis was verified. (PeuB et al., 2020)

Always pressured to an efficient energy consumption as shown in section 2.2 the cavefish must avoid the energy-draining innate immune response where possible. Also, A. mexicanus would more frequently encounter a parasite/pathogen in the surface area rather than in caves. It is shown, that adaptive immune responses are according to energy-trade-off hypotheses beneficial, if a low parasite abundance prevalences. (Mayer et al., 2016)

Biotic and abiotic factors applying selective pressure towards a genetic composition in the gene pool of the A. mexicanus Pachón ecotype, which shifts the immune cell genesis towards the adaptive immune response. This leads to the following hypothesis and expectation:

There are significant differences in the blood cell compositions between the ecotype of Astyanax mexicanus inhabiting the Pachón cave and the surface ecotype.

In particular, I expect to see differences in the ratio of the relative abundance of granulocytes and monocytes, who are associated with the innate immune response, and lymphocytes being associated with the adapted immune response.

The methods applied in this thesis allow to determine the size of erythrocytes, where also adaptations were found as stated in section 2.2 due to the hypoxia in cave waters. Hence, I can add the following hypothesis:

The size of erythrocytes is significantly larger in Pachón cavefish compared to surface fish of A. mexicanus.

3 Materials and Methods

3.1 Fish origin

All fish are offsprings of either a population originated in the cave Pachón or of a surface population of A. mexicanus and were bred and kept under controlled laboratory conditions in Münster.

3.2 Animal husbandry

All fish were supervised in the aquatic facility at the Institute for Evolution and Biodiversity at the University Münster. The net volumes of the tanks were either 15 liters (275x400x200 mm) housing 512 fish or 95 liters (820x400x300 mm), housing 20-50 fish. UV, mechanical and biological methods were applied for water filtering. Water temperature was at 21 °C, the pH was held over 8. Both ecotypes were maintained under a 12h:12h light-dark rhythm. They were fed on Fridays with granulate (Gemma micro, Gemma silk) and on Wednesdays with red bloodworms (~12:00 am). All housing conditions and collection methods are regularly approved by the Health and Veterinary Office of Münster (recently, 04 April 2023).

3.3 Sample collection and blood cell fixation

Eighteen fish were euthanized and dissected between 08:00 am and 10:00 am to reduce possible disruptive factors like the circadian rhythm. They were incubated in a tricaine methane sulfonate solution (MS22, 0.5 g/L) by a licensed coworker for at least 15 minutes to ensure brain death. Standard length from the tip of the mouth to the base of the caudal fin and the weight with a precise scale was measured and the distribution of the sample for examination is shown in Table 1. In the appendix in Table 7 the dataset for all specimens is given. Some samples were not further used in this thesis. The sex was determined by a dissection of the gonads (Nfemale = Nmale = 9), ovaries and eggs were found in the female organism.

Table 1 shows average weight and length of the 18 samples used in this thesis. It was differed between the two ecotypes of A. mexicanus.

Table 1: Biometric data from individuals of A. mexicanus. Weight and length were measured

Illustrations are not included in the reading sample

To reduce contamination probability and for safety reasons nitrile gloves were worn during the fixation and staining process. Blood was taken from the caudal vein with an Eppendorf One-Channel 10-100pl mechanical micropipette after cutting through the muscle tissue of the caudal fin at the caudal peduncle as shown in Figure 4. The fish was used for further investigations for another project. The obtained blood was applied on slides. Blood smears of a thin blood layer were then generated by gently pressing one bloody and one clean slide together and pulling the slides along the longitudinal axis of the slides.

This way, thin blood smears on both slides were generated.

Illustrations are not included in the reading sample

Figure 4: Access point for the blood samples and standard length of fish. Standard length was measured along the shown distance of the black bars. In red line indicates the cut for accessing the caudal vein. On the left there is an individuum of A. mexicanus cavefish form, on the right there is the surface fish form.

After air-drying for 30 seconds, the slides with the blood smears were incubated in ethanol (c > 98 %) for at least five minutes to fix the cells onto the slide. Next, slides were labeled, air-dried for at least 2 hours and were then ready to use for the staining process. The staining procedure took place two to six weeks after the fixation process. Before and after staining, specimens were kept in a dark and closed slide storage box at room temperature.

3.4 May-Grünwald/Giemsa stain

Staining was done following the Pappenheim protocol by the manufacturer (Sigma-Aldrich). May- Grünwald and Giemsa are the components of this two-dye stain.

However, the protocol showed weaknesses in the staining result; hence, the goal was to optimize the Pappenheim staining.

In an experimental approach, the following steps of the workflow were changed individually to look for possible microscopical changes in the coloring of the blood cells in comparison to the standard protocol.

- Age of the fixed blood smears
- Rinsing method
- Length and dilution of the May-Grünwald staining step
- Length and dilution of the Giemsa staining step
- pH of the PBS buffer in between the two stains

The decision for the final staining protocol (results are shown in section 4.1) was however in the end to only lower the pH of the buffer from 7.2 to 7.0.

Also, blood samples were not anticoagulated in EDTA. Premixed solutions by Sigma-Aldrich were used. Chemicals were stored at room temperature. Stains were kept sealed and under aluminum foil. (Sigma-Aldrich #GS-10)

The steps of the workflow included:

- Incubation of the slides in May-Grünwald stain for 5 minutes
- Incubation in PBS-buffer for 90 seconds (pH = 7.0)
- Incubation in Giemsa stain for 15 minutes
- Rinsing of each side of the slide (5-10 seconds) with demineralized water
- Air-drying with a cold fan for fastening the drying process

The following chemicals were used:

GIEMSA STAIN

- GS500, modified, 0.4 % w/v, in a buffered methanol solution, pH 6.9, with stabilizers.
- Diluted with aqua dest. (c = 1:20)

MAY-GRÜNWALD STAIN

- MG500, 0.25 % w/v, in methanol.

PHOSPHATE BUFFERED SALINE

- Tablets dissolved in 1 liter of aqua dest.
- The buffer was generated using NaOH and HCl at a pH of 7.0 at 25 °C.

3.5 Light microscopical imaging and image processing

To be able to examine the specimens by light microscopy and for preservation reasons, MERCK NeoMount™ was applied on the specimens as a color-stable and xylene-free mounting medium. A xylene, e.g. MERCK NeoClearTM substitute to prevent the occurrence of turbidity was not necessary.

The cover glass was put carefully on the applied Neo-Mount™ to prevent air bubbles. Slides were dried in a horizontal position. It took 60 minutes for the specimens to be completely dried and hardened. For safety reasons all work (staining and mounting) was done under the fume hood.

Microscopical brightfield investigation with the ZEISS Axioskop 2 was then conducted after the Köhler process with 100x magnification and under oil immersion with ZEISS immersion oil 518 N.

The collecting of images was done systematically to prevent that cells were photographed twice and to reduce biases. The slides were examined macroscopically and then the regions of interest were determined. Promising areas were those with a smooth surface without ripples and holes. (Burton, 2021)

Illustrations are not included in the reading sample

Figure 5: Example of decision making for the regions of interests on a stained slide. Red lines show the theoretical line along the microscope stage was moved. Areas of the vertical line were photographed. Additionally, in some cases a second horizontal was also considered. The yellow area depicts a spot that should be avoided due to its unfavorable characteristics. Arrows show moving direction.

The microscope stage was cautiously moved along the theoretical vertical line as marked in Figure 5. Images were snapped with the microscopical camera ZEISS Axiocam 305 color approximately every 100-200 pm. Only if the region of interest (ROI) allowed a doubtless counting an image was saved. If a ROI appeared to spongy or crowded because of contaminations or artefacts, the region was skipped. To reach a total cell number of at least 1000 per sample, an additional theoretical horizontal line was drawn, and the microscope stage was moved along that line in the opposite direction of the vertical line. Image processing was done with ZEISS Zen Blue Version 3.5. It aimed at enhancing the image quality to facilitate the segmentation and to export images for segmentation and counting. The largest format 2464 x 2056 pixels was selected. The software calculated a theoretical pixel size of 0.035 pm x 0.035 pm. Automatic exposure was generated by the Zeiss Software preprogramed for “Best Fit” image analysis. The exposure time was mostly between 20 ms and 60 ms. Shading correction and white balancing was set manually. The noise filter was enabled. Finally, the images were converted from .czi files to .tif files.

3.6 Image segmentation

For image segmentation and counting, the deep-learning artificial intelligence (AI) tool Biodock™ was used (Biodock. Biodock, 2023). The algorithm, shown in Figure 6 is based on “Segment Anything Model (SAM)”, which includes promptable segmentation (Kirillov et al., 2023).

Illustrations are not included in the reading sample

Figure 6: Concept map of the AI-Segment-Anything Model (SAM). The model consists in three parts the prompt encoder processes annotatations of the user, while the image encoder processes the information of the image data. The lightweight mask decoder component connects both encoders and generates valid masks. By adding annotations and retraining the model can be further improved. (Kirillov et al., 2023)

The workflow included 3 steps:

1. Annotation of labels and first training

After uploading the training data as .tif files, which included 67 images from samples that are not used for the analysis, 7 classes were assigned, so that the cells could be grouped to them as objects. Assigned classes included erythrocytes, immature erythrocytes, thrombocytes, lymphocytes, monocytes, granulocytes, and dead cells. Dead cells were excluded before calculating the relative abundance of the classes. The minimal count of object per class is 10, but mostly 25 objects per class are recommended. Table 2 shows, that the minimal requirements for the training process was met.

Table 2: Defined classes (cell types) and amount of labels assigned to the respective cell type during AI training. Labels were made during the training process and assigned to one of those classes.

Illustrations are not included in the reading sample

2. Evaluation and retraining

Next, the training process was started with the Model “Alligator” at a threshold of 0.05. An average precision (AP) > 90 % for mask segmentation along all classes was reached by the trained AI-model (e.g. AP of erythrocytes = 94.4 %). The result was reevaluated, which means it was checked for errors. Errors included wrong classification, poor segmentation precision, and no recognition of overlapping cells. Correcting the segmentation result was done six times times to increase the average precision. When the result of the training results met the requirements, a test was performed with random images from the dataset. One image per sample was taken and counted manually to receive control data and to compare it with the results of the AI segmentation and counting.

3. Data acquisition

In the end, the entire dataset was segmented and counted, and the results were downloaded as .csv files for analyzing the results concerning the relative abundance.

For drawing data concerning the erythrocytes area, data was excluded. All objects classified as erythrocytes smaller than 10.000 pixel were filtered out, to exclude cropped cells at the edge of images. Due to segmentation mistakes, also objects with a large diameter were filtered out.

3.7 Statistical analysis

For data plotting and the statistical analyses, the Software R (Version 4.2.3) was used. (R Core Team. R: A language and environment for statistical computing, 2018)

The packages “ggstatsplot”, “ggplot2”, “ggsignif”, “ggpubr“, “dplyr“, and “viridis“ were installed. The Shapiro-Wilk test and the Levene test were performed to ensure the conditions for the ANOVA test are met.

An ANOVA test was performed to check for significant differences between ecotypes, sexes, and their interaction between each other. The null hypothesis was set at p < 0.05, which would indicate a significant difference. For a p value above 0.05 the hypothesis would be rejected.

For validation of the result of the AI, a manual cell count of a random sample was performed. To check for normality, the Shapiro-Wilk test was performed. To calculate, whether the variance of the data is equally distributed or not, the Levene test was performed.

After that, the erythrocytes area of both ecotypes was compared, and a Welch two-sample t-test was performed to check for any significant differences.

4 Results

Samples from 18 fish (Npachon = 9, Nsurface = 9) were taken. Female and male fish were also equally distributed. Six types of cells could be classified: Erythrocytes, immature erythrocytes, thrombocytes, lymphocytes, monocytes and granulocytes. Further classification between those classes, e.g. differencing also progenitors was not made and it was not differed between types of granulocytes (basophiles, neutrophiles, eosinophiles). This thesis included a total amount of 38,855 cells.

4.1 Improvement of the staining

The staining of the original protocol of the manufacturer resulted in poor erythrocytes cytoplasm. Therefore, changes in parameters of the staining were altered individually to look for possible changes in the staining quality. Changes included the rinsing method, length and dilution of the May-Grünwald staining step , Length and Dilution of the Giemsa staining step and pH of the PBS in between the two stains. In the following conclusions from those changes are shown:

Blood smears of slides that were fixed months ago resulted more pale in comparison to stained blood smears that were fixed hours ago.

When changing the rinsing method (PBS 7.0 instead of demineralized water or longer than 10 seconds), it resulted in an irregular heterogeneous staining, where the stain of some cells was washed out, as seen in Figure 17.

When the dilution the of May-Grünwald Stain was reduced from not diluted to 1:2, the cytoplasm of the monocytes was poorly stained, while the cytoplasm of the erythrocytes appeared beige-pale. When extending the incubation time in the May-Grünwald stain for 1-2 minutes, the erythrocytes cytoplasm appeared to have more contrast.

When the dilution of the Giemsa stain was above 20 min, the cells appeared to be generally dark and hardly differentiable. The same was the case when the dilution of Giemsa was reduced from 1:20 to 1:10 as seen in Figure 18 in the appendix.

Remarkably high resolution of the chromatin was observed, when incubating for 20 min in Giemsa and in a PBS of pH = 6.8, as seen in Figure 19.

At a pH of 7.4 of the PBS, the Erythrocytes' cytoplasm appeared greenish grey. When lowering the pH to 7.0 such a coloring was not observed anymore. In the appendix in Figure 20 the effect of lowering the pH of the PBS is shown.

4.2 Staining results of the samples

Table 3 shows the qualitive staining results. In total, six cell classes were identified. Dead cells were abducted from all other results but had to be segmented and assigned as a class in Biodock. In example of a typical blood cell layer is shown in the appendix (Figure 16)

Table 3: Examples of identified cell in blood samples of A. mexicanus. Blood smears were stained with May-Grunwald/Giemsa following the Sigma-Aldrich protocol. Images were taken with 100x objective (NA =1.30) under oil immersion. The scale was theoretically calculated by the microscopical image processing software ZEISS Zen blue 3.5. Size was calculated as follows Npixel of cell * 0.0352 pm[2].

Illustrations are not included in the reading sample

4.3 Relative abundances of different blood cell types

The results in this sub chapter show the relative abundance of different blood cells (dependent variables) of individuals of A. mexicanus belonging to the samples “Pachon-female” (n = 5), “Surface-female” (n = 4), “Pachon-male” (n = 4) or “Surface-male” (n = 5). The relative abundances of the following blood cell types were analyzed: erythrocytes, immature erythrocytes, thrombocytes, monocytes, lymphocytes, and granulocytes. Also, dead cells, which could not be classified, were found, and were subtracted from N total (see equation below). The relative abundance was calculated according to the following equation:

Illustrations are not included in the reading sample

With N cell type of interest being the counted number of a certain blood cell type, i.e. erythrocytes, and N total being the total number of all cells less the number of dead cells.

Outliers had a strong impact on the mean because of the small sample size. Therefore, the median (Q2) and the interquartile range (IQR) were chosen for the descriptive statistics.

Figure 7 shows the relative abundance of erythrocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2 = 92.5 % and the interquartile range was IQR = 2.6 %. In female surface fish, the median was Q2 = 92.4 % with an IQR of 12.1 %. In male Pachón cavefish, a median of 91.8 % with an IQR of 4.6 % was found. For male surface fish, a median of 97.2 % and an IQR of 4.1 % were found.

To check if the data is normally distributed, a Shapiro test was performed. The result of the test indicates a normal distribution (p = 0.84) with a slight heteroscedasticity (Levene-Test: p = 0.02)

Next, it was examined if there is a significant effect of the factors on the dependent variable, i.e. if ecotype and/or sex of the fish influence the relative abundance of erythrocytes. For this, the parametric ANOVA test was performed, which offers the possibility to compare between more than two factors. Also, its conditions (normality and homoscedasticity) were sometimes met. Besides it can test whether the effect of an ecotype on the dependent variable is the same for both sexes or if there is a differential effect depending on the sex of the subject.

According to the result of the ANOVA test, significant differences in the relative abundance of erythrocytes between sex and ecotype were not found. Within the ecotype, a p value of 0.347 was calculated, within the sex p = 0.581 was calculated and the effect of ecotype on the erythrocyte abundance was the same for both sexes (p = 0.335).

Illustrations are not included in the reading sample

Figure 7: Relative abundance of erythrocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed in box plots showing the median as a horizontal line in the box and the mean as an “x”. Boxes are drawn from the 25th to the 75th percentile. Whiskers extend until the 95th percentile and the 5th percentile. Single data points are indicated as dots. NS = not significant, as no significant differences were measured between sex or ecotype or between the interaction of ecotype and sex.

Figure 8 shows the relative abundance of immature erythrocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2 = 0.218 % and the interquartile range was IQR = 0.810 %. In female surface fish, the median was Q2 = 0.141 % with an IQR of 0.296 %. For male surface fish, a median of 0.075 % and an IQR of 0.296 % were found. Immature erythrocytes were found in only two (x = 0.105 %) out of 4 samples of male Pachón cavefish.

To check if the data is normally distributed, a Shapiro-Wilk test was performed. The result of the test indicates a normal distribution (p = 0.84) with a slight heteroscedasticity (Levene test: p = 0.02) According to the result of the ANOVA test, significant differences in the relative abundance of immature erythrocytes between sex and ecotype were not found. Within the ecotype a p value of 0.405 was calculated, within the sex p = 0.981 was calculated, while the effect of ecotype on the immature erythrocyte’s abundance was the same for both sexes (p = 0.139).

Illustrations are not included in the reading sample

Figure 8: Relative abundance of immature erythrocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed as described in Figure 7.

Figure 9 shows the relative abundance of thrombocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2 = 5.57 % with IQR = 0.68 %. In female surface fish, the median was Q2 = 4.86 % with IQR = 8.46 %. In male Pachón cavefish, a median of 5.81 % with IQR = 5.07 % was found. For male surface fish, a median of 1.66 % and an IQR of 2.88 % were found. The result of the Shapiro test indicates a normal distribution (p= 0.54) with heteroscedasticity (Levene-Test: p=0.006)

According to the result of the ANOVA test, significant differences in the relative abundance of thrombocytes between sex and ecotype were not found. Within the ecotype a p value of 0.225 was calculated, within the sex p = 0.516 was calculated, while the effect of the ecotype on the thrombocytes’ abundance was the same for both sexes (p = 0.203).

Illustrations are not included in the reading sample

Figure 9 : Relative abundance of thrombocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed as described in Figure 7.

Figure 10 shows the relative abundance of lymphocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2= 0.888 % with IQR = 0.860 %. In female surface fish, the median was Q2= 1.248 % with IQR = 1.985 %. In male Pachón cavefish, a median of 1.116 % with IQR = 0.738 % was found. For male surface fish, a median of 0.393 % and an IQR of 0.991 % were found. The result of the Shapiro-Wilk test indicates a normal distribution (p= 0.35) with Homo- scedasticity (Levene-Test: p=0.49)

According to the result of the ANOVA test, significant differences between sex and ecotype were not found. Within the ecotype a p value of 0.772 was calculated, within the sex p = 0.518 was calculated, while the effect of the ecotype on the lymphocyte’s abundance was the same for both sexes (p = 0.398).

Illustrations are not included in the reading sample

Figure 10: Relative abundance of lymphocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed as described in Figure 7.

Figure 11 shows the relative abundance of monocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2= 0.411 % with IQR = 0.170 %. In female surface fish, the median was Q2= 0.153 % with IQR = 0.077 %. In male Pachón cavefish, a median of 0.266 % with IQR = 0.113 % was found. For male surface fish, a median of 0.208 % and an IQR of 0.060 % were found. The result of the Shapiro-Wilk test does not indicate a normal distribution (p < 0.01) but Homoscedasticity (Levene-Test: p = 0.72)

According to the result of the ANOVA test, significant differences between sex and ecotype were not found. Within the ecotype a p value of 0.661 was calculated, within the sex p = 0.654 was calculated, while the effect of the ecotype on the monocytes abundance was the same for both sexes (p = 0.215).

Illustrations are not included in the reading sample

Figure 11: Relative abundance of monocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed as described in Figure 7.

Figure 12 shows the relative abundance of granulocytes in the four tested samples. In female Pachón cavefish, the median of the relative abundance was Q2= 0.091 % with IQR = 0.044 %. In female surface fish, the median was Q2= 0.035 % with IQR = 0.252 %. In male Pachón cavefish, a median of 0.066 % with IQR = 0.322 % was found. For male surface fish, a median of 0.277 % and an IQR of 0.320 % were found. The result of the Shapiro-Wilk test does not indicate a normal distribution but indicate with Homoscedasticity (Levene-Test: p=0.61)

According to the result of the ANOVA test, significant differences between sex and ecotype were not found. Within the ecotype a p value of 0.862 was calculated, within the sex p = 0.390 was calculated, while the effect of ecotype on the granulocytes abundance was the same for both sexes (p = 0.555).

Illustrations are not included in the reading sample

Figure 12: Relative abundance of granulocytes in two different ecotypes and sexes of A. mexicanus. The diagram shows the results for both sexes (female and male) of Pachón cavefish as well as surface fish. Data is displayed as described in Figure 7.

In conclusion, the results neither indicate any significant influence of the sex on the blood cell abundance nor any significant influence of the ecotype on the abundance.

Table 4 shows the summary of the relative abundance of blood cell types in both ecotypes without distinguishing between the sexes. With a larger sample (n = 9), the mean and standard deviation were considered. It is shown that erythrocytes are by far the most abundant cell type in the blood of all tested samples in both groups (x > 90 %). The second most abundant cells in both ecotypes are thrombocytes (Pachón: x = 6 %, surface: x = 4 %). Lymphocytes are the most abundant leucocytes in the analyzed blood cells (x = 1 %), whereas the relative abundances of monocytes and granulocytes are lower than 1 %.

Table 4: Mean relative abundance of blood cells [%] in two ecotypes of A. mexicanus. The table shows the mean (x) and the standard deviation (SD) of each cell class in the entire dataset (n = 18). All numbers in the table are depicted in [ %].

Illustrations are not included in the reading sample

Also, the relation of the number of monocytes (N monocytes) plus the number of granulocytes (N granulocytes) per number of lymphocytes (N lymophcytes) was calculated. This ratio is called the M/L ratio. The following equation was used for the calculation:

Illustrations are not included in the reading sample

The M/L ratios of both ecotypes regardless of the sex (Pachón: n = 9, surface: n = 9) were compared (see Figure 13). In the Pachón cavefish group, a median of Q2 = 0.472 % (IQR = 0.351 %) and in the surface fish group a median of Q2 = 0.333 % (IQR = 0.599 %) were found. Due to two outliers the standard deviation is much higher in the Pachón group (SD = 0.635 %, x = 0.656 %) than in the surface group (SD = 0.333 %, x = 0.47 %).

No significant differences in the M/L ratio between the two ecotypes were calculated after a Welsh’s t- test (p = 0.46).

Illustrations are not included in the reading sample

Figure 13: M/L ratio in two different ecotypes of A. mexicanus. The diagram shows the results for Pachón cavefish as well as surface fish. Data is displayed in violin and box plots showing the median as a horizontal line in the box and the mean as a red dot. Boxes are drawn from the 25th to the 75th percentile. Whiskers extend until the 95th and 5th percentile. Single data points are indicated as dots. Visualization was done with GGstatsplot2. (Patil, I. Visualizations with statistical details: The ggstatsplot approach, 2021)

4.4 Different counting methods - manual counting versus counting by Biodock

Cells were counted by using the trained AI model of the Biodock program. For validation of the AI- models counting results, a subset of the data was counted manually, and the same subset was analyzed

Illustrations are not included in the reading sample

Figure 14: Comparison of the two different cell counting methods applied on a random sample of blood cells of A. mexicanus . The diagrams show the results in absolute numbers for each cell type obtained with the manual counting (blue bars) and with the AI program (Biodock). The results of two different training versions of the AI program are shown (version 5: green bars, version 6: yellow bars). Biodock version 6 was finally used to perform the segmentation and analysis of the blood cells in A. mexicanus. A random control sample of taken by adding one image per sample resulting in 18 images and 970 cells accoding to the manual count.

The subset includes one random image per sample (N = 18) In total, 970 cells were counted manually, while the latest version of Biodock counted 15 cells less. Table 5 compares the relative abundance of each cell type neglecting dead cells.

There is less deviation concerning the relative count of version 6 (0.247 %) than in Version 5 (0.523 %) in relation to manual counting. Figure 14 depicts the results of the cell counting in absolute numbers and includes dead cells. An improvement of precision in the cell count can be observed, which does not totally match with the result of the manual counting.

Table 5: Relative abundance of blood cells [%] of A. mexicanus in a random control sample. The table shows the relative count of each cell class in that random sample. Two different counting methods were applied. A random control sample of taken by adding one image per sample resulting in 18 images and 970 cell according to the manual count. All numbers depicted in this table are in [%]. * Relative standard deviation of the relative count compared with manual counting.

Illustrations are not included in the reading sample

4.5 Differences between Pachón cavefish and surface fish concerning the erythrocytes area

The AI program Biodock also provides pixel counts of perimeter and object area. Therefore, the size of erythrocytes was evaluated statistically. Outliers were excluded according to Table 8 in the Appendix. Objects with an area below 12,500 Pixel were always cropped located at the edge of the segmented image. Those false positives would wrongly manipulate the average erythrocyte size. Also, the segmentation rarely made mistakes, when assigning two separated labels but assign it to one object. This could be made visible with a graph as seen in Figure X. in the appendix, there is no correlation between area and perimeter anymore, when that segmentation mistake happened.

Table 6: Statistical description of the ecotypes Pachón and surface and of female, male concerning their erythrocytes area. The cleaned dataset was taken. Visualization in Figure 15.

Illustrations are not included in the reading sample

Figure 15 shows the violin plot of the cleaned data of the two categories “Pachón” and “Surface”. See Table 6 for comparing the mean, median and SD between Pachón and surface as well as female and male. Testing for significant difference in the size of the erythrocytes between both ecotypes was done with a Welch two-sample t-test. The difference between both ecotypes concerning the erythrocytes size is significant (p < 0.001). There are no significant differences between male and female fish. There is a 95 % chance to pick a cell from SF that is between 6.9 ^m[2] and 7.1 ^m[2] smaller than the one from CF.

Illustrations are not included in the reading sample

Figure 15: Differences between Pachón cavefish and Surface fish concerning the erythrocytes area. The dataset with cleaned data was taken, to exclude cells that were cropped by the edge of the segmented image and to exclude cells with two seperated areas. X = mean, O = median, Brackets in light grey: standard deviation. p < 0.001. 6.9 inn[2] < CI95 < 7.1 inn[2]. CI = Confidence Interval.

5 Discussion

5.1 Discussion of the fixation and blood sampling

High quality standards for fixation and staining are essential for discriminating between different types of blood cells. This facilitates the annotation of objects to classes during the AI training process. Also, the AI's precision will increase with clear staining results. In the following, potential sources of error and recommendations for improvement of the fixation and staining procedure are discussed.

Several staining protocols recommend methanol instead of ethanol for fixation. In this experimental approach difference between those two liquids could not be recognized.

Holes in the cell layer later during microscopic investigations were observed. A quick incubation of the clean slides in ethanol and drying before the fixation, which dissolves any hydrophobic films would be recommended.

Many slides could not be used as the overall blood smears were covered with undifferentiable matter, possible tissue residues. It is likely that the first pl of the fish’s blood, which is taken from the caudal vein, possibly contains more unwanted organic matter than the rest. Not using the first pl of a fish would be recommended.

To ensure thin blood smears of constant quality (Burton, 2021), applying the standard technique for generating blood smears, could be taken into consideration. The alternative technique, that I applied did not always delivered promising thin blood smears. Nevertheless, the advantage was, that the number of slides used for investigation could be doubled.

5.2 Discussion of the May-Grünwald-Giemsa stain

The panoptic Pappenheim stain is a two-step and three-dye method with eosin Y and methylene blue dyes of May-Grünwald solution and azure B, eosin Y, methylene blue solution of Giemsa solution. (Piaton et al., 2016).

The Pappenheim staining method applied in this thesis, combining the effects of the May-Grünwald and Giemsa stains, achieved acceptable results enabling a distinction between several blood cell types. Because of the polychromatic effect, cell components were stained differently. This bases on the isoelectric point of their proteins and the H+-availability of the milieu influenced by the pH of the PBS and the dilutions' buffer. (Mulisch, Welsch, 2015)

In optimal conditions, acidophilic/eosinophilic components of the cytoplasm like hemoglobin and granola are stained in pink-orange-red colors as the eosin Y binds to those structures due to electrostatic adsorption. The acidic and therefore basophilic chromatin milieu of the nuclei binds base dyes like azure B, as the nuclei are stained in purple colors. . (Mulisch, Welsch, 2015).

Interactions between methylene-blue and its oxidation products explain the blue cytoplasm of lymphocytes or the grey cytoplasm of monocytes and immature erythrocytes and neutrophiles, which are all basophilic structures (Horobin, 2011), (Binder et al., 2012).

However, monocytes' and granulocytes cytoplasm also appeared often pale and spongy. Reasons for such a hemogram could be a too long time between fixation and staining of the blood smears, as can be concluded from the experimental approach of staining, where the duration between fixation and staining was more than one month. It is recommended to stain within two weeks after the fixation (Binder et al., 2012). Other recommendations state to stain directly after the fixed slides are dried (Piaton et al., 2016). Also, the incubation time of slides in the May-Grünwald and Giemsa staining solution could have been lengthened for 10-20 %, to intensify the color of the stains (Horobin, 2011).

Moreover, after cleaning the container with the staining solution, precipitation was observed, which also can lead to a reduced intensity and strong varying in the coloring due to a decline of the dissolved stains concentration. It is recommended to replace the staining solution after 5 throughputs (Binder et al., 2012) or after one day (Piaton et al., 2016). On top, dyes could be oxidated because of air contact, as staining containers were maybe not properly sealed (Piaton et al., 2016).

Furthermore, there could be irregularities in the staining results, as no standardized automated staining machine (e.g. Tarmac™) was used, that would have reduced many of the above-mentioned mistakes.

5.3 Appearance of blood cells

In the following the identification of cell classes is discussed. The erythrocytes were the easiest cells to recognize manually as shape and color were different from all other cells. Also, immature erythrocytes could be labeled with much confidence because their shape was similar to those of the erythrocytes. Tough, however, was to differ between thrombocytes and lymphocytes, as their shape and color were comparable. Because thrombocytes and the nuclei of erythrocytes are both basophilic, a rare confusion cannot be excluded, which explains lower abundance of erythrocytes and higher abundance of thrombocytes compared to finding from other publications (Traver et al., 2003b). In this paper flow cytometry was applied counting. 97.6 % of the cells in the peripheral blood were erythrocytes or immature erythrocytes, 0.6 % were thrombocytes or prethrombocytes, 1.0 % were lymphocytes and less than 1.0 % were monomyelocytes.

Moreover, no further identification of myelomonocyte and lymphoid progenitors was done nor a differentiation between prethrombocytes and thrombocytes, as the dataset was too small (< 40.000 cells) to find enough cells for a new class for training the AI-model.

Hence, all granulocytes were put into one class and were not further differentiated, although there is evidence for eosinophiles and neutrophiles and heterophiles in fish. Blasts and macrophages were not found either, although these mature immune cells were documented in fish (PeuB et al., 2020).

In total, the staining allowed a confident and successful identification of 6 cell classes which also reflects findings concerning identifications of peripheral blood cells of fish in other publications. For instance, in Chinook salmon for instance, erythrocytes, lymphocytes, thrombocytes, monocytes and neutrophils had been found. (Lulijwa et al., 2019)

5.4 AI quantification and cell classifying

AI for cell counting was selected, because this method offered the opportunity to analyze a large dataset without typical human biases. A human being is exposed to subjective impression and losses of concentration while an artificial intelligence can recognize patterns with the ability to objectively interpret images on pixel level, which is nearly invisible to the human eye. (LeCun et al., 2015)

The results delivered by the trained deep-learning AI model of Biodock™ could be validated with the approach to count a random subset of images (N=18 and 970 cells). The AI models count only deviates 0.2 % from the manual counting, when comparing the relative abundance of blood cells, with the calculated result from manual counting. This is the mean difference of the AI counting vs the manual counting. Nevertheless, for lymphocytes, immature erythrocytes and granulocytes, this statistic is not conclusive, as a large divergence in absolute numbers between both counting methods was shown. The granulocytes divergence is explainable because only few granulocytes (11) were found for training. Also due to their polymorph characteristics such as divers segmented nucleus and irregular distributed granola, the deep learning segmentation was complicated. One reason for the divergence in the lymphocytes is the difficulty of distinguishing from thrombocytes due to the similar color and shape. Therefore, conclusions for my hypothesis should be used with caution.

5.5 Hypothesis conclusions

5.5.1 Immune cell abundance

No significant differences were found in the abundance of blood cells between the two ecotypes Pachón cavefish and surface fish of A. mexicanus, even not in the abundance of immune cells. The M/L ratio specifically shows that there is no significant shift in the immune cell composition concerning the ratio between mono-myeloid and lymphoid immune cells in both ecotypes. Also no significant difference in the abundance of blood cells were found between male and female fish.

Therefore, the hypothesis, which had prognosed a manifestation of the adapted immune system in the cavefish ecotype that could have revealed a lower myelomonocyte/lymphocyte ratio of immune cells in A. mexicanus cavefish compared to the surface fish, could not be supported.

Reasons for that conclusion could be the hygienic environment of the laboratory and hence a healthy infection status of the fish. Whereas in wild populations, the species is exposed to a richer parasite abundance and therefore a higher infection risk. In that case, also a shift in the immune system strategy was found in the blood cell composition of the Pachón ecotype (PeuB et al., 2020).

5.5.2 Erythrocytes’ size

The hypothesis that the erythrocytes size is dependent to the ecotype of A. mexicanus can be reinforced. Pachón cavefish have on average significantly 22 % larger erythrocytes than the surface ecotype of A. mexicanus. Due to ongoing hypoxia in the cave milieu, there is a strong evolutionary pressure to an efficient O2 and CO2 transportation. The larger the area, the shorter is the diffusion distance for oxygen and carbon dioxide. (Boggs et al., 2022)

In comparison with that publication, the findings in this thesis show a smaller average erythrocytes area for both ecotypes. This indicates weaknesses of the segmentation by the AI model, which had probably not recognized the outer pixel locating at the edge of a segmented object. (Kirillov et al., 2023)

6 Conclusion

Due to severe evolutionary pressures affecting genes regulating physiology and morphology of A. mex- icanus, it was expected to measure a significant shift in the immune system strategy towards a more sensitive adapted response.

However, in this thesis, an expression of the phenotypical manifestation of genetical changes favoring the lymphoid strain of hematopoiesis for the immune response could not be shown in healthy, lab-raised A. mexicanus.

Weaknesses in the staining protocol could have led to mistakes in the segmentation process, which could ultimately have led to mistakes concerning the immune cell count. Because of the small dataset (< 40,000 cells) already a small amount of wrongly classified cells could have affected the M/L ratio, which determines the focus of the immune system.

Significant differences however were found concerning the erythrocytes size between the two ecotypes, supporting the hypothesis that hypoxia, which is also measured in cave waters, represents a strong selective pressure towards larger erythrocytes.

7 Outlook

Further field investigations could also include richness and prevalence of microparasites like bacteria and viruses to describe a complete picture of the parasite abundance on and in A. mexicanus cavefish of different ecotypes. This would empirically indicate evidence to the theoretical framework, that parasite diversity losses correlate with biodiversity losses (Lafferty, 2012). This would give crucial premises for evaluating the focus of the immune system's response in cavefish as an mathematical framework predicts, that establishing a shift towards an adapted immune system response is evolutionary beneficial for a species (Mayer et al., 2016). To find evidence for this, a broad field investigation could be expanded on many other Astyanax mexicanus ecotypes inhabiting separated caves of the same area, to assess the convergence concerning their immunological niche conformance. Also, including infected lab-raised cavefish to the investigation could verify shifts.

AI seems to be a convenient method for the analysis of blood smears if trained bias-free. Larger datasets however and precise staining will be necessary to get trustful results. For example, blood smears were stained using an automatic stainer and were analyzed with AI using a dataset of more than 4.000.000 cells in a recent study concerning leukemia (Kockwelp et al., 2024).

8 Appendix

Table 7: Overview of all slides being stained and the belonging biometric data. The index number on the left delivers the respective individual number of the fish. In total 22 fish were dissected. SF: A. mexicanus surface form, CF: A. mexicanus cave form

Illustrations are not included in the reading sample

Illustrations are not included in the reading sample

Figure 16: Typical blood cell layer of a sample of A. mexicanus. Photographed with 40x objective according to section 3.5. The Sigmal Aldrich protocol was applied, but with the PBS incubation at pH = 7.2 in between the two-dye-stain. Notice the dark purple, small round cells (=thrombocytes); the slightly bigger cells and with an still recognizable, light blue cytoplasm and a large purple nucleus (=lymphocytes); and the largest cells in that image having a greyish cytoplasm and purple nucleus (=erythrocytes)

Illustrations are not included in the reading sample

Figure 17: Image shows mistakes in the duration of the rinsing at the end of the staining procedure. As slide was incubated for longer than a minute in PBS, the stain of the nuclei started to get washed out.

Illustrations are not included in the reading sample

Figure 18: Comparision between blood cells stained with a dilution of 1:20 of Giemsa (left) and with a dilution of 1:10 of Giemsa (right). The other staining parameters were performed according to teh Sigma Aldrich protocol. Noticable is however is the high resolution of the erythrocytes nuclei in the right image.

Illustrations are not included in the reading sample

Figure 19: Blood cell stain when slides were incubated in Giemsa (c = 1:20) stain for 20min instead of 15 min. The Sigmal Aldrich protocol was applied, but Incubation in PBS pH=6.8 was performed. The dark chromatin appeared in a remarkable high resolution inside the purple nuclei of the erythrocytes. That effect is also documented in publications (Binder et al., 2012).

Illustrations are not included in the reading sample

Figure 20: Blood cells stained accoding to the Sigma Aldrich protocole but the slides were incubated in PBS, pH= 7.2 (left) and PBS pH= 6.8 (right) between the two-dye May-Grünwald Giemsa stain. The cytoplasm of erythrocytes appear reddish with a pH =6.8, while incubate in a PBS pH= 7.2, the cytoplasm greyish-green.

Table 8: Description of the subsets for analyzing the Erythrocytes area.

Illustrations are not included in the reading sample

Illustrations are not included in the reading sample

Figure 21: A snapshot of the diagram provided of the Biodock program points out the segmentation results concerning erythrocytes of surface male fish. There is perimeter on the x-axis and area on the y- axis. Units are pixel. Correlation between area and perimeter is obvious. Objects with a perimeter above 945 pixel and with an area below 12500 pixel were excluded later in R.

9 Acknowledges

I would like to thank my supervisor Dr. Robert Peuß and Prof. Dr. Joachim Kurtz for giving me the opportunity to write my bachelor’s thesis at the Institute for Evolution and Biodiversity at the University Münster and for organizing the frame.

I am incredibly grateful for the allowance to conduct an experimental thesis honoring their trust for giving me the freedom in the lab. I really appreciated that my mentor and supervisor Dr. Robert Peuß spent his time for periodical meetings for consulting me about my thesis.

I thank the technical staff Ilka, Kathrin, Anke and Julia, who for their patience and time to answer question regarding the lab and the maintaining of the microscope.

During my time at the Kurtz Group, the PhD student Marc Bauhus also guided me. I really appreciated his constant support especially to always get prompt answers in urgent times of questions and his advises concerning the method. Special thanks also go to Dr. Peter Czuppon, as he let me participate in his course “Introduction in Statistics” and for taking care of the coffee machine. Also, I am indebted to Dr. rer. nat. Thomas Zobel, that he showed up and helped a fellow student and me with the image processing.

And last but not least, I would like to thank Ulla Pebesma and Michelle Borgers for their support during the sample collection. Without Michelle Borgers, I would have not had fish for my thesis.

In total, I really appreciated being a member of the Kurtz group. There was so much sympathy, cooperativeness, enthusiasm, and an overall positive atmosphere that really helped me to endure this time and makes me miss the weekly team/lab meeting.

A big thank goes to my sister Barbara who took time for questions, motivation, encouragement, and proofreading.

10 References

Aspiras, A. C.; Rohner, N.; Martineau, B.; Borowsky, R. L.; Tabin, C. J. Melanocortin 4 receptor mutations contribute to the adaptation of cavefish to nutrient-poor conditions. Proc. Natl. Acad. Sci. U.S.A. 2015, 112 (31), 9668-9673. DOI: 10.1073/pnas.1510802112.

Binder, T.; Diem, H.; Fuchs, R.; Gutensohn, K.; Nebe, T. Pappenheim-Färbung: Beschreibung einer hämatologischen Standardfärbung - Geschichte, Chemie, Durchführung, Artefakte und Problemlösungen: Pappenheim Stain: Description of a hematological standard stain - history, chemistry, procedure, artifacts and problem solutions. J. Lab. Med. 2012, 36 (5), 293-309. DOI: 10.1515/lab- med-2012-0027.

Biodock. Biodock, 2023.

Biology and Evolution of the Mexican Cavefish; Keene, A. C., Yoshizawa, M., McGaugh, S. E., Eds.; Academic Press, 2016.

Boggs, T. E.; Friedman, J. S.; Gross, J. B. Alterations to cavefish red blood cells provide evidence of adaptation to reduced subterranean oxygen. Sci. Rep. 2022, 12 (1), 3735. DOI: 10.1038/s41598- 022-07619-0.

Burton, Erin. Module 6: Blood Smear Technique and Reticulocyte Counting: Module 6.2: Blood Smear Procedure. https://pressbooks.umn.edu/cvdl/chapter/module-7-2-blood-smear-procedure/ (accessed January 12, 2024).

Culver, D. C.; Pipan, T. Sources of Energy in Subterranean Environments. In The Biology of Caves and Other Subterranean Habitats; Culver, D. C., Pipan, T., Eds.; Oxford University Press, 2019; pp 24-42. DOI: 10.1093/oso/9780198820765.003.0002.

Espinasa, L.; Bonaroti, N.; Wong, J.; Pottin, K.; Queinnec, E.; Rétaux, S. Contrasting feeding habits of post-larval and adult Astyanax cavefish. Subt. Biol. 2017, 21, 1-17. DOI: 10.3897/sub- tbiol.21.11046.

Espinasa, L.; Pech, A. Biogeographical affinities of the aquatic community of Refugio Cave, a newly discovered Astyanax cave. Subt. Biol. 2023, 46, 77-86. DOI: 10.3897/subtbiol.46.102043.

Fumey, J.; Hinaux, H.; Noirot, C.; Thermes, C.; Rétaux, S.; Casane, D. Evidence for late Pleistocene origin of Astyanax mexicanus cavefish. BMC evolutionary biology 2018, 18 (1), 43. DOI: 10.1186/s12862-018-1156-7.

Gross, J. B.; Furterer, A.; Carlson, B. M.; Stahl, B. A. An integrated transcriptome-wide analysis of cave and surface dwelling Astyanax mexicanus. PloS one 2013, 8 (2), e55659. DOI: 10.1371/jour- nal.pone.0055659.

Horobin, R. W. How Romanowsky stains work and why they remain valuable - including a proposed universal Romanowsky staining mechanism and a rational troubleshooting scheme. Biotech. Histo- chem. 2011, 86 (1), 36-51. DOI: 10.3109/10520295.2010.515491.

Jeffery, W.; MARTASIAN, D. Evolution of Eye Regression in the Cavefish Astyanax: Apoptosis and the Pax-6 Gene. Am. Zool. 1998, 38 (4), 685-696. DOI: 10.1093/icb/38.4.685.

Kirillov, Alexander, Mintun, Eric, Ravi, Nikhila, Mao, Hanzi, Rolland, Chloe, Gustafson, Laura, Xiao, Tete, Whitehead, Spencer, Berg, Alexander C., Lo Wan-Yen, Dollar, Piotr, Girshick, Ross. Segment Anything: arXiv:2304.02643, 2023.

Kling, J. Insulin resistance grows in the dark. Lab Anim. 2018, 47 (6), 152. DOI: 10.1038/s41684- 018-0075-y.

Kockwelp, J.; Thiele, S.; Bartsch, J.; Haalck, L.; Gromoll, J.; Schlatt, S.; Exeler, R.; Bleckmann, A.; Lenz, G.; Wolf, S.; Steffen, B.; Berdel, W. E.; Schliemann, C.; Risse, B.; Angenendt, L. Deep learning predicts therapy-relevant genetics in acute myeloid leukemia from Pappenheim-stained bone marrow smears. Blood advances 2024, 8 (1), 70-79. DOI: 10.1182/bloodad- vances.2023011076.

Kondera, E. Haematopoiesis and haematopoietic organs in fish. Anim. Sci. Genet. 2019, 15 (1), 9-16. DOI: 10.5604/01.3001.0013.4535.

Krishnan, J.; Persons, J. L.; PeuB, R.; Hassan, H.; Kenzior, A.; Xiong, S.; Olsen, L.; Maldonado, E.; Kowalko, J. E.; Rohner, N. Comparative transcriptome analysis of wild and lab populations of As- tyanax mexicanus uncovers differential effects of environment and morphotype on gene expression. J. Exp. Zool. (Mol. Dev. Evol.) 2020, 334 (7-8), 530-539. DOI: 10.1002/jez.b.22933.

Lafferty, K. D. Biodiversity loss decreases parasite diversity: theory and patterns. Phil. Trans. R. Soc. B 2012, 367 (1604), 2814-2827. DOI: 10.1098/rstb.2012.0110.

LeCun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 521 (7553), 436-444. DOI: 10.1038/nature14539.

Lieschke, G. J.; Trede, N. S. Fish immunology. Curr. Biol. 2009, 19 (16), R678-82. DOI: 10.1016/j.cub.2009.06.068.

Lulijwa, R.; Alfaro, A. C.; Merien, F.; Burdass, M.; Young, T.; Meyer, J.; Nguyen, T. V.; Trembath, C. Characterisation of Chinook salmon (Oncorhynchus tshawytscha) blood and validation of flow cytometry cell count and viability assay kit. Fish Shellfish Immunol. 2019, 88, 179-188. DOI: 10.1016/j.fsi.2019.02.059.

Maldonado, E.; Rangel-Huerta, E.; Rodriguez-Salazar, E.; Pereida-Jaramillo, E.; Martinez-Torres, A. Subterranean life: Behavior, metabolic, and some other adaptations of Astyanax cavefish. J. Exp. Zool. (Mol. Dev. Evol.) 2020, 334 (7-8), 463-473. DOI: 10.1002/jez.b.22948.

Mayer, A.; Mora, T.; Rivoire, O.; Walczak, A. M. Diversity of immune strategies explained by adaptation to pathogen statistics. Proc. Natl. Acad. Sci. U.S.A. 2016, 113 (31), 8630-8635. DOI: 10.1073/pnas.1600663113.

Medley, J. K.; Persons, J.; Biswas, T.; Olsen, L.; PeuB, R.; Krishnan, J.; Xiong, S.; Rohner, N. The metabolome of Mexican cavefish shows a convergent signature highlighting sugar, antioxidant, and Ageing-Related metabolites. eLife 2022, 11. DOI: 10.7554/eLife.74539.

Miranda-Gamboa, R.; Espinasa, L.; Verde-Ramírez, M.; Hernández-Lozano, J.; Lacaille, J. L.; Espi- nasa, M.; Ornelas-García, P. A new cave population of Astyanax mexicanus from Northern Sierra de El Abra, Tamaulipas, Mexico. Subt. Biol. 2023, 45, 95-117. DOI: 10.3897/subtbiol.45.98434.

Moldovan, O. T.; Kovác, E.; Halse, S. Cave Ecology; Ecological Studies 235; Springer Nature Switzerland AG: Cham, 2018.

Mulisch, M.; Welsch, U. Romeis -Mikroskopische Technik, 19. Auflage; Springer-Verlag: Berlin Heidelberg, 2015.

nomadseason. Monthly climate in El Abra, Tamaulipas, Mexico. https://nomadseason.com/climate/ mexico/tamaulipas/el-abra.html (accessed January 27, 2024).

Pacioglu, O.; Tusa, I. M.; Popa, I.; Itcus, C.; Plavan, G.-I.; Boufahja, F.; Baba, §. C. Aquatic subterranean food webs: A review. Glob. Ecol. Conserv. 2023, 48, e02704. DOI: 10.1016/j.gecco.2023.e02704.

Patil, I. Visualizations with statistical details: The ggstatsplot approach; Journal of Open Source Software, 2021.

PeuB, R.; Box, A. C.; Chen, S.; Wang, Y.; Tsuchiya, D.; Persons, J. L.; Kenzior, A.; Maldonado, E.; Krishnan, J.; Scharsack, J. P.; Slaughter, B. D.; Rohner, N. Adaptation to low parasite abundance affects immune investment and immunopathological responses of cavefish. Nature Ecol. Evol. 2020, 4 (10), 1416-1430. DOI: 10.1038/s41559-020-1234-2.

Piaton, E.; Fabre, M.; Goubin-Versini, I.; Bretz-Grenier, M.-F.; Courtade-Saïdi, M.; Vincent, S.; Belle- année, G.; Thivolet, F.; Boutonnat, J.; Debaque, H.; Fleury-Feith, J.; Vielh, P.; Egelé, C.; Bellocq, J.-P.; Michiels, J.-F.; Cochand-Priollet, B. Guidelines for May-Grünwald-Giemsa staining in haematology and non-gynaecological cytopathology: recommendations of the French Society of Clinical Cytology (SFCC) and of the French Association for Quality Assurance in Anatomic and Cytologic Pathology (AFAQAP). Cytopathology 2016, 27 (5), 359-368. DOI: 10.1111/cyt.12323.

Poulson, T. L. Food Sources. In Encyclopedia of Caves (Second Edition); White, W. B., Culver, D. C., Eds.; Elsevier, 2012; pp 323-334. DOI: 10.1016/B978-0-12-383832-2.00045-1.

Protas, M.; Conrad, M.; Gross, J. B.; Tabin, C.; Borowsky, R. Regressive Evolution in the Mexican Cave Tetra, Astyanax mexicanus. Curr. Biol. [Online] 2007, 17 (5), 452-454. https://www.sci-encedirect.com/science/article/pii/S0960982207008901.

R Core Team. R: A language and environment for statistical computing; R Foundation for Statistical Computing, 2018.

Rohner, N. The cavefish Astyanax mexicanus. Nat. Methods 2023, 20 (7), 948-950. DOI: 10.1038/s41592-023-01916-w.

Santacruz, A.; Hernández-Mena, D.; Miranda-Gamboa, R.; Pérez-Ponce de León, G.; Ornelas-García, C. P. Host-parasite interactions in perpetual darkness: Macroparasite diversity in the cavefish Asty- anax mexicanus. Zool. Res. 2023, 44 (4), 782-792. DOI: 10.24272/j.issn.2095-8137.2022.376.

Traver, D.; Herbomel, P.; Patton, E. E.; Murphey, R. D.; Yoder, J. A.; Litman, G. W.; Catic, A.; Amemiya, C. T.; Zon, L. I.; Trede, N. S. The zebrafish as a model organism to study development of the immune system. Adv. Immunol. 2003a, 81, 253-330.

Traver, D.; Paw, B. H.; Poss, K. D.; Penberthy, W. T.; Lin, S.; Zon, L. I. Transplantation and in vivo imaging of multilineage engraftment in zebrafish bloodless mutants. Nat. Immunol. 2003b, 4 (12), 1238-1246. DOI: 10.1038/ni1007.

Urry, L. A.; Cain, M. L.; Wasserman, S. A.; Minorsky, P. V.; Reece, J. B. Campbell Biologie, 11., aktualisierte Auflage; Studium Biologie; Pearson: Hallbergmoos, 2019.

van der Weele, C. M.; Jeffery, W. R. Cavefish cope with environmental hypoxia by developing more erythrocytes and overexpression of hypoxia-inducible genes. eLife 2022, 11. DOI: 10.7554/eLife.69109.

Weber, J. N.; Steinel, N. C.; Peng, F.; Shim, K. C.; Lohman, B. K.; Subramanian, S.; De Lisle, S. P.; Bolnick, D. I. Evolutionary gain and loss of a pathological immune response to parasitism. Science 2022, 377, 1206-1211.

Wilson, E. J.; Tobler, M.; Riesch, R.; Martínez-García, L.; García-De León, F. J. Natural history and trophic ecology of three populations of the Mexican cavefish, Astyanax mexicanus. Environ. Biol. Fish 2021, 104 (11), 1461-1474. DOI: 10.1007/s10641-021-01163-y.

Xiong, S.; Krishnan, J.; Peuß, R.; Rohner, N. Early adipogenesis contributes to excess fat accumulation in cave populations of Astyanax mexicanus. Dev. Biol. 2018, 441 (2), 297-304. DOI: 10.1016/j.ydbio.2018.06.003.

Frequently asked questions

What is the study about?

The study analyzes the blood cell composition of the cave-dwelling and surface-dwelling forms of the Astyanax mexicanus fish, and compares the erythrocyte size between these ecotypes. It investigates potential adaptations related to the immune system and oxygen transport in cavefish.

Where does the Pachón cavefish live?

The Pachón cavefish inhabits a natural subterranean freshwater network in the Pachón cave, located in the northern part of the Sierra El Abra region in Northeast Mexico.

What are the key characteristics of the Pachón cave environment?

The Pachón cave environment is characterized by perpetual darkness, low nutrition availability (oligotrophic), and low-oxygenized waters.

What are some adaptations that the Astyanax mexicanus cavefish have made to thrive?

Some morphological adaptations are the eye regression and the loss of pigmentation. Physiological adaptations include altered glucose metabolism, hyperphagia, increased fat storage, and more efficient oxygen transport.

How is the immune system of vertebrates divided?

The immune system of vertebrates is divided into the innate and adapted immune responses.

What were the main hypotheses tested in the study?

The two main hypotheses were: 1. There are significant differences in the blood cell compositions between the Pachón cavefish and the surface fish ecotypes. 2. The size of erythrocytes is significantly larger in Pachón cavefish compared to surface fish.

Where did the fish used in the study come from?

All fish were offspring of populations originating in the Pachón cave or surface waters and were bred and kept under controlled laboratory conditions in Münster.

How were the blood samples collected and prepared?

Blood was taken from the caudal vein after euthanizing the fish. Blood smears were generated on slides, air-dried, and fixed in ethanol before staining.

What staining method was used?

The May-Grünwald/Giemsa stain (Pappenheim method) was used to stain the blood smears.

How was image analysis conducted?

Light microscopy was used for generating images, which were analyzed with a trained deep-learning AI-model by Biodock for image segmentation and cell counting.

What blood cell types were identified and analyzed?

The main blood cell types identified and analyzed were erythrocytes, immature erythrocytes, thrombocytes, lymphocytes, monocytes, and granulocytes.

What statistical methods were used to analyze the data?

ANOVA tests, Shapiro-Wilk tests, Levene tests, and Welch's t-tests were performed to determine significant differences between ecotypes, sexes, and erythrocyte size.

What were the main results regarding blood cell composition?

No significant differences in blood cell composition were measured between the two ecotypes or between sexes. There was also no significant shift in the immune systems strategy that could be proved.

What were the main results regarding erythrocyte size?

Erythrocytes from cavefish were significantly larger (22%) on average compared to the surface form, indicating that hypoxia seems to be a strong selective pressure for the cavefish.

What are the potential reasons for not finding a significant difference in blood cell composition?

The hygienic environment of the laboratory may have influenced the immune systems of the fish, compared to the wild state. Weaknesses in the staining protocol could have led to segmentation errors, which could ultimately have led to mistakes concerning the immune cell count.

What does M/L ratio stand for?

The M/L ratio is calculated as monocytes plus granulocytes per number of lymphocytes. It may indicate a shift in the immune cell composition.

What future research directions are suggested?

Further field investigations could examine microparasite richness and prevalence to develop a clearer picture of parasite abundance. Also, the investigations should be expanded to many other Astyanax mexicanus ecotypes inhabiting separate caves to assess their immunological conformance. Including infected lab-raised cavefish in the investigation might also verify shifts in the immune system.