Retinal Topography Maps in R

I’m happy to report that we published a paper in Journal of Vision this week. It’s a methods paper on an open-source R package for retinal topography mapping and comparison. The paper represents part 1 of Brian’s senior thesis — major kudos! (Publication of part 2 is still in the works…)

Cohn, B. A., Collin, S. P., Wainwright, P. C., & L. Schmitz (2015). Retinal topography maps in R: New tools for the analysis and visualization of spatial retinal data. Journal of Vision, 15(9):19, 1–10. doi:10.1167/15.9.19.

Figure 1. An example of the functionality of Retiina. Panel (a) shows a simplified outline and the sampling locations of a retinal wholemount. Panel (b) shows the reconstructed map, while (c) illustrates the fit error of the model. [Figure 1 of Cohn et al. 2015, Journal of Vision]

Figure 1. An example of the functionality of Retina. Panel (a) shows a simplified outline and the sampling locations of a retinal wholemount. Panel (b) shows the reconstructed map, while (c) illustrates the fit error of the model. [Figure 1 of Cohn et al. 2015, Journal of Vision]

Retinal topography maps are a very widely used tool in visual ecology, as they summarize important information about the distribution of different cell types across the entire retina. These cells can be photoreceptors or ganglion cells, among others, and such cells have often very different densities across the retinal hemisphere. Humans, for example, have one very specialized area with lots of photoreceptors and ganglion cells, the fovea, whereas the peripheral parts of the eye have far less cells. Importantly, these high-density areas correlate with visual acuity. Other species feature very different distribution of cells in the retina, sometimes multiple peaks, or a streak, or a combination of different features. A good taste of the diversity of retinal topographies across vertebrates can be found here, with hundreds of examples.

A major idea and research approach in the field of visual ecology is to link the diversity of retinal topographies to the lifestyle of organisms. And this is what I originally set out to do as part of my postdoc research with Peter Wainwright and Shaun Collin. Early on we recognized that comparisons between different eyes (be it of the same species or eyes from different species) wasn’t necessarily straightforward. Two major problems emerged: (1) the incisions required to flatten the eye cup in order to make a wholemount were never quite consistent; and (2) the reconstruction of topography, often done by manual interpolation, didn’t seem 100% accurate or objective. Obviously we weren’t the first to notice this. I’m not going to review the entire body of literature (we tried to do this as thoroughly as possible in the paper, but I want to highlight two recent contributions by David Sterratt and Eduardo Garza-Gisholt.

Sterratt managed to provide a new tool to tackle the problem of incisions and folding, implemented in the R package Retistruct. That’s a big step forward. Garza-Gisholt explored approaches to visualize cell densities in R using a few different methods. Both studies combined were an excellent starting point for us to tackle writing a software that provided a one-stop visualization tool for retinal topography maps, which we named ‘Retina.’

‘Retina’ is very easy to learn, and with stereology count data at hand, you can generate several maps per hour. Useful features include generation of average maps and a range of descriptive statistics that allow for a deeper assessment of the modeled map. An extensive tutorial with example data, and, of course, all code, is available on Github. If you would like more detailed instructions on how to use the software, please do not hesitate to contact us.

If you are planning on working on retinal maps in the future, it would be great if you gave our software a try. We hope it will be useful! There are many aspects of quantitative retinal mapping that are understudied and any additional empirical study will be very helpful in moving our field forward. Please also let us know if you have any questions, concerns, and suggestions. Our goal is to further develop Retina based on feedback and requests from users.

Specifically, what criteria are important for you when making retinal topography maps?

What features of the current version do you like (or dislike)?

We would love to hear from you.

Posted in own research, R, vision | Tagged , , , , , , ,

Sniper-scope: mapping and understanding the visual system of nature’s underwater sniper

By Samantha Decker (Scripps College) [Edited by Lars Schmitz, as part of BIOL 167 “Sensory Evolution”, an upper division class at the W.M. Keck Science Department. Written for educational purposes only].

How animals turn the chaos of environmental stimuli into an accurate, integrated and useful perception is one of the greatest questions of biology, psychology, neuroscience and even philosophy. Careful research can help answer some of the questions that Locke, Socrates, Plato, Descartes, Wundt, Gibson and curious toddlers alike have all had such as “When I see red am I seeing what you call red?” or maybe more advanced “How does light turn into what I see when I open my eyes?” or “How do I know what I’m looking at is real? And what does real even mean anyways?” Before this gets too Matrix-y, let’s bring some scientific understanding into the discussion of vision and perception. But how?

In order to understand what is happening between the time when a light wave hits a retina and the perception of an object, careful investigation of the visual system must be conducted. The birth of neuroscience as a field was pushed by the Nobel-winning discoveries of Hubel and Wiesel with their careful characterization of the cat visual system. Since then, more research on mammal visual systems has occurred and we have a much clearer understanding of how visual stimulus is coded in certain brain regions and integrated to form a perception. Expanding this type of investigation to other species allows for an understanding of how perception evolved and a clearer understanding of what happens between stimulus, perception and behavior.

Archer fish are unique among fish for their sniper-like style of hunting. They shoot a powerful jet of water from their mouth knocking prey off overhanging branches into the water where they snatch it up. Check out this cool video.

.

They are able to down targets that move rapidly and adjust the stream of water based on perceived size, speed, and direction of the prey all while dealing with the refraction difference between air and water (Schuster et al, 2006; Gerullis and Schuster, 2014). This remarkable fish is an interesting model organism to study vision due to the complexity of its visual environment and behavior as well as the evolutionary knowledge gained by investigating our not-so-distant (and fishier) relatives in the teleost fish family.

The Archer fish. [commons.wikimedia.org/wiki/File:ArcherFish.jpg]

Figure 2. The Archer fish. [commons.wikimedia.org/wiki/File:ArcherFish.jpg]

Using the archer fish as a model organism to understand vision provides insight into human vision as well as providing information on how this amazing sensory system has evolved and what allows a species to become a visual learner. Recently, Ben-Tov et al. experimentally defined the visual receptive field in this fascinating model organism. Knowing the organization of the visual receptive field can lead to all sorts of interesting research such as evolution of brain regions, developmental biology, characterizing types and networks of neurons all leading to the illumination of the ‘black box’ surrounding how a stimulus turns into perception. The authors investigated the archer fish’s optic tectum and sought to analyze and define the functional organization and receptive field properties of the cells in these areas. To do this, they used extracellular measurements in the optic tectum to record responses from singles cells and develop a retinotectal map. Then, using the same extracellular recording methods, the receptive fields of those cells were determined.

Ben-Tov et al. used the same method to answer both their questions about retinotectal map as well as receptive fields of those cells. They took anesthetized and immobilized archer fish and performed surgery to open the skull above the optic tectum (recording set up in Figure 3b). Then, extracellular recordings of single cells were recorded as various stimulus types were shown to the fish (Figure 3). The various stimulus types were selected to see if the activity of the neuron they were recording from changed when they changed the visual stimuli. These changes were then quantified by counting action potentials recorded (visualized as dots on raster plots like in Figure 6 a and c and insets on figure 7) and using standard index equations used to quantify the cell’s ‘preference’ for certain stimulus.

The archer fish and the experimental setup. A: an archer fish shoots at an insect above water level. B: a schematic view of the experimental setup, which consists of a monitor to present the stimuli, a recording setup of 1 extracellular electrode. C–F: the 4 types of visual stimuli used in the experiment. C: on-off diffuse flash, a repeating loop consisting of 1 s of white screen followed by 1 s of black screen. D: static bar, a stationary bar in 8 different orientations (orientations varied by 22.5°). E: moving bar, a bar moving in 8 different directions (directions varied by 45°). The direction of motion was orthogonal to the bar's orientation. F: sinusoidal gratings, drifting in different spatial and temporal frequencies. Spatial frequency varied between 0.03 and 0.7 cycles/deg (6 logarithmic steps), and temporal frequency varied between 1.5 and 9 cycles/s (6 linear steps). Ben Tov et al, 2013

Figure 3. The archer fish and the experimental setup. A: an archer fish shoots at an insect above water level. B: a schematic view of the experimental setup, which consists of a monitor to present the stimuli, a recording setup of 1 extracellular electrode. C–F: the 4 types of visual stimuli used in the experiment. C: on-off diffuse flash, a repeating loop consisting of 1 s of white screen followed by 1 s of black screen. D: static bar, a stationary bar in 8 different orientations (orientations varied by 22.5°). E: moving bar, a bar moving in 8 different directions (directions varied by 45°). The direction of motion was orthogonal to the bar’s orientation. F: sinusoidal gratings, drifting in different spatial and temporal frequencies. Spatial frequency varied between 0.03 and 0.7 cycles/deg (6 logarithmic steps), and temporal frequency varied between 1.5 and 9 cycles/s (6 linear steps). Taken from Ben Tov et al., 2013 (Fig.1).

Retinotectal Map:

Retinotectal mapping refers to a mapping between the visual field (what the fish can see in the environment) and the surface of the optic tectum (the superficial brain region where visual processing was thought to be). This mapping is shown in the color coding seen in Figure 4, parts a and b. Part (a) shows an image of an archer fish brain with the optic tectum circles and then ‘zoomed in’ below with their map schematically shown. Part (b) shows the location in the archer fish’s visual field that corresponds to each color which is mapped onto the tectum in part a.

Figure 4

Figure 4. Space and size organization of receptive fields in the archer fish optic tectum. A: a photo of the archer fish brain with the left optic tectum delineated (top) and the electrode positions on the left optic tectum (bottom). B: topographical arrangement of receptive fields in the right visual field of the archer fish. The black circle represents the fish eye. Each color corresponds to the electrode position shown in A. Solid rectangles represent the average receptive field location and dimension, and contour rectangles represent the area into which all the receptive fields belonging to the same grid fall. Taken from Ben Tov et al, 2013 (Fig.2)

After they had recorded and located where in the visual field a certain cell responded to and determined the mapping, Ben-Tov et al. continued recording from that cell and determined if it was selective to different types of stimulus (Figure 3c-f). They determined that cells fall on a continuum between highest response to a diffuse light stimulus turning on, off or both (Figure 5). This shows that most cells respond to both on and off stimulus, so they continued on to determine the receptive field of the cells and whether they were selective for stimulus type.

Figure5

Figure 5. Cell responses to ON-OFF stimulus. Distribution of the response ratio of ON and OFF responses shows that the cells fall on a continuum of response patterns between ON and OFF classes. The most abundant pattern is a symmetrical ON-OFF response. Taken from Ben Tov et al., 2013 (Figure 3d)

Receptive Field:

Ben-Tov et al. found cells that responded strongly only when a bar stimulus was in a certain orientation (Figure 6 a and b) as well as some that didn’t seem to care about orientation and responded equally to the bar stimulus regardless of the orientation (Figure 6 c and d). They determined that they could categorize cells in the optic tectum as either orientation-tuned or orientation-agnostic cells.

Figure6

Figure 6. Orientation selectivity profiles assessed using the static bar stimulus. A: raster plots of the responses to the 8 oriented bars for an orientation-tuned cell. Black lines represent the start and end time of the bar display. B: orientation tuning profile for the orientation-tuned cell from A. The dashed black line represents the fitted von Misses distribution function. C: raster plots of the responses to the 8 oriented bars for an orientation-agnostic cell. D: orientation tuning profile for the orientation-agnostic cell from C. Taken from Ben Tov et al., 2013 (Fig.4)

Another type of stimulus Ben-Tov et al. tried was a moving bar. After finding the center of the receptive field of the cell they moved the stimulus in various directions relative to the fish. They found three distinct types of response profiles. Orientation-tuned cells showed a preference for bars moving along a specific axis. The example shown in Figure 7a is a cell that is highly responsive when the bar moves along the nasal-temporal axis (from the fish’s nose to the side of its head). Alternately, the direction-tuned cell in Figure 7b is only responsive when the bar moves from nasal to temporal but not temporal to nasal. And a final cell response profile was directional agnostic where the cell equally responded to the bar moving along any axis (Figure 7c).

Figure 7. Direction selectivity profiles assessed using the moving bar stimulus. A: orientation-tuned cell profile. B: direction-tuned cell profile. C: direction-agnostic cell profile. D: distribution of selectivity (with 95% confidence interval) for agnostic, orientation-tuned, and direction-tuned cells. Taken from Ben Tov et al., 2013 (Fig.5)

Figure 7. Direction selectivity profiles assessed using the moving bar stimulus. A: orientation-tuned cell profile. B: direction-tuned cell profile. C: direction-agnostic cell profile. D: distribution of selectivity (with 95% confidence interval) for agnostic, orientation-tuned, and direction-tuned cells. Taken from Ben Tov et al., 2013 (Fig.5)

Yet another stimulus type Ben-Tov et al. explored was the sinusoidal grating, trying to identify frequencies that elicited maximal responses. They found low pass cells which had a maximum response to low temporal frequencies, and high pass cells which had a maximum response to high temporal frequencies or no preference all pass (left side of bar graph Figure 8 labeled temporal tuning). To clean up this data a bit, they fitted a Gaussian function to look at difference in response to spatial frequencies and found two types. A low pass for spatial tuning which has a maximum response at the lowest spatial frequency, and a band pass for spatial tuning which had a maximum response in the middle of the frequency range (right side of Figure 8 labeled spatial tuning). Overall, the data show that there are cells in the optic tectum that are clearly stimulus-specific.

Figure 8. Distribution of tuning profiles (with 95% confidence interval). Light gray, temporal frequency tuning; dark gray, spatial frequency tuning. We calculated the confidence interval for the possibility of the existence of an additional tuning profile group that we have not observed in our data (see MATERIALS AND METHODS). For both the spatial frequency and temporal frequency tuning, we found that with 95% of certainty, the proportion of a possible missing group is between 0 and 0.12. Taken from Ben Tov et al., 2013 (Fig.6H)

Figure 8. Distribution of tuning profiles (with 95% confidence interval). Light gray, temporal frequency tuning; dark gray, spatial frequency tuning. We calculated the confidence interval for the possibility of the existence of an additional tuning profile group that we have not observed in our data (see MATERIALS AND METHODS). For both the spatial frequency and temporal frequency tuning, we found that with 95% of certainty, the proportion of a possible missing group is between 0 and 0.12. Taken from Ben Tov et al., 2013 (Fig.6H)

Other research labs had previously focused on the archer fish’s special behavior enabled by their remarkable terrestrial/aquatic vision. If you’ve ever tried to look underwater or looked at your straw sitting in a glass you’ve noticed the distortion that occurs when looking at an object in water versus in air. The archer fish’s behavior that has been the focus of research centers on their ability to learn how to compensate for this distortion to accurately determine the size of prey and adjust hunting behavior as fit (Gerullis and Schuster, 2014). Archer fish also learn to alter hunting strategy based off of speed of the target as well as learning this activity as juveniles by observing peers and successful adults (Schuster et al, 2006). To understand this behavior more fully, knowledge of the visual system must be expanded. It had been determined previously determined that there are directionally selective retinal ganglion cells in archer fish (Tsvilling et al, 2012). Ben-Tov et al. expand this investigation and show that there are directionally selective (as well as orientation selective) cells in the tectum (in addition to the retina). This substantially advances the field by providing valuable information about organization of visual processing in the archer fish. This ‘zooming out’ within the visual processing system provides valuable information and allows others to leverage this knowledge to gain a deeper understanding of this valuable model organism.

Characterization of the intricate details of the teleost visual system makes it possible to investigate how visual processing evolved within teleosts. Such studies allow for comparisons to well characterized visual processing systems and could result in improved understanding of how the evolution of organized sensory systems and brain regions results in perception that is accurate and useful for the organism.

Ben-Tov et al also made an attempt to make a comparison to other species’ functional units for visual processing. It is possible that the optic tectum (the dorsal-ventral axis) of the archer fish was reversed compared to some other fish species, which might be a result of the archer fish’s unique visual behavior of having a visual environment above and below water. This is an intriguing hypothesis, which clearly should be investigated in more detail. One could do this by comparisons between archer fish and close relatives that do not rely on terrestrial vision for hunting.

Ben-Tov et al. also speculated as to the analogue of their findings in mammals, an interesting developmental and evolutionary comparison. There may be functional similarities between the archer fish’s optic tectum and the mammalian superior colliculus, yet little data analysis or statistics are in place to back this up at this point. Nevertheless, understanding the evolution of optic tectum structures may allow for further understanding of disorders and development in humans. It would also be interesting to investigate some of the behaviors in archer fish while using the map and recording from cell types characterized in this paper to more deeply understand the coding process that orchestrate the complex behaviors of Archer fish. Another possible future study would be to see where the cells characterized in this paper project (using lentiviral, GFP or CLEAR brain techniques) to further ‘zoom out’ within the organization of visual processing.

References

Ben-Tov, M., Kopilevich. I., Donchin, O., Ben-Shahar, O., Giladi, C., Segev, R. 2013 Visual receptive field properties of cells in the optic tectum of the archer fish. Journal of Neurophysiology 110(3), 748-759. (DOI: 10.1152/jn.00094.2013)

Schuster, S., Wohl, S., Griebsch, M., Klostermeier, I. 2006 Animal Cognition: How Archer Fish Learn to Down Rapidly Moving Targets. Current Biology 16(4), 378-383. (doi:10.1016/j.cub.2005.12.037)

Tsvilling, V., Donchin, O., Shamir, M., Segev, R. 2012 Archer fish fast hunting maneuver may be guided by directionally selective retinal ganglion cells. European Journal of Neuroscience 35(3), 436-44 (DOI: 10.1111/j.1460-9568.2011.07971.x)

Gerullis, P., and Schuster, S. 2014 Archerfish Actively Control the Hyrdrodynamics of their Jets. Current Biology. 24(18), 2156-2160. (doi:10.1016/j.cub.2014.07.059)

Posted in sensory evolution blogs, vision | Tagged , , ,

The “Nose” Knows: Olfactory Receptor Losses Explain the Transition to Herbivory in Flies

By Madison Dipman (Pomona College) [Edited by Lars Schmitz, as part of BIOL 167 “Sensory Evolution”, an upper division class at the W.M. Keck Science Department. Written for educational purposes only].

A puzzling discrepancy exists in the world of insects. Although herbivorous species make up half of all known insect orders, the switch to a plant-only diet evolved in merely one-third of all living orders. Why is this the case, and what does it mean for insects desiring a change in diet?

Scientists have long hypothesized that this incongruity may be associated with the inherent challenges of colonizing plants. To avoid predation and parasitism by insects, plants have evolved a series of toxic chemical and physical defenses that impair the ability of insects to take advantage of nutrient-rich structures and potential homes for their offspring. Switching from feeding on yeast and other microbes on decaying plant tissues to targeting living plants requires an intricate combination of physiological, morphological, and behavioral adaptations—certainly a bit trickier than giving up meat on a whim or for a New Year’s resolution. Identifying the functional genomic changes that underlie the transition to herbivory in insects may reveal how this novel innovation arose from preexisting chemosensory circuits and why it has caused such massive adaptive radiation events in the paleontological record.

No nose?! But how do insects smell?

Like other high order animals, flies sense odors with olfactory organs located on their heads. Although fly “noses” on the antennae and maxillary palp look quite different from mammalian noses, the underlying odorant receptor neurons (ORNs) used for smell are morphologically similar to vertebrate ORNs (Vosshall and Stocker, 2007; Figure 1).

The position of olfactory (pink) and gustatory (blue) neurons on the fly (left). Scanning electron image of a fly head with major chemosensory organs (antenna and maxillary palp) labeled (right). [Figure 1a,b of Vosshall et al., 2007]

The position of olfactory (pink) and gustatory (blue) neurons on the fly (left). Scanning electron image of a fly head with major chemosensory organs (antenna and maxillary palp) labeled (right). [Figure 1a,b of Vosshall et al., 2007]

In Drosophila, odorants are detected by distinct subsets of olfactory receptor neurons (ORNs), which are localized in sensilla that cover the surface of the two olfactory organs: the antennae and maxillary palp. A sensillum is a broad term that describes an arthropod sensory organ associated with an insect’s tough outer covering. After odors are recognized, the ORNs send action potentials via axons to the antennal lobe of the brain, which processes olfactory information and initiates a behavioral response (de Bruyne et al., 2009; Hansson et al., 2010).

The responses of most insect ORNs rely on members of a large family of odorant receptor (OR) genes, which constitutes one of the most diverse gene families in insects. Despite significant variation in OR sequences and divergence within the genus Drosophila for 40 million years, functional ORN responses have been conserved across millions of years. For example, the OR gene Or42b is highly conserved across species, and is necessary for attraction and orientation to chemical structures in yeast volatiles. Because similar compounds activate Or42b across many Drosophila species, researchers postulate that volatile cues and the associated receptors for yeast detection are conserved across the Drosophilidae (Goldman-Huertas et al., 2014).

What does this have to do with herbivory?

Insect olfactory systems have evolved to monitor volatile chemicals in the environment and respond to olfactory-triggered cues that vary depending on the needs and habitat of the insect species. Previous studies have found that the diversification of the chemosensory repertoire from a relatively stable number of genes may reinforce or encourage adaptations, especially if these genes evolve to confer novel functions or are redeployed in new developmental contexts (Cande, Prud’homme, and Gompeil, 2013).

Goldman-Huertas and other researchers at the University of Arizona applied the idea that chemonsensory genes may diversify and encourage adaptations to the aforementioned herbivory issue and set out to test the hypothesis that the functional loss of chemosensory genes played a critical role in the transition to herbivory in insects. Changing the genetic repertoire, they argued, would in turn reorganize neurological processes involved in detection of plant volatiles, thereby drastically affecting behavior.

Drosophilidae makes an excellent study system for examining herbivory evolution in insects, since it boasts the ultimate genomic model for olfactory studies, Drosophila melanogaster—the fruit fly—as well as several well-documented transitions to herbivory. This 2014 study focuses on Scaptomyza flava, an herbivorous close relative of D. melanogaster. The ancestral niche for the genus Scaptomyza is microbe-feeding, but Scaptomyza species use decaying leaves and stems rather than fermenting fruit for olfactory-directed behaviors, such as feeding and oviposition (i.e. laying eggs) (Lapoint, O’Grady, and Whiteman, 2013; Figure 2).

Adult female S. flava fly with green abdomen after feeding on mustard leaf tissue. This behavior is one example of how S. flava uses plants for survival. [Figure 1 of Goldman-Huertas et al., 2014]

Adult female S. flava fly with green abdomen after feeding on mustard leaf tissue. This behavior is one example of how S. flava uses plants for survival. [Figure 1 of Goldman-Huertas et al., 2014]

Based on a time-calibrated Bayesian phylogeny, it appears that herbivory evolved only a single time within the genus about 13.5 million years ago and may have followed close association of flies with decaying plant tissues (Lapoint, O’Grady, and Whiteman, 2013; Figure 3). A Bayesian phylogeny incorporates the prior probability of an event occurring and the likelihood of an event occurring, allowing it to quickly produce both a complex tree estimate and measures of uncertainty for the groups on the tree (Holder and Lewis, 2003).

Time-calibrated Bayesian phylogeny of Drosophilidae species, including Drosophila and Scaptomyza species. Pie graphs at each node indicate the probability of a change to herbivory (green) or retention of ancestral microbe-feeding (white) traits. Herbivorous taxa are indicated by the leaf and bracket. [Figure 1 of Goldman-Huertas et al., 2014]

Time-calibrated Bayesian phylogeny of Drosophilidae species, including Drosophila and Scaptomyza species. Pie graphs at each node indicate the probability of a change to herbivory (green) or retention of ancestral microbe-feeding (white) traits. Herbivorous taxa are indicated by the leaf and bracket. [Figure 1 of Goldman-Huertas et al., 2014]

Since each ORN has characteristic spike amplitude, it is possible to determine the activity of a single neuron with electrophysiological assays (Goldman et al., 2005). In a single-unit electrophysiology, a sensillum (and the ORNs in the surface) is stimulated by application of an odor. A fine-tipped electrode is then inserted into the sensillum to measure the electrical activity of the OR neurons by recording the action potentials from the cell (Figure 4). Using this technique and genomic mapping, researchers have been able to functionally describe the OR gene family in Drosophila species.

Head of a fly showing fluorescence of the maxillary palp (left). Micrograph of the fly maxillary palp, demonstrating the use of an Or promoter to drive expression of a gene that confers fluorescence (middle). Single-unit electrophysiology (right), in which a recording electrode is placed through a fluorescently-labeled sensillum. [Figure 1B of Goldman et al., 2005]

Head of a fly showing fluorescence of the maxillary palp (left). Micrograph of the fly maxillary palp, demonstrating the use of an Or promoter to drive expression of a gene that confers fluorescence (middle). Single-unit electrophysiology (right), in which a recording electrode is placed through a fluorescently-labeled sensillum. [Figure 1B of Goldman et al., 2005]

Goldman-Huertas et al. collaborated with neuroscientists to perform these assays on S. flava and D. melanogaster. They measured the electrical responses in the insects’ antennae generated by olfactory receptors following presentation of two different scents: yeast and live plant (Arabidopsis) volatile compounds, which are both emitted in nature and would indicate the presence of a viable food source contingent on feeding preferences. The antennae of S. flava were more strongly stimulated by Arabidopsis volatiles than yeast, and the antennae of D. melanogaster were significantly more responsive to yeast volatiles than Arabidopsis (Figure 5). They also tested other simpler compounds associated with yeast and leaf tissue, and determined that overall, S. flava was less sensitive to short chemical compounds associated with yeast strains (e.g. aliphatic esters), which may explain S. flava’s lack of attraction to yeast volatiles.

Boxplot results from the electroantennogram assay. S. flava had diminished antennal responses to yeast volatiles and enhanced responses to plant-related Arabidopsis volatiles, whereas D. melanogaster was more sensitive to yeast volatiles than Arabidopsis. [Figure 2b (top) and 2c (bottom) of Goldman-Huertas et al., 2014).

Boxplot results from the electroantennogram assay. S. flava had diminished antennal responses to yeast volatiles and enhanced responses to plant-related Arabidopsis volatiles, whereas D. melanogaster was more sensitive to yeast volatiles than Arabidopsis. [Figure 2b (top) and 2c (bottom) of Goldman-Huertas et al., 2014).

This experiment was supplemented by a behavioral test carried out with a four-field olfactometer apparatus that created four independent airfields, two of which were exposed to yeast volatiles. Gravid adult females were placed in the arena, and their presence in either yeast or control fields was recorded. D. melanogaster flies spent significantly more time in yeast-volatile fields than S. flava, while S. flava did not spend more time in yeast-volatile fields, dividing time evenly between the yeast and control fields (Figure 6).

Apparatus for the four-field olfactometer assay, in which filtered air is blown through four corners of an arena and establishes four independent airfields that flies can choose between, [Figure 1 of Lehrman et al., 2013] (left). Two of these airfields were exposed to yeast volatiles. Results from the behavioral observation, indicating herbivorous S. flava flies did not spend a significantly higher proportion of time in yeast fields, consistent with a loss of attraction to yeast volatiles, [Figure 2a of Goldman-Huertas et al., 2014] (right).

Apparatus for the four-field olfactometer assay, in which filtered air is blown through four corners of an arena and establishes four independent airfields that flies can choose between, [Figure 1 of Lehrman et al., 2013] (left). Two of these airfields were exposed to yeast volatiles. Results from the behavioral observation, indicating herbivorous S. flava flies did not spend a significantly higher proportion of time in yeast fields, consistent with a loss of attraction to yeast volatiles, [Figure 2a of Goldman-Huertas et al., 2014] (right).

Neurological assays and behavioral observations revealed that the smell of yeast, which is abundant on rotting fruit, is minimally detected by antennae of S. flava and does not attract the flies. This is in stark contrast to the response of D. melanogaster. As any person who has left a piece of fruit out for too long has observed, fruit flies are drawn to this compound with extreme fervor. Although S. flava flies were not attracted to these yeast volatiles, their antennal chemoreceptors were sensitive to the Arabidopsis compound responsible for the smell of freshly cut grass, which is common in leafy plants.

Based on the lack of attraction and minimal electrical response to yeast volatiles in S. flava, Goldman-Huertas and collaborators predicted that olfactory genes important for sniffing out yeast must have been lost or altered in herbivorous Scaptomyza species. To characterize the changes in the OR gene family responsible for this behavioral modification, they sequenced the genome and annotated OR genes in herbivorous S. flava and then compared the findings to the thoroughly documented functional ORs in D. melanogaster. In the Scaptomyza lineage, only four widely conserved ORs were uniquely lost (Or22a, Or85d) or pseudogenized, indicating they are no longer functional (Or9a, Or42b). As expected, these ORs that function in microbe-feeding flies and are lost in herbivorous species play a role in detecting yeast volatiles, leading them to conclude that loss-of-function mutations were critical for the transition to herbivory in insects.

The researchers then set out to confirm their hypothesis that functional losses of portions of the OR gene family implicated in detecting yeast volatiles played a role in the ecological transition to herbivory in insects. This would be supported if OR gene losses coincided with the transition to herbivory in Scaptomyza according to their maximum likelihood ancestral state reconstruction model, which assumes the phenotypes that developed were statistically most likely but does not assume that all events are equally likely to happen. According to their model, one OR gene loss coincided with the evolution of herbivory, but losses of two other OR genes preceded the switch to plant-feeding (Figure 7). In addition, the researchers found no evidence of accelerated chemosensory gene loss in S. flava compared with other microbe-feeding Drosophila species, but this may be due to insufficient loss events to parameterize the complex model or the involvement of other gene families.

OR gene losses mapped onto a Scaptomyza phylogeny based on the results from PCR screens and genomic data. Three of the four OR genes lost coincided with or preceded the evolution of herbivory in the lineage ca.13.5 mya. [Figure 3a of Goldman-Huertas et al., 2014]

OR gene losses mapped onto a Scaptomyza phylogeny based on the results from PCR screens and genomic data. Three of the four OR genes lost coincided with or preceded the evolution of herbivory in the lineage ca.13.5 mya. [Figure 3a of Goldman-Huertas et al., 2014]

Taken together, these findings suggest that ancestral Scaptomyza had already lost conserved yeast-volatile sensors and possibly gained new olfactory pathways, which were later co-opted by herbivorous lineages to aid in the colonization of plant species. Examples of these ancestral but non-herbivorous species may include flies that live within decaying leaves or in tunnels in leaves produced by other insects (Goldman-Huertas et al., 2014). Sister groups of many herbivorous insect lineages also feed on dead organic material and fungi, which may be a precursor to full-fledged herbivory. None of these species are herbivorous, yet they are closely associated with plants, signifying the first steps of a major trophic shift.

The transition to herbivory in Scaptomyza likely involved many changes in olfactory cues, and loss-of-function mutations are not sufficient to explain this novel behavioral shift, since losing the ability to detect yeast does not inherently lead to colonizing plants. But what is missing in the story? The researchers tested for evidence of episodic positive selection in S. flava OR genes and found two ORs with a signature of positive selection. Other experimental and functional tests are needed to verify whether positive selection fixed changes to chemosensory genes in the Scaptomyza lineage, and if so, how these amino acid changes contributed to the development of herbivorous behaviors. Because it is still unclear where host-finding behaviors arose, further research on the annotated library of ORs is crucial, but potential candidates include the ORs with signatures of episodic positive selection.

But why does this matter?

The finding that the functional loss of chemosensory genes contributed to trophic shifts may be a major theme in the tree of life and one that facilitates rampant species radiation events. This study provides a valuable framework for tackling challenging questions of how these novel choices came to be as a result of subtle, targeted changes to a portion of chemosensory gene families. Armed with this knowledge, researchers may be able to answer similar questions in other insect lineages as well as animal species whose feeding preferences and behaviors diverge from those of their ancestors.

Understanding how trophic transitions are mediated by changes to the chemoreceptor repertoire also has numerous practical implications. There are many insect species (including S. flava itself) that are considered pests due to their negative impact on plant or animal lives. The “easy” solution for controlling insect pests since the Green Revolution of the 1960s and 70s has been the use of insecticides. However, abundant scientific research has demonstrated pesticides are not only harmful to other organisms and the environment, but also often fail in the long-term due to the evolution of pesticide resistance. By applying pesticides and other insect repellants to our crops and our bodies, we as humans may be changing the course of insect evolution.

The progression of experiments used in the Goldman-Huertas study may be applied to these organisms to better understand how insects made the shift from feeding on microbes to a particular plant species that humans rely on as an economically valuable crop. From these studies, researchers may be able to inform integrated pest management strategies, which seek to control pests but also minimize disruption to agroecosystems. Similar studies would also be valuable if they targeted vectors of human diseases, such as malaria-carrying mosquitoes, with the aim of deducing the genetic basis behind the trophic transition to consumption of human blood and ultimately designing repellants that could protect human health. Changes in insect behavior are intricately linked to the health and economy of humans, and there is still much to be learned about human-insect-plant relationships that may be gleaned from studying evolution in insects.

References

Cande, J., Prud’homme, B., Gompeil, N. 2013 Smells like evolution: the role of chemoreceptor evolution in behavioral change. Current Opinion in Neurobiology 23, 152-158. (DOI 10.1016/j.conb.2012.07.008)

de Bruyne, M., Smart, R., Zammit, E., Warr, C.G. 2009 Functional and molecular evolution of olfactory neurons and receptors for aliphatic esters across the Drosophila genus. J Comp Physiol A 196(2), 97-109. (DOI 10.1007/s00359-009-0496-6)

Goldman, A.L., van der Goes van Naters, W., Lessing, D., Warr, C.G., Carlson, J.R. 2005 Coexpression of two functional odor receptors in one neuron. Neuron 45(5), 661–78. (DOI 10.1016/j.neuron.2005.01.025)

Goldman-Huertas, B., Mitchell, R.F., Lapoint, R.T., Faucher, C., Hildebrand, J.G., Whiteman, N.K. 2014 Evolution of herbivory in Drosophilidae linked to loss of behaviors, antennal responses, odorant receptors, and ancestral diet. PNAS 112(10), 3026-3031. (DOI 10.1073/pnas.1424656112)

Hansson, B.S., Knaden, M., Sachse, S., Stensmyr, M.C., Wicher, D. 2010 Towards plant-odor related olfactory neuroethology in Drosophila. Chemoecology 20(2), 51-61. (DOI 10.1007/s00049-009-0033-7)

Holder, M., Lewis, P.O. 2003 Phylogeny estimation: traditional and Bayesian approaches. Nature Reviews Genetics 4, 275-284. (DOI 10.1038/nrg1044)

Lapoint, R.T., O’Grady, P.M., Whiteman, N.K. 2013 Diversification and dispersal of the Hawaiian Drosophilidae: The evolution of Scaptomyza. Molecular Phylogenetics and Evolution 69, 95-108. (DOI 10.1016/j.ympev.2013.04.032)

Lehrman, A., Boddum, T., Stenberg, J.A., Orians, C.M., Bjorkman, C. 2013. Constitutive and herbivore-induced systemic volatiles differentially attract an omnivorous biocontrol agent to contrasting Salix clones. AoB Plants 5, plt005. (DOI 10.1093/aobpla/plt005)

Vosshall, L.B., Stocker, R.F. 2007 Molecular Architecture of Smell and Taste in Drosophila. Annual Review of Neuroscience 30, 505-533. (DOI 10.1146/annurev.neuro.30.051606.094306)

Posted in sensory evolution blogs | Tagged , ,

Blind Sharks: Detect Magnetic Fields?

By Kimberly Coombs (Claremont McKenna College) [Edited by Lars Schmitz, as part of BIOL 167 “Sensory Evolution”, an upper division class at the W.M. Keck Science Department. Written for educational purposes only].

Sharks are most commonly known as visual predators; however, this does not tell the whole story of why sharks are such successful predators. It is true that sharks have highly developed vision. Yet, sharks actually rely upon a variety of different senses that have allowed them to evolve as apex predators of the oceans. For example, sharks rely upon hearing, their lateral line, electroreception, and chemoreception (Hart & Collin, 2015). All of these senses are important for sharks to be effective hunters and maintain their position at the top of the food chain. What happens if a shark loses one of its senses (i.e., that particular sense is no longer functional)? This is exactly what O’Connell and his fellow colleagues decided to study.

O’Connell et al. (2014) were interested in studying whether lemon sharks, Negaprion brevirostris, can detect magnetic fields if they were visually-deprived (Figure 1).

Figure 1. Lemon shark gracefully swimming in the ocean (By Albert Kok, GFDL [https://commons.wikimedia.org/wiki/File:Lemon_shark2.jpg]  or CC-BY-SA-3.0 [http://creativecommons.org/licenses/by-sa/3.0], via Wikimedia Commons).

Figure 1. Lemon shark gracefully swimming in the ocean (By Albert Kok, GFDL [https://commons.wikimedia.org/wiki/File:Lemon_shark2.jpg] or CC-BY-SA-3.0 [http://creativecommons.org/licenses/by-sa/3.0], via Wikimedia Commons).

For this experiment, O’Connell et al.were only concerned with lemon sharks’ senses of electroreception and vision. At this point, you are probably wondering what in the world is electroreception, since I have mentioned it twice now. Calm down and pay attention. Electroreception is a shark’s ability to detect electrical fields emitted by inanimate objects and other animals (Hart & Collin, 2015). WHOA! Did you just read that correctly, sharks can detect electrical fields? Yes, sharks are amazing creatures, how else do you think they became top predators of the seas? Sharks’ electrosensory system is more correctly known as the ampullae of Lorenzini and is mostly used to avoid predators, orient themselves, find and capture prey, and choose a mate (O’Connell et al., 2014, Hart & Collin, 2015, Hutchinson et al., 2012). Through the use of their electrosensory system, sharks are also able to detect magnetic fields. Sharks are further equipped with a tapetum lucidum, a layer of tissue that sits within the eye. This structure increases the amount of light available to the photoreceptors in the eye, especially in low light conditions; as a result, the nocturnal vision of sharks is greatly enhanced.

O’Connell et al.were specifically interested in lemon sharks’ detection of magnetic fields when visually deprived, such as when they are in turbid waters (poor visibility waters). Sharks’ ability to sense magnetic and electrical fields was first discovered in 1935, when blindfolded small spotted catsharks, Scyliorhinus canicula, demonstrated escape responses when a steel wire was brought near their heads (Hart & Collin, 2015).

In order to test shark magnetic field detection, the authors gathered 24 juvenile lemon sharks (14 male, 10 female) of roughly the same size and placed them in a holding pen. They set up an experimental pen that contained three compartments (experimental arena, recovery and acclimation pen, and a corridor). The experimental arena was further split into four zones: magnet zone, control zone, separation, and observation. The control and magnet zones were identical in that they both had three 1.75m tall polyvinyl chloride (PVC) columns facing perpendicular to the ground. However, the control zone had sham magnets or clay bricks placed in the PVC, while the magnet zone had grade C8 barium ferrite magnets placed in the PVC (Figure 2).

Figure 2. O’Connell et al. (2014) experimental set up of the three compartment pen. A) Shows all three compartments, recovery and acclimation pen, experimental arena, and corridor. B) Observation zone placed up against the substrate in the control and magnet zones. One shark at a time was placed into the recovery/acclimation pen from the holding pen before and after each trial. The shark was then guided through the corridor into the experimental arena to begin the 30min trial (From O’Connell et al., 2014).

Figure 2. O’Connell et al. (2014) experimental set up of the three compartment pen. A) Shows all three compartments, recovery and acclimation pen, experimental arena, and corridor. B) Observation zone placed up against the substrate in the control and magnet zones. One shark at a time was placed into the recovery/acclimation pen from the holding pen before and after each trial. The shark was then guided through the corridor into the experimental arena to begin the 30min trial (From O’Connell et al., 2014).

They denoted four different types of treatments for the sharks and had six sharks per treatment. The treatments consisted of a control (no manipulation to the sharks vision), procedural control ‘eyebrow’ (sharks had one suture above each eye), procedural control ‘one eye’ (sharks had either the left or the right nictitating membrane closed), and ‘visually deprived’ (sharks had both nictitating membranes closed). Hold up, what is a nictitating membrane you ask? Well, the nictitating membrane acts as an opaque third eyelid in several species, such as sharks. The upper and lower eyelids of sharks are mostly immobile; therefore, the nictitating membrane may be drawn over the eye (essentially, the nictitating membrane acts as our eyelids do when we blink) (Gruber & Schneiderman, 1975; Figure 3).

Figure 3. Lemon shark undergoing surgery to close nictitating membrane. A) Lemon shark prepped to begin surgery. D) Nictitating membrane surgically closed over eye of lemon shark (From O’Connell et al., 2014).

Figure 3. Lemon shark undergoing surgery to close nictitating membrane. A) Lemon shark prepped to begin surgery. D) Nictitating membrane surgically closed over eye of lemon shark (From O’Connell et al., 2014).

After surgery was complete, experimental testing began. O’Connell et al. had one shark enter the experimental arena at a time and three different behaviors were recorded for a 30min testing period: entrances (shark went to observation zone and swam through the PVC), visits (shark swam in one of the observation zones), and avoidances (shark changed direction suddenly and/or accelerated away after visiting one of the observation zones).

O’Connell et. al. found that juvenile lemon sharks do rely heavily on their electrosensory system to forage and navigate through turbid waters. They discovered that the magnets acted as a repellent for the sharks, causing the sharks to change their swimming behavior abruptly in order to avoid the magnets. This is consistent with other studies that found that white sharks also avoid magnets placed in the water (O’Connell et al., 2014). The authors further found that the visually deprived sharks got the closest to the magnets before exhibiting an avoidance behavior than the other experimental shark groups (Figure 4).

Figure 4. Histograms representing the percentage of avoidance distance from the magnet zone for all treatment types of lemon sharks. The secondary y axis also shows the magnetic field strength generated by the magnets. A) Control shark, B) Eyebrow shark, C) One-eye Shark, D) Visually deprived shark (n=6) (From O’Connell et al., 2014).

Figure 4. Histograms representing the percentage of avoidance distance from the magnet zone for all treatment types of lemon sharks. The secondary y axis also shows the magnetic field strength generated by the magnets. A) Control shark, B) Eyebrow shark, C) One-eye Shark, D) Visually deprived shark (n=6) (From O’Connell et al., 2014).

They believe this could be due to the electrosensory system of the sharks operating on a shorter range than vision. The control and the one eye sharks were farthest away from the magnets when they exhibited an avoidance behavior, which O’Connell et al.believe is due to both their visual and electrosensory systems being stimulated. The visually deprived sharks visited the magnet zone more frequently than the other shark types, yet they entered through the PVC much less than the other sharks (Figure 5, 6).

Figure 5. Mean entrance (total entrances/total visits) and avoidance (total avoidances/total visits) ratios for each lemon shark treatment for the magnet zone (n=6) (From O’Connell et al., 2014).

Figure 5. Mean entrance (total entrances/total visits) and avoidance (total avoidances/total visits) ratios for each lemon shark treatment for the magnet zone (n=6) (From O’Connell et al., 2014).

Figure 6. Box and whisker plot showing the median, 25th percentile, and 75th percentile of visit quantities to magnet zone prior to entering through the PVC for the first time for each lemon shark treatment (n=6) (From O’Connell et. al., 2014).

Figure 6. Box and whisker plot showing the median, 25th percentile, and 75th percentile of visit quantities to magnet zone prior to entering through the PVC for the first time for each lemon shark treatment (n=6) (From O’Connell et. al., 2014).

This result suggests that when lemon sharks are more visually deprived, they rely more heavily on electroreception. This is congruent with other studies that have shown a reliance on short-range senses, such as electroreception and mechanoreception, when environmental parameters restrict vision (Hutchinson et al., 2012, O’Connell et al., 2014). Though the authors’ method of simulating turbidity may have been invasive and not exactly representative of what it would be like in the natural environment, this is the first study to show evidence of context-dependent switching (the ability to change behavior in response to the current biological and ecological state) in shark electroreception.

So what? Why should you care? Well, this is very important from an evolutionary standpoint. Since context-dependent switching has been found in sharks, it is highly possible that this ability may also be found in other elasmobranchs. In fact, context-dependent switching has been documented in other organisms, such as teleosts, mammals, and amphibians (O’Connell et al., 2014)-. This indicates that context-dependent switching probably evolved with a common ancestor of elasmobranchs, teleosts, mammals, and amphibians. Studies should be looking into whether other elasmobranchs have this capability as well as where in the tree of life did species start to demonstrate context-dependent switching.

Furthermore, it is not presently known if suturing closed the nictitating membrane of sharks represents how these sharks would respond in natural turbid waters. If it is accurate, the authors suggest that using magnet repellents in the field would be successful at manipulating shark behavior in turbid water environments. Meaning that if accidental shark attacks and shark bycatches wanted to be avoided, magnets should be placed in the water as sharks will avoid these areas. Future studies should be focused on examination of turbid water conditions, as this factor might be the quintessential factor for determining the most successful repellents to use in turbid waters. Also, studying turbid water conditions will aid in the understanding of how and why sharks’ change the senses they rely upon in these conditions. Researchers should look at the effectiveness of magnetic repellents in areas that are well known as being turbid, such as inshore areas influenced by eutrophication, runoff, and riverine input. Such studies should further look at the impact magnetic repellents have on other species besides sharks, as these repellents may have a negative impact on other species and could cause a change in the ecology of the area the repellents are emplaced.

 

References

Gruber, S.H., Schneiderman, N. 1975. Classical conditioning of the nictitating membrane response of the lemon shark (Negaprion brevirostris). Behavior Research Methods and Instrumentation 7(5), 430-434.

Hart, N.S., Colling, S.P. 2015. Sharks senses and shark repellents. Integrative Zoology 10, 38-64. (DOI: 10.1111/1749-4877.12095).

Hutchinson, M., Wang, J.H., Swimmer, Y., Holland, K., Kohin, S., Dewar, H., Wraith, J. Vetter, R., Heberer, C., Martinez, J. 2012. The effects of a lanthanide metal alloy on shark catch rates. Fisheries Research 131-133, 45-51. (DOI: 10.1016/j.fishres.2012.07.006).

O’Connell, C.P., Andreotti, S., Rutzen, M., Meÿer, M., Matthee, C.A., He, P. 2014. Effects of Sharksafe barrier on white shark (Carcharodon carcharias) behavior and its implications for future conservation technologies. Journal of Experimental Marine Biology and Ecology 460, 37-46. (DOI:10.1016/j.jembe.2014.06.004).

O’Connell, C.P., Guttridge, T.L., Gruber, S.H., Brooks, J., Finger, J.S., He, P. 2014. Behavioral modification of visually deprived lemon sharks (Negaprion brevirostris) towards magnetic fields. Journal of Experimental Marine Biology and Ecology 453, 131-137. (DOI: 10.1016/j.jembe.2014.01.009).

Posted in electroreception, magnetoreception, sensory evolution blogs | Tagged , , ,

New lab webpage

The blog remains a the good old wordpress location, but I have moved the info about my lab and research to a new webpage. The main reason behind is that I wanted to have better control over publishing code from research and teaching projects.

Posted in Uncategorized

Phylogenetic trees in R

I wrote this little tutorial as an introductory chapter for the NESCent Academy on Macroevolution back in July 2014. It’s meant to provide a brief overview of the basic structure of tree objects in R and illustrate some of the tree manipulation and visualization options.

Given that I’ll be teaching a module on comparative methods in our ‘Research Methods in Organismal Biology’ course soon, I figured I may as well make all scripts and exercises accessible on this blog. Once you have mastered this, explore many more options in Emmanuel Paradis’ book (Analysis of Phylogenetics and Evolution with R) and Liam Revell’s blog.

Let’s first load some libraries that we need for this exercise.

    library(ape)
    library(geiger)

Trees as “phylo” objects

Let’s begin by simulating a tree. There are many options for doing this. We’ll use the rtree() function of the ‘ape’ package. The tree is stored as an object called “phy”.

    phy <- rtree(n=10) 
    # n specifies the number of tips we want.

How does the tree look like? We can plot it with the following line.

    plot(phy)

 

index01So, now we wonder how the phylogenetic information is encoded. And no, it’s not a black box. By typing “phy” in the command line we can retrieve some basic information about the object.

    phy 
## 
## Phylogenetic tree with 10 tips and 9 internal nodes.
## 
## Tip labels:
## t3, t8, t5, t4, t7, t2, ...
## 
## Rooted; includes branch lengths.
# A tree with n tips and (n-1) nodes.
# There are tip labels, the tree is rooted, and we have branch lengths.

But how is this information organized within the phylo object? We can find out with the str() function, which displays the structure of an R object.

    str(phy)
## List of 4
##  $ edge       : int [1:18, 1:2] 11 12 13 13 12 14 15 15 14 16 ...
##  $ tip.label  : chr [1:10] "t3" "t8" "t5" "t4" ...
##  $ edge.length: num [1:18] 0.944 0.996 0.842 0.542 0.627 ...
##  $ Nnode      : int 9
##  - attr(*, "class")= chr "phylo"
##  - attr(*, "order")= chr "cladewise"

The output tells us that a phylo object is a list of four components (+ two attributes): edge, tip.label, edge.length, and Nnode. We can display these items separately.

    phy$edge
##       [,1] [,2]
##  [1,]   11   12
##  [2,]   12   13
##  [3,]   13    1
##  [4,]   13    2
##  [5,]   12   14
##  [6,]   14   15
##  [7,]   15    3
##  [8,]   15    4
##  [9,]   14   16
## [10,]   16    5
## [11,]   16   17
## [12,]   17   18
## [13,]   18    6
## [14,]   18    7
## [15,]   17    8
## [16,]   11   19
## [17,]   19    9
## [18,]   19   10
    phy$tip.label
##  [1] "t3"  "t8"  "t5"  "t4"  "t7"  "t2"  "t10" "t1"  "t9"  "t6"
    phy$edge.length
##  [1] 0.9437 0.9960 0.8423 0.5422 0.6265 0.6445 0.1634 0.7275 0.8058 0.8304
## [11] 0.9346 0.5120 0.9839 0.7717 0.2805 0.1774 0.8578 0.3754
    phy$Nnode
## [1] 9

And of course, we can also store these components as new objects with names of your choice.

    branches <- phy$edge
    species <- phy$tip.label
    brlength <- phy$edge.length
    nodes <- phy$Nnode

But what does this all mean? Let’s explore this with a hand-written tree. Assume you have a tree with 6 species (A through F). The phylogenetic relationships of species A:F can be described in bracket form (“parenthetic format”, or Newick).

    mini.phy <- read.tree(text = "((((A,B), C), (D,E)),F);")
    plot(mini.phy)

index02The structure of the object contains only three components this time because we didn’t provide branch length:

    str(mini.phy)
## List of 3
##  $ edge     : int [1:10, 1:2] 7 8 9 10 10 9 8 11 11 7 ...
##  $ tip.label: chr [1:6] "A" "B" "C" "D" ...
##  $ Nnode    : int 5
##  - attr(*, "class")= chr "phylo"
##  - attr(*, "order")= chr "cladewise"

But how are the edges defined? When we type…

    mini.phy$edge
##       [,1] [,2]
##  [1,]    7    8
##  [2,]    8    9
##  [3,]    9   10
##  [4,]   10    1
##  [5,]   10    2
##  [6,]    9    3
##  [7,]    8   11
##  [8,]   11    4
##  [9,]   11    5
## [10,]    7    6

… we see a matrix of 10 rows and 2 columns. This matrix represents unique combinations of node- and tip numbers, defining each branch segment of the tree.

    plot(mini.phy, label.offset=0.2)  
    # the label.offset argument moves the names a bit to the right
    nodelabels() # add node numbers
    tiplabels()  # add tip numbers

 

index03For example, the branch (or edge) leading to species “C” is identified by row 6: 9, 3. Both nodelabel() and tiplabel() function are quite flexible and afford the opportunity to visually anhance your tree plot. Here’s an example for mini.phy object. We begin by creating an vector of different colors, with the same length as number of species in the tree.

    mycol<-c("blue", "blue", "blue", "red", "red", "red")

Now we plot the tree, moving the taxon names a bit to the right, and add the tiplabels without text, using symbols instead.

    plot(mini.phy, adj=0, label.offset=0.75, lwd=2)
    tiplabels(pch=21, col="black", adj=1, bg=mycol, cex=2)

 

index04Swapping sisterclades, identifying clades/tips, dropping tips

It sometimes may be useful to rotate the tree about a specific node, i.e. swap sister clades. This can be carried out with the rotate() function. Let’s continue to work with mini.phy:

    plot(mini.phy, label.offset=0.2)  
    nodelabels()                      
    tiplabels()                       

 

index05How about we swap clades (D, E) and (A, B, C)? Their most recent common ancestor is found at node 8.

    rot.phy <- rotate(mini.phy, node=8)

And now let’s see what happenend:

    plot(rot.phy, label.offset=0.2)   
    nodelabels()                      
    tiplabels()                       

 

index06It will also be very helpful to select all tips in a given clade. This is implemented in the ‘geiger’ package; the tips() function finds all descendants of a node.

    cladeABC <- tips(rot.phy, node=9) 
    # node 9 defines the clade composed of (A, B, C)
    
    cladeABC
## [1] "A" "B" "C"

Another helpful command allows for tree pruning, i.e. cutting of tips or entire clades. For example, we can delete the entire cladeABC group:

    pruned.phy <- drop.tip(rot.phy, cladeABC)
    plot(pruned.phy, label.offset=0.2)    
    nodelabels()                          
    tiplabels()                          

 

index07Or we can drop tips (1 or multiple) randomly. Liam Revell explained how to do this nicely on his blog. To prune tips, say m=2 random tips, enter:

    m=2
    pruned.phy2 <- drop.tip(rot.phy, sample(rot.phy$tip.label)[1:m]) 
    
    # m=1 drops 1 single tip, of course!
    
    plot(pruned.phy2, label.offset=0.2)   

index08It may also be useful to select all branches in a specific group of tips. This is implemented in the ‘ape’ package; the which.edge() function finds all edges in a specified group. For example, let’s identify all branches of the cladeABC group as defined above.

    cladeABCbranches <- which.edge(rot.phy, cladeABC) 
    
    # cladeABC was defined earlier, using the tips() function
    
    cladeABCbranches 
## [1] 6 7 8 9
    # this should be a numerical vector containing 6, 7, 8, 9

And as we can see, rows 6-9 of the edge.length matrix represent the expected branches. Let’s first plot the tree, and then look at the edge matrix for cross-checking.

    plot(rot.phy, label.offset=0.2)   
    nodelabels()                      
    tiplabels()                       

index09

    rot.phy$edge # compare edge matrix and tree plot
##       [,1] [,2]
##  [1,]    7    8
##  [2,]    8   11
##  [3,]   11    4
##  [4,]   11    5
##  [5,]    8    9
##  [6,]    9   10
##  [7,]   10    1
##  [8,]   10    2
##  [9,]    9    3
## [10,]    7    6

Here would be a good opportunity to show you how to assign different branch colors. For example, how can we emphasize the branches of the clade formed by A, B, and C? We first create a color vector, only consisting of grey colors. Then we’ll assign black to all branches of clade ABC.

    clcolr <- rep("darkgrey", dim(rot.phy$edge)[1]) 
    clcolr[cladeABCbranches] <- "black"
    plot(rot.phy, lwd=3, edge.color=clcolr)

 

index10Let’s conclude this section with one last exercise: combining trees. Assume you have two different phylogenies, with two different sets of taxa (no overlap). Another assumption is that you have knowledge how the trees may fit together. Then the bind.tree() function of ‘ape’ package can help. The function takes in two phylo-objects. The position of where the trees are bound is defined by tip- or node number within the first tree. Note that you can also specify the “root” as binding position.

    tree1 <- rtree(n=10); plot(tree1); nodelabels(); tiplabels()

index11

    tree2 <- rtree(n=10); plot(tree2)

index12

    combined.tree <- bind.tree(tree1, tree2, where=1) 
    plot(combined.tree)

index13

Fully resolved and polytomous trees

All tree examples so far were fully resolved, i.e. each tree was fully binary. It’s very easy to access visually for small trees, but we can also do this more formally:

    is.binary.tree(mini.phy)
## [1] TRUE

Let’s make a tree that is not fully resolved.

    poly.phy <- read.tree(text = "(((A,B,C),(D,E)),F);")
    plot(poly.phy)

index14

  # Is this tree binary?

    is.binary.tree(poly.phy)
## [1] FALSE

OK. Many comparative methods require fully resolved trees. But what to do if that’s not the case? The multi2di() function can resolve polytomies, either randomly or in the order in which they appear in the tree. The default setting is to resolve polytomies randomly.

    resolved.phy <- multi2di(poly.phy)
    is.binary.tree(resolved.phy)
## [1] TRUE
    plot(resolved.phy) # visual inspection.

index15If you repeat the above lines a few times you will see the effect of randomly resolving the polytomy.

Modifying tree shape and other plotting options

There are many options for formatting and beautifying trees in R. Here are some basics. Let’s begin by simulating a tree once more.

    phy <- rtree(n=10) # n specifies the number of tips we want.

    # The default plot produces a rightwards tree

    plot(phy)

index16 The tree orientation can be changed by modifying the “direction”- argument. Try it out!

    plot(phy, direction="upwards") 

index17

    # other options are "rightwards" (default), "leftwards", and "downwards".

The font size of the tip labels (species names) can be changed with the cex argument.

    plot(phy, 
         direction="upwards", 
         cex=2.5) 

index18Try a few different settings! If we don’t want the species names displayed, you can do the following:

    plot(phy, 
         direction="upwards", 
         show.tip.label=FALSE)

index19And if you would like thicker branches, do this:

    plot(phy, 
         direction="upwards", 
         show.tip.label=FALSE,
         edge.width=3) 

index20Color of the branches can be controlled by the edge.color argument:

    plot(phy, 
         direction="upwards", 
         show.tip.label=FALSE, 
         edge.width=3, 
         edge.color="blue") 

index21Try a few different settings! There are many other options… Type ?plot.phylo in the console to read more.

Another useful command I’d like to introduce is the ladderize() function. It reorganizes the tree structure, normally yielding much more readable trees. Let’s make a bigger tree to really visualize the ladderize-effect.

    big.phy <- rtree(n=50)
    ladderized.phy <- ladderize(big.phy)

Let’s create a plot with two panels, contrasting ‘before” and ’after’. With par() we can include the mfrow option, which specifies number of rows and columns in the plot matrix.

    par(mfrow=(c(1,2))) # 1 row but 2 plot panels

    plot(big.phy, direction="upwards", 
         show.tip.label=FALSE,
         edge.width=1,
         edge.color="blue")

    plot(ladderized.phy, direction="upwards", 
         show.tip.label=FALSE,
         edge.width=1,
         edge.color="blue")

index22

    par(mfrow=c(1,1)) #for subsequent single-panel plots

Finally, let’s change the type of the tree. This is done by modifying the “type” argument. The default is set to “phylogram.” Other options include “cladogram”, “fan”, “unrooted”, and “radial”. Try all of them!

    plot(ladderized.phy, type="fan", 
         show.tip.label=FALSE,
         edge.width=3,
         edge.color="blue")

index23

Tons of options! And kind of fun.

Posted in R | Tagged , , , , , | 1 Comment

Blind cavefish: social butterfly or awkward turtle?

By Nour Bundogji (Pitzer College), Niti Nagar (Claremont McKenna College), and Sachin Shah (Claremont McKenna College) [Edited by Lars Schmitz, as part of BIOL 167 “Sensory Evolution”, an upper division class at the W.M. Keck Science Department. Written for educational purposes only].

Would you have ever guessed that the evolution of sensory systems may dictate your social behaviors? Well, this certainly seems to be the case for cavefish, Astyanax mexicanus. Many studies have shown that fish are quite social creatures. In fact, Anna Greenwood, a scientist in the Human Biology Division at Fred Hutchinson Cancer Research Center said, “the motivation to be social is common among fish and humans, where some of the same brain regions and neurological chemicals that control human social behavior are probably involved in fish social behavior.” The two types of social behaviors that A. mexicanus exhibit are schooling and shoaling. As part 1 of an amazing twin paper in Current Biology, Kowalko and colleagues (2013) distinguished between these two behaviors by defining schooling as the tendency of fish to synchronize their behavior and swim together, while shoaling is the tendency of fish to group with other fish of the same species.  Both these social behaviors provide a number of advantages such as deterring attacks from predators and foraging for food.

Figure 1. Beautiful Mexican blind cavefish. [By H. Zell (Own work) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons]

Figure 1. Beautiful Mexican blind cavefish. [By H. Zell (Own work) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0)%5D, via Wikimedia Commons]

In this study, several populations of A. mexicanus (Figure 1) were used to investigate the mechanism responsible for loss of schooling and shoaling behaviors: the sighted surface dwelling and three independently evolved blind cave-dwelling fish (Tinaja, Pachon, and Molino populations).

So, what exactly are the reasons behind the absence of schooling and shoaling behavior in cavefish? Kowalko and colleagues proposed four hypotheses behind this loss. For example, it may be due to the absence of large predators in caves, which minimizes the need for fish to travel close together in schools. In addition, food is fairly scarce in many caves, which may make these behaviors unfavorable. This loss could otherwise be due to the fact that cavefish have lateral line systems that are different from their surface dwelling counterparts. Lastly, the disappearance of this behavior may be due to genetic components that are independent of vision loss.

Kowalko and her fellow researchers first investigated the differences in schooling and shoaling behaviors in the two types of A. mexicanus. They measured schooling by recording the tendency of fish to follow a model school of plastic fish (Figure 2).  The surface fish followed the model school and the three independently evolved cavefish did not display any schooling behaviors.

Figure 2. Kowalko et al. designed this apparatus shown above to measure schooling behavior [from Kowalko et al., 2013]

Figure 2. Kowalko et al. designed this apparatus shown above to measure schooling behavior [from Kowalko et al., 2013]

Tests for shoaling were integrated in the tests for schooling where the researchers measured the average distance to the nearest fish. They found that surface fish swam closer together than any other cavefish population (Figure 3).

Figure 3. Shoaling was measured by quantifying the average nearest neighbor distance and average inter-individual distance. Results show surface swam closer together than any other cavefish. Thus, cavefish have lost the tendency to school or shoal. [from Kowalko et al., 2013]

Figure 3. Shoaling was measured by quantifying the average nearest neighbor distance and average inter-individual distance. Results show surface swam closer together than any other cavefish. Thus, cavefish have lost the tendency to school or shoal. [from Kowalko et al., 2013]

Surface fish show aggregating behavior even when they are raised in isolation, so there must be some kind of genetic determination of this behavior. When looking at the genetics of schooling behavior, the Kowalko team crossed surface fish with Tinaja cavefish to determine whether this trait was dominant or recessive. The F1 generation of fish portrayed that schooling is a dominant behavior. However, the F2 generation revealed fish with schooling and non-schooling behaviors indicating social behavior has a polygenic basis.

Cavefish have larger number and size of cranial neuromasts in comparison to surface fish (Yoshizawa et al., 2014). Kowalko and his fellow researchers used a similar reasoning to test their lateral line hypothesis by comparing the number of neuromasts to the proportion of time spent schooling and shoaling for both surface and cavefish. However, they failed to find a relationship between the two, concluding that cave environments do not have an effect on the evolutionary loss of these behaviors.

So, what is responsible for the loss of schooling and shoaling? To determine this, they tested whether loss of schooling would be seen in cavefish that lost visual function during development. Removing one lens from surface fish larvae produced a significant effect on shoaling behavior such that fish swam farther away from one another. When surface fish had both lenses removed, they schooled significantly less from those with one lens removed. Thus, it was concluded that vision is essential for schooling and shoaling behaviors (Figure 4).

Figure 4. Variation of lenses in surface fish (C-E) compared to cavefish (F). Lens removal resulted in less social behavior (G,H). [from Kowalko et al., 2013]

Figure 4. Variation of lenses in surface fish (C-E) compared to cavefish (F). Lens removal resulted in less social behavior (G,H). [from Kowalko et al., 2013]

To further investigate this finding, these researchers looked at visual function, defined as the ability to sense light. This was compared between the F1 generation, cavefish, and surface fish. The surface fish and the F1hybrids spent most of their time in the dark (avoiding the light), indicating their ability to detect light. On the other hand, the cavefish did not indicate a preference for the light or dark environment, indicating their inability to detect light. Many non-schooling fish have little visual function, as they display no light preference. However, some light detecting fish do not show schooling behavior, suggesting that there may be a visually independent component to schooling and shoaling.

Overall, vision seems to be the determining factor responsible for these behaviors in fish. This indicated that the evolution of sensory systems can contribute to the complexity of social behaviors in vertebrates. In this study, it was further suggested that two eyes are responsible for following the haphazard movement of fish, which is important for maintaining normal schooling and shoaling behaviors.

Although the researchers were effective in showing the importance of vision-dependent mechanisms in social behavior, there were some minor shortcomings to their investigation, leaving room for future studies. Previous work has shown that experimentally blinded surface fish exhibit schooling behaviors, which contradicts results found in this study (Partridge and Pitcher, 1980). This discrepancy suggests that further studies investigating visual-independent factors should be conducted to confirm either finding. For example, this study did not account for discrepancies in habitats or other selective mechanisms, such as the presence of predators or the scarcity of food. Furthermore, only Tinaja cavefish were tested when examining the effect of the lateral line system. It was also found last year that benthic sticklebacks (another really important species for understanding evolution; see Greenwood et al. 2013 — part 2 of the twin paper!) displayed differences in social behavior dependent on the evolution of the lateral line system, which wasn’t the case for A. mexicanus. Thus, it would be advantageous to conduct the same lateral line system test for more cavefish populations and other species.

References

Greenwood, A. K., Wark, A. R., Yoshida, K., Peichel C. L. 2013. Genetic and neural modularity underlie the evolution of schooling behavior in Threespine Sticklebacks. Current Biology 23: 1884-1888. (DOI: 10.1016/j.cub.2013.07.058)

Kowalko, J.E., Rohner, N., Rompani, S. B., Peterson, B. K., Linden, T. A., Yoshizawa, M., Kay, E. H., Weber, J., Hoekstra, H. E., Jeffery, W. R., Borowsky, R., T., Clifford J. 2013. Loss of schooling behavior in cavefish through sight-dependent and sight-independent mechanisms. Current Biology 23, 1874-1883. (DOI: 10.1016/j.cub.2013.07.056).

Partridge, B. L., Pitcher, T. J. 1980. The sensory basis of fish schools: roles of lateral line and vision. Journal of Comparative Physiology 135, 315-325. (DOI: 10.1007/BF00657647).

Yoshizawa, M., Jeffery, W. R., van Netten, S. M., McHenry, M. J. 2004. The sensitivity of lateral line receptors and their role in the behavior of Mexican blind cavefish (Astyanax mexicanus). Journal of Experimental Biology 217: 886. (DOI: 10.1242/​jeb.094599).

Posted in mechanosensation, sensory evolution blogs, vision | Tagged , , , , , , | Leave a comment