This manuscript describes the setup, implementation, and analysis of boldness, aggression, and shoaling in zebrafish and testing for the presence of a behavioral syndrome. A standardized approach for behavioral quantification will allow for easier comparison across studies. Modifications to this protocol are possible as each assay can be run individually.
A behavioral syndrome exists when specific behaviors interact under different contexts. Zebrafish have been test subjects in recent studies and it is important to standardize protocols to ensure proper analyses and interpretations. In our previous studies, we have measured boldness by monitoring a series of behaviors (time near surface, latency in transitions, number of transitions, and darts) in a 1.5 L trapezoidal tank. Likewise, we quantified aggression by observing bites, lateral displays, darts, and time near an inclined mirror in a rectangular 19 L tank. By dividing a 76 L tank into thirds, we also examined shoaling preferences. The shoaling assay is a highly customizable assay and can be tailored for specific hypotheses. However, protocols for this assay also must be standardized, yet flexible enough for customization. In previous studies, end chambers were either empty, contained 5 or 10 zebrafish, or 5 pearl danios (D. albolineatus). In the following manuscript, we present a detailed protocol and representative data that accompany successful applications of the protocol, which will allow for replication of behavioral syndrome experiments.
There is a growing body of literature investigating the associations between distinct behaviors within individual animals from a given population. These associations are termed behavioral syndromes, and the measurements typically include boldness, aggression, exploratory behavior, and sociability1-5. Behavioral syndromes are valuable for both direct and indirect reasons. Directly, knowledge of behavioral syndromes can provide a more complete view of evolutionary theory, population structure, and population dynamics3. Indirectly, knowledge regarding behavioral associations may inform fields that quantify behavior such as pharmacology6 , toxicology7, behavioral genetics8,9, and endocrinology10. Because of these direct and indirect benefits, an increased knowledge of behavioral syndromes is especially valuable in commonly used model organisms such as the zebrafish. Studies using zebrafish are found in a variety of disciplines, including the analysis of behavioral syndromes11-13. To advance knowledge in behavioral syndrome research, and because other disciplines also use behavioral measurements for hypothesis testing, reliable and succinct descriptions of behavior are required for valid analyses and interpretations and standardized protocols will facilitate inter-study comparisons within species. Our protocol was developed to measure a boldness-aggression-shoaling behavioral syndrome in a population of lab-reared zebrafish14. However, the basis of the protocol (tracking individual fish, ensuring proper randomization, and appropriate analyses) can be easily modified for a variety of alternative behavioral measures. Additionally, boldness, aggression, or shoaling assays can be run individually for the testing of distinct hypotheses. Therefore, while it is our goal to describe how to conduct a behavioral syndrome study and the protocol for successful individual level behavioral measurement, each facet of this procedure can stand alone.
The literature on behavioral syndromes spans several taxonomic groups, from arthropods to humans4 and, in order to measure a behavioral syndrome, at least two behavioral contexts must be quantified. Unfortunately, there is often little consistency in the assays that are used to quantify the behavioral measurement across the axes of behavior. For example, in fish, boldness may be measured using T-maze assays, open-field assays, or introduction of a novel or foreign stimulus15. Aggression studies in fish might involve dyad interactions, video stimulus assays, or clay model assays12,16,17. Likewise, analysis of shoaling behavior, which typically involves the measurement of shoalmate preference, may be performed in different types of tanks, with different methods to determine association time21-23. In this protocol a specific subset of the overall behavioral assay repertoire is presented. Specifically, this protocol presents a methodology to track individuals through boldness, aggression, and shoaling assays in a way that facilitates comparisons within individuals to determine whether the comparisons are consistent across all individuals within a population. We have performed this protocol with zebrafish and convict cichlids (Amatitlania nigrofasciata) in previous studies14,18, and it will work with any similarly sized freshwater fish.
Boldness assays are conducted in a 1.5 L trapezoidal tank that has a horizontal line delineating equal sized areas in the upper and lower portions of the tank. Quantified behaviors include the number of transitions by the test fish between the upper and lower portion of the tank, the time spent in each portion, the number of darts, and the latency to enter the upper portion. The aggression assay is performed in a 19L rectangular tank that includes a 3 inch x 5 inch mirror inclined at about 22° situated in the lower left corner of the tank19. Quantified behaviors include the total amount of time spent by the target fish interacting with the mirror20, along with specific aggressive indicators – number of bites, lateral displays, darts between the test fish and its reflection. For these specific indicators, bites are defined as quick lunges toward the mirror with an open mouth, lateral displays are defined as the flaring of lateral, pectoral, anal and dorsal fins in the direction of the mirror, and darts are any erratic movements that are not directed toward the mirror. Lastly, the shoaling assay quantifies the behavior of a test fish in center chamber of a tri-chambered tank. The side chambers of the tank are either empty, or contain a "target shoal" of fish, and the time the test fish spends near each side chamber is measured21-23. A single composite score, referred to as Strength of Shoaling (SoS), is calculated for each individual test fish, specific to the stimuli, and can be used in downstream analyses14. All behaviors are scored by a single viewer, or multiple viewers using free behavioral quantification software known as JWatcher24.
Testing the presence of a behavioral syndrome is primarily a statistical endeavor, and it is advisable to follow the guidelines as presented by Budeav 201025. Specifically, it is recommended to perform a principal components analysis (PCA) on centered and normed data in which the inputs are the vectors of an individual's behaviors in assays with multiple behavioral measurements (i.e., boldness and aggression). The PCA, performed on a correlation matrix, reduces the dimensionality of the behavioral measurements, and thus extracts the most important knowledge that explains a majority of the variation. The extracted components can then be interpreted based on high factor loadings for the individual behavior of interest and regression scores can be extracted for each individual on the basis of the explanatory components. These regression scores can then be compared to the SoS measurement and other various non-behavioral measurements such as fish size or sex.
This workflow has been implemented in a study of zebrafish behavioral syndromes in which a sex specific behavioral syndrome that exists between boldness and shoaling14 was discovered. In this situation, bolder zebrafish males are more likely to associate with a larger, more aggressive species (D. albolineatus), but this association is lost in females. This workflow was also implemented in a study of juvenile convict cichlid kin (Amatitlania nigrofasciata)18 in which a behavioral syndrome was not discovered, potentially indicating behavioral plasticity of the species. Therefore, the following protocol is presented with a goal of delineating the nature of three specific assays (boldness, aggression, and shoaling) in the framework of studying an individual level behavioral syndrome.
The following methodologies for the housing, care, and study of zebrafish have been approved by the Saint Joseph's University IACUC.
1. Zebrafish Housing and Care
2. Randomization and Tank Setup
3. Conducting the Aggression Assay
4. Conducting the Boldness Assay
5. Conducting the Shoaling Assay
6. Data Quality Control
7. Analyze Data
Depending on the nature of the study, and specific protocol employed, several distinct results are possible in a behavioral syndromes experiment. The following tables and figures, where indicated, are adapted from our previous study published in the journal Behavioural Processes14 and the journal Zebrafish17. When the proposal (as described above) is carried out in its entirety, two sets of results, 'within assay correlations' and 'between assay correlations,' are expected (Figure 1).
The first set of results describes within assay consistency. Specifically, the results describe how behaviors are correlated with each other within the boldness assay and the within the aggression assay separately. Tables 1 and 2 describe what these correlations should look like for the boldness and aggression assays and the data presented are adapted from a previous study14. For the boldness assay, it is expected that number of transitions and time spent in the upper portion of the tank are positively correlated, but latency to enter the upper portion is negatively correlated. Representative results for individuals measured on the boldness axis of behavior in the boldness tank (see step 4.1) are given in Table 1. For the aggression assay, it is expected that bites, lateral displays, and time spent near the mirror are all correlated. Representative results for individuals measured on the aggression axis of behavior in the aggression tank (see step 3.1) are given in Table 2. Darts are not usually correlated with aggressive behavior although, in some cases, nonaggressive fish tend to dart (hence a negative correlation). Lastly, a Strength of Shoaling (SoS) measurement for the shoaling assay provides an individual-level measure of shoaling tendency. Shoaling is a highly repeatable behavior across a variety of shoaling assays when measured using SoS (ICC = 0.641)14 and several of our previous studies have confirmed shoaling behavior in zebrafish23,38. Since all individuals will be measured for each of the behaviors, we can calculate Spearman correlations across these behaviors and expect some behaviors to be correlated, as identified for zebrafish in the representative results.
The second portion of the results investigates the presence of a behavioral syndrome. The results of the integrative analysis of all behaviors quantified in the boldness and aggression behavioral assays are summarized using a Principal Components Analysis (PCA). There should be two to four interpretable components based on high loadings for each behavior. If a single eigenvector that explains a good portion of the variation includes behaviors from both the boldness and aggression assays, then a behavioral syndrome has been observed. However, if behaviors from the boldness and aggression assays do not overlap on a single eigenvector, then the study does not describe a behavioral syndrome. Representative results are presented for the absence of a boldness-aggression behavioral syndrome (Figure 2). In this example, component 1 is most strongly representative of aggression behaviors while component 2 is most strongly associated with boldness behaviors. Because aggression and boldness behaviors are not represented by the same component, it can be concluded that there is the absence of a boldness-aggression behavioral syndrome (boldness vectors are roughly orthogonal to aggression vectors). Also, these behaviors are not influenced by sex because the distribution pattern is the same between males and females. To observe if boldness or aggression behaviors are associated with shoaling, the extracted regression scores for each fish are correlated with the fish's SoS measurement. It is important to ensure that any correlating components are actually describing an interpretable set of behaviors. If there are high absolute value Spearman correlations between any component and the shoaling measurement, then a shoaling behavioral syndrome is present. Regardless of the specific results, it is important to provide feasible biological interpretations to all observations. Representative results are provided for a description of a boldness-shoaling behavioral syndrome (Table 3)14. While the data presented represent typical results for zebrafish, they should not be interpreted as measurements of data quality. There are several reasons why the data will look different across different experiments such as species or population differences.
Figure 1. Proposed workflow for a boldness, aggression, and shoaling behavioral syndromes experiment. Typical measurements are listed for each assay. An outline of how to analyze the assays individually and then how to integrate the analyses is also listed. Please click here to view a larger version of this figure.
Figure 2. Representative results of a PCA indicating that a boldness-aggression behavioral syndrome is absent. The red vectors indicate behaviors which contribute to the component scores (left and bottom axes) and the points represent individual scores along each component (right and top axes). Males are represented by blue points and females are represented by green points. Note that 'erratic behavior' refer to 'darts' and 'aggression rate' can be computed by adding the number of bites and lateral displays divided by the total time interacting with the mirror. Please click here to view a larger version of this figure.
Transitions | Latency | Darts | ||||
ᴩ | p | ᴩ | p | ᴩ | p | |
Surface Time | 0.65 | <0.001* | -0.68 | <0.001* | 0.42 | 0.007* |
Transitions | -0.40 | 0.012* | 0.33 | 0.041* | ||
Latency | -0.47 | 0.002 | ||||
*α <0.05 |
Table 1: Boldness Spearman rank correlations in a novel tank assay. This table had been modified from Way et al., 2015.
Lateral Displays | Darts | Mirror Time | ||||
ᴩ | p | ᴩ | p | ᴩ | p | |
Bites | 0.69 | <0.001* | -0.10 | 0.547 | 0.63 | <0.001* |
Lateral displays | -0.06 | 0.735 | 0.66 | <0.001* | ||
Darts | -0.18 | 0.285 | ||||
*α <0.05 |
Table 2: Aggression Spearman rank correlations in a novel tank assay. This table had been modified from Way et al., 2015.
ᴩ | p | |
PCA Components | ||
Boldness | 0.339 | 0.035* |
Aggression | 0.075 | 0.65 |
Dart | -0.012 | 0.944 |
Table 3: Using the PCA components to confirm the presence of a shoaling and boldness behavioral syndrome present in this population. This table had been modified from Way et al., 2015.
The protocol will determine if there are consistent associations in boldness, aggression, and shoaling behaviors in zebrafish. If there are consistent associations in a given population between any of these behaviors, then a behavioral syndrome is present. By studying a population's natural behavioral syndrome, researchers can have a more complete understanding of its behavioral dynamic, population structure, and possibly evolutionary history3. Furthermore, manipulating the environment that affects these behavioral syndromes, like in pharmacology6, toxicology7, behavioral genetics8,9, and endocrinology10 studies, can shed light on how different factors might affect behavioral associations. The specific composition of the aggression-boldness-shoaling behavioral syndrome found in this protocol encompasses common behaviors well studied in the model organism zebrafish11-13. While this methodology can be applied to any fish species, the wide application for using zebrafish may help with the interdisciplinary standardization of this protocol. Each assay described above can be more easily controlled and manipulated than other comparable techniques. For example, the aggression assay uses the inclined mirror to elicit a wider array of aggression responses rather than using a live or clay model, video stimulus, or vertical mirror image stimulus17. The behaviors measured in boldness assay are more easily quantified than in a t-maze or large open field test15. Lastly, the shoaling assay described can easily track an individual, which is essential for studying behavioral syndromes at the individual level, and can allow for unlimited combinations of target shoals to ask novel experimental questions. While this protocol can be altered to accommodate a variety of behavioral questions, the standardization of it may lead to further insight on how DNA, cellular mechanisms, cellular development, and introduced chemicals can affect individual and population behavioral syndromes.
In order to be confident that the observed associations are true, the protocol must be followed carefully. For the most part, the protocol is modifiable according to the specific hypothesis; however, there are several steps that must be performed to ensure confidence in the results. First, it is important that the individuals are housed in an appropriate tank, at optimal conditions (see protocol) and are fed properly. The water conditions should be consistent in housing racks, acclimation tanks, and test tanks. If the water is not consistent, the fish will take longer to acclimate to the conditions, and the behavioral measurements will not be properly captured. It is also important that water is recycled between boldness and aggression tanks so that artificial odors are not carried over between assays. Second, it is of utmost importance that all behaviors are scored properly. To ensure this, all assays are videotaped to allow for multiple viewings, and for several training opportunities. A single trained scorer can accurately measure fish behavior, but training does take some time. If a careful training period is ignored, then the confidence of the results of the behavioral study is low. A careful study with low sample sizes performed by a well-trained observer is more powerful than a study with high sample size performed by a poorly trained observer. Lastly, quality control of the collected data is also critical. The steps that are taken to randomize individuals, and remove biases should result in useable, reliable data, and data transformations are recommended for preprocessing according to the specific statistical test (see protocol). If these steps are not carefully done, the output of the analyses may not be reliably interpretable.
As previously mentioned, the technique is heavily modifiable according to the specific test hypotheses. There exist other behavioral measurements that can be investigated to be related in a behavioral syndrome among a number of diverse fish species39-41. While the tenets of the protocol remain the same, the specific questions can be easily altered to allow for diverse hypothesis testing. For example, a researcher may use the aforementioned protocol to test for the presence of an aggression and exploratory activity behavioral syndrome in convict cichlids (Amatitlania nigrofasciata). The question is related, but the study organism is different from the one described by the protocol. However, the general steps remain largely the same. Individuals must be randomized and tracked through a series of assays, with acclimation periods in between assays, the water must be consistent and fresh, and the scientist should be properly trained. One major difference is the behaviors that are being measured are likely to change according to the specific assay, and the behavioral associations may change according to the specific species and original environment.
The limitations of the protocol are linked with some of the traditional uncertainties of behavioral studies. Specifically, if the assays are scored unreliably then it is likely the interpretations of the findings are erroneous, and it would be difficult to identify this error. To overcome this limitation, it is possible that two well-trained scorers observe behaviors. To assess their reliability, calculate intraclass correlation coefficients on measurements applied on the same fish, and, if necessary, adjust for differences. Alternatively, if available, automatic tracking software such as Ethovision can be implemented and validated by a well-trained observer to increase throughput and accuracy42. Furthermore, there are different possible interpretations regarding the naming and scoring of the "boldness" behaviors. Other studies have termed the behavior described43 and others have described the behavior as "exploratory"44,45. In our work the behavior was described as "boldness" as it was measuring behavior in an environment unfamiliar to the focal individual. However, while the term may be subject to alternative interpretations, this does not affect the protocol or the analysis. Additionally, while we expect measurements within boldness and aggression assays to be strongly correlated within a population, there is likely to be some instances of low to no correlation of some behaviors. This limitation is overcome by the strength of the PCA, because it keys in on important sources of variability and, even if measurements are not correlated, the analysis will extract behavioral variation consistent in the collected data. Lastly, as is the case with all scientific methods, if the protocol is universally adopted and is consistently carried out by several labs, and there is some unforeseen, unmeasured confounding introduced by the protocol, then this potential deleterious element persists in the literature, and it becomes difficult to dismantle. Pharmacologically confirming that a representative set of assays elicit the intended behavioral response requires a more thorough understanding of behavioral, neurological and hormonal responses. Behavioral syndromes can help explain the basis of behavior, but this laboratory limitation may be able to be more regularly addressed in future studies. Nevertheless, results have been provided that validates the use of this protocol in zebrafish, and, with proper modifications, the protocol can be extended to a variety of hypotheses in a number of different fish species. By carefully following a detailed protocol for the housing, selection and testing of various behaviorally parameters will enable researchers to make more specific comparisons across a wide array of studies.
The authors have nothing to disclose.
This work was supported by a Howard Hughes Medical Institute Education Grant and an internal grant from the Saint Joseph’s University chapter of Sigma Xi. We would also like to thank the three anonymous reviewers who helped strengthen the protocol and interpretations.
Zebrafish Rack System | Aquaneering Inc | Cat. # ZS550 | |
Pet Valu Tropical Fish Food, 224.0g | Pet Valu | Cat. # 31700 | |
Premium Grade Brine Shrimp Eggs, 16 oz | Brine Shrimp Direct | ||
1.5 L Trapezoidal Tank | Pentair Aquatic Ecosystems | Cat. # itsts-a | |
19L rectangular tank | That Fish Place | 211932 | |
76L rectangular tank | That Fish Place | 212180 | |
Hitachi KP-D20A CCD Camera | Prescott's, Inc. | ||
Nikon AF Nikkor 35-105mm f/305~4.5s MACRO lens | Nikon Corporation | ||
ArtMinds Square Mirror, Value Pack 3"x3" | Michaels | Cat. # 10334162 | |
Jwatcher | |||
SPSS Statistics Base | IBM | ||
R | The R Foundation |