Hydraulic Fracturing: What is it and what does it mean for the environment?

Original title: A Greater Need for Research into Environmental Effects of Hydraulic Fracturing in the U.S. By Bryan White, December, 2014

Hydraulic Fracturing Schematic.

Hydraulic Fracturing Schematic. CC 3.0 by Mikenorton. Wikimedia Commons.

Hydraulic fracturing, a shale gas mining process which allows untapped gas resources to be harvested from ancient rock, offers an economic boon for the United States. However, the pathways through which hydraulic fracturing can extoll negative effects on the environment have remained largely uninvestigated. One pathway, groundwater contamination has the potential to contaminate streams and other fresh water sources above ground. Groundwater contamination can occur when wastewater, the liquid byproduct of shale gas well formation, is recovered and either reused or sent to a water treatment facility. Another pathway is the direct contamination of underground freshwater aquifers from fracturing fluids due to the seepage of liquid through both existing and newly formed cracks in shale rock. A third pathway, increased seismic activity can be triggered during the well injection process leading to felt seismicity (earthquakes felt by humans). The differences in potential that these pathways have for causing environmental harm will be explored in this paper, and the current status of experimental evidence demonstrating (or failing to demonstrate) environmental harm will be reviewed.

Hydraulic fracturing, otherwise known as “fracturing” or “fracking”, became an industry standard practice during the 1980”s, although shale gas plays (geologic contiguous areas where shale gas can be harvested from) have been mined since the early 1900”s (Curtis 2002, Vidic 2013). Fracturing is the process in which many cracks, fractures, are created in Devonion (~300 million years old) shale rock via the injection of a high pressure liquid slurry. Shale rock is a type of sedimentary rock that consists of compacted layers of mud, silt, and other organic matter accumulated over millions of years. This densely compacted material facilitates the generation of both thermogenic (generated via intense heat and pressure) and biogenic (microbially generated) methane (CH4) gas creation (Curtis 2002, Cokar 2013). The high pressures induced on the shale rock by this injected slurry causes many fractures the size and type of which are determined by the pre-existing tectonic conditions (Hubbert and Willis 1972). When fractures are formed by the injection, normally unattainable natural gas begins to seep out of the rocks, a process called desorption, and can be collected under negative pressure at a surface well. During well completion, a percentage of the fracturing fluids used to create the fractures can be recovered and either reused or treated.

Due to the proprietary nature of fracturing fluids, the specific contents of fluids are largely unknown. According to the Oil and Gas (http://www.ogj.com), slurry injected into shale rock consists of 90.21% water, 8.9% Proppant, and 0.44% other (Saba 2013). Proppant is a technical term for a mixture of water and either ceramic particles or sand grains. These fine, sand-sized particles help to keep fractured shale rock open so that gas can be freely desorbed and extracted along a negative pressure gradient. Other proprietary chemicals used in hydraulic fracturing can include a pH Adjusting Agent (0.01%), a Breaker (0.009%), a Cross-linker (0.006%), an Iron Controller (0.004%), a Corrosion Inhibitor (0.001%), a Biocide (0.001%), a Friction Reducer (0.08%), a Surfactant (0.08%), potassium chloride (0.05%), a Gelling Agent (0.05%), and Acid (0.11%) (Saba 2013).

The unknown content of fracturing fluids makes tracing their pathway through the fracturing process difficult and makes the direct detection of groundwater and aquifers difficult, so field tests using methane have been devised. A recent study by Osborn et al. (2011) aimed to indirectly test for the contamination of fracturing fluids by tracing the origin of detectable methane. Shale gas plays (geologic areas deemed economically valuable) can contain two types methane: biogenic methane derived from microbes and thermogenic methane derived from intense geologic processes (Curtis 2002). The ratios of these two types of methane gas vary from shale gas play to play (Curtis 2002), and the isotopic signature of each type of methane differs depending on its origin (biogenic or thermogenic, Osborn et al. 2011). However, the analysis of Osborn et al. (2011) lacked enough data points in critical areas (water wells near inactive fracturing wells) to prove that methane contamination had occurred. Jackson et al. (2013) expanded on the Osborn et al. (2011) field test. The improved data coverage of Jackson et al. (2013) suggests methane contamination is occurring and that the source of methane contamination is indeed from fracturing wells. However, since the presence of methane may be from naturally occurring geological processes, the specificity of this field test remains uncertain without pre-fracturing data, which both studies lack. This highlights the absolute need for pre-emptive sampling before hydraulic fracturing occurs. Furthermore, due to play-specific differences in biogenic and thermogenic gas signatures (Curtis 2002) it remains uncertain how these data can be applied outside of Appalachian Basin plays they were tested in.

Above-ground environmental contamination typically occurs following the injection of fracturing fluids when water is either recovered and reused on site or recovered and transported off-site where it undergoes water treatment for decontamination (Olmstead et al. 2013). Current recovery rates of fracturing fluids (flowback wastewater) have been estimated to be 10% with 90% of fluids remaining submerged (Rahm et al. 2013). Recovered waters can contain high levels of salts (Olmstead et al. 2013), as well as naturally occurring radionuclides (Olmstead et al. 2013). Along with drastically increased salt levels, total suspended solids can also be increased (Olmstead et al. 2013) to levels harmful to aquatic organisms. Two radionuclides found in hydraulic fracturing waste fluids are barium and radium which may bioaccumulate in stream sediments (Warner 2013). Radionuclides are atoms experiencing unstable nuclei and are prone to emit ionizing alpha, beta, and gamma radiation that bioaccumulates in the environment. Treatment of waste water successfully removes ~90% of radionuclides (Warner 2013). The bulk of the accumulation of contaminants output from above-ground wastewater treatment into streams occurs at or near where effluent is discharged into the environment. While it is clear that wastewater contaminants appear in effluent discharge and downstream of treatment plants, it is not clear whether salt, radionuclides, or TSS occur in toxic levels in those streams.

The occurrence of induced seismicity and earthquakes associated with hydraulic fracturing has been well documented (Hsieh and Bredehoeft 1981; Fehler et al. 1987), although occurrences of “felt seismicity”, earthquakes felt by humans, are actually quite rare. The first such account of felt seismicity occurred in 1960 when the U.S. military injected waste fluid into a borehole at the Rocky Mountain Arsenal, Colorado (Davies et al. 2013). Earthquakes generated by the injection of waste fluids caused earthquakes ranging up to 5.3M and caused significant structural damage in nearby towns and the use of the borehole was stopped in 1966 (Davies et al. 2013). More recent reports suggests that microseismic activity might trigger fault activation up to ~2k away from the site of hydraulic fracturing as fluid moves through permeable fracture fault systems (Holland 2013), which supported the conclusion of Rozhko (2010) regarding the mechanism of microseismic induction fluid diffusion. Davies et al. (2013) also concludes that hydraulic fracturing can activate faults and that the most likely method of fault activation is injection of fracturing fluids into pre-existing faults, although the actual magnitude of most earthquakes generated by hydraulic fracturing are less than 1 M on the Richter scale suggesting that the majority of seismic events are microearthquakes not felt on the surface. As the ranges of hydraulic fracturing expand, the risks associated with induced seismicity will also likely increase, potentially leading to a “point of no return” where natural faults have been permanently altered by fracturing activities.

In conclusion, hydraulic fracturing may present negative environmental impacts as a result of the bioaccumulation of salts and radionuclides, methane contamination of surface and subterranean aquifers, and an increase in geologic activity surrounding fracturing sites. A wide range of pollutants are hydraulically injected into wells, and injected water is rendered potentially toxic from naturally occurring elements (radionuclides). Most of the water injected into wells is recovered and either treated or reused in other wells reducing or eliminating effluent to sub-toxic levels, but little is known about current levels of potential environmental toxicity due to bioaccumulation. With new research and advances in fracturing technology potentially allowing completed wells to be modified into large subterranean bioreactors (Cokar 2013), the lifespan of wells may be increased drastically and any potentially negative environmental effects associated with well existence prolonged. The potentially harmful environmental effects associated with hydraulic fracturing warrants continued research into the how bioaccumulation of contaminants might occur, refining field tests for detecting well contamination, and understanding how increased seismic activity might permanently change the geologic landscape of an area.

Literature Cited

Cokar, M., Ford, B., Gieg, L. M., Kallos, M. S., & Gates, I. D. (2013). Reactive reservoir simulation of biogenic shallow shale gas systems enabled by experimentally determined methane generation rates. Energy & Fuels, 27(5), 2413-2421.
The authors performed laboratory experiments to determine the rates of natural gas produced by microbes under conditions similar to shale gas reservoirs. Methanogenesis was simulated in the laboratory by using core rock and water samples taken from shale gas reservoirs. Interiors of cores from different depths (ranging from 250 m to 350 m) were crushed and incubated to remove any surface-dwelling microbes, then inoculated with fresh water obtained from a shale reservoir. Kinetic rate equations for gas production were derived from the inoculated samples giving a rate equation for each depth. Microbes inoculated on different depths produced markedly different gas outputs but there was not a pattern associated with increased or decreased depth. Model parameters derived from rate experiments were input into a mathematical model to predict gas output for a 2500 day period (6.8 years) from a shale gas reservoir which closely matched output from an actual reservoir. The authors conclude that in active shale gas wells biogenic gas makes up roughly 12% of extracted gas. Biogenic gas output might be increased so long as electron donors (H2 or acetate) are present in sufficient quantities.

Curtis, J. B. (2002). Fractured shale-gas systems. AAPG Bulletin, 86 (11), 1921-1938.
This paper covers a brief history of shale gas mining followed up by a more detailed overview of 5 current shale-gas plays in the United States. The author also overviews current economic status of each shale-gas play, geological and geochemical composition, and the amount of gas that is economically feasible to recover. The concepts of biogenic vs. methanogenic gas are introduced and discussed.

Davies, R., Foulger, G., Bindley, A., & Styles, P. (2013). Induced seismicity and hydraulic fracturing for the recovery of hydrocarbons. Marine and Petroleum Geology, 45, 171-185.
This article focuses on the “felt seismicity”, earthquakes felt by humans, generated by hydraulic fracturing and injection mining. The authors began by a general introduction to how earthquakes can be caused by the injection of fluids, either fracturing fluids or other waste water, into boreholes. Induced earthquakes are typically caused by the progressive loading of stress from shifting tectonic plates, but any type of progressively changing stress can cause earthquakes. Typically faults are lubricated either by magma or water which allows the fault to overcome the friction on the fault plane. The injection of fracturing fluids increases stress on faults, can create new cracks and fractures in the rock surrounding faults, and makes it easier for fault slips to occur where there already exists pressure. The authors sought to understand the causes for the largest sized (highest magnitude) earthquakes. They conducted a meta-study consisting of 198 seismic examples from 66 published papers. They broke down the likely triggers of seismicity into specific subcategories and attributed each earthquake to a particular category. The authors found that hydraulic fracturing could trigger faults but the majority of published examples of induced seismicity were from conventional mining activities. Induced seismicity due to hydraulic fracturing also produces only very small earthquakes (~3 M) that are not likely to be felt by humans. However, this does not rule out the possibility that hydraulic fracturing could re-activate a larger fault and cause felt seismicity (~4 to 8 M).

Fehler, M., House, L., & Kaieda, H. (1987). Determining planes along which earthquakes occur: method and application to earthquakes accompanying hydraulic fracturing. Journal of Geophysical Research: Solid Earth (1978-2012), 92(B9), 9407-9414.
Provides a statistical method for finding the specific fault plane of an earthquake caused by hydraulic fracturing. They used this method to find the fault plane of microearthquakes and determined that those microearthquakes were caused by hydraulic fracturing.

Holland, A. A. (2013). Earthquakes Triggered by Hydraulic Fracturing in South-Central Oklahoma. Bulletin of the Seismological Society of America, 103(3), 1784-1792.
A case study report of 116 microearthquakes (0.6 to 2.9 M) that occurred near a well undergoing hydraulic fracturing in south-central Oklahoma. Authors found a very strong temporal correlation between fracturing events and seismic events. Interestingly the authors did not find a strong correlation between injection volume and earthquake magnitude. Authors also note that the majority of wells in Oklahoma have not been suggested to undergo any induced seismicity.

Hubbert, M. K., & Willis, D. G. (1972). Mechanics of hydraulic fracturing. Society of Petroleum
Engineers of AIME, 210, 153-168. Develops detailed physical model of induced seismicity due to hydraulic fracturing. The extent and type of fractures caused by fluid injection is largely dependent on the pre-existing tectonic stress of an area.

Jackson, R. B., Vengosh, A., Darrah, T. H., Warner, N. R., Down, A., Poreda, R. J., Osborn, O. G., Zhao K., & Karr, J. D. (2013). Increased stray gas abundance in a subset of drinking water wells near Marcellus shale gas extraction. Proceedings of the National Academy of Sciences, 110(28), 11250-11255.
This paper is a follow-up experiment from the Osborn 2011 paper. The authors collected water from shallow water wells used for drinking in Pennsylvania and traced the source of any methane found in those wells. The authors again use the method of assigning methane origin to shale gas wells (hydraulic fracturing) if it was geochemically produced thermogenic methane. Methane was assigned to non-fracking sources if it was found to be biogenic (originating from methanogenic microbes). The authors found dissolved methane in 82% of water wells sampled. They also found that the water samples with the most elevated methane levels had the most thermogenic origin, suggesting fracturing was the source of methane contamination. Ethane, propane, and helium isotopic analysis was also done and supported isotopic origins similar to those found in shale gas reservoirs for those isotopes as well, although sample sizes were much lower for the other isotopes. The authors concluded contamination was occurring in the Marcellus shale area and that the most likely sources of contamination are faulty steel casings and inadequate cement sealing.

Olmstead, S. M., Muehlenbachs, L. A., Shih, J. S., Chu, Z., & Krupnick, A. J. (2013). Shale gas development impacts on surface water quality in Pennsylvania. Proceedings of the National Academy of Sciences, 110(13), 4962-4967.
The authors conducted a statistical analysis using GIS software and the coordinates of sampled surface water locations in relation to fracking operations. To estimate surface water contamination previously collected data on Cl- and total suspended solids (TSS) were analyzed. Cl- is used during the fracking process and can harm aquatic ecosystems by increasing the amount of dissolved heavy metals or phosphates in water. TSS are also harmful to aquatic ecosystems and can clog and block out sunlight, increase temperatures, and reduce available dissolved oxygen. The authors found that the density of shale gas wells present upstream from a watershed did not statistically increase Cl- concentrations. However, increased density of water treatment facilities that receive fracking fluids did. This suggests that for Cl- levels increased Cl- in watersheds is a result of treated water rather than shale gas well contamination. For TSS, the effects of well pads (5-15 acre industrial sites surrounding a well head) were found to significantly increase downstream TSS levels. This suggests runoff from well heads are contributing to stream TSS, although the specific mechanism contributing to increased TSS is unclear.

Osborn, S. G., Vengosh, A., Warner, N. R., & Jackson, R. B. (2011). Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proceedings of the National Academy of Sciences, 108(20), 8172-8176.
Authors hypothesize that fracturing fluids injected into wells may leak into nearby freshwater aquifers and cause contamination. Contamination from fracturing fluids could increase the amounts of radioactive materials and toxic gases in fresh groundwater and pose a human health risk. The authors analyzed water samples from 68 wells for methane concentrations, dissolved salts, and dissolved isotopes. They sampled water from two treatments: active, a water source within 1 km of an active fracking well, and inactive, a water source greater than 1 km from a fracking well. The authors detected methane in 85% of drinking-water wells sampled and methane concentration increased nearer to gas wells. Furthermore, the authors found that the methane detected in drinking-water wells had a radioactive isotopic signature more closely related to deep, thermogenically derived methane rather than shallower, microbially generated biogenic methane. However, there are some issues with the representation of the data presented here. For example, Figure 3 shows many of the water samples taken nearest to active wells have higher methane concentrations, however, no samples were taken where a drinking-water well was nearest to a non-active gas well. In order to conclude shale gas wells are the source of increased methane, drinking sources near non-active wells (less than 1000 m) need to be sampled.

Rozhko, A. Y. (2010). Role of seepage forces on seismicity triggering. Journal of Geophysical Research: Solid Earth (1978-2012), 115(B11).
The author develops a mathematical model based on poroelasticity to demonstrate microseismic events occur due to seepage (the diffusion of injection fluids) into boreholes. The author validated his model against published data from both hydraulic fracturing injection wells and a geothermal injection well. In the hydraulic fracturing well, microseismic events lasted about 5 hours following injection, 1 hour after injection stopped, and traveled about 600 m. Injection into the geothermal well lasted about 60 hours and an additional 20 hours after injection had stopped. Microseismic events also only occurred roughly 600 m from the borehole at the geothermal well site. The nonlinearity of microseismic events is due to the diffusion of fluid and elastic stress response of rock. This model suggests that well depletion, the removal of fluid pressure, would also cause microseismic events.

Saba, T. (2013). Evaluating claims of groundwater contamination from hydraulic fracturing. Oil & Gas Journal, 111(7).
This an article published in the Oil and Gas Journal, which is not a peer-reviewed journal. However, it does include detailed statistics from sources active in the gas and oil industry and so is relevant to this review. Unfortunately peer-reviewed articles containing detailed statistics of the ingredients used in fracking fluids are not currently available. It is important because these are the values and names of chemicals that the gas and oil industry feels comfortable publishing to the public. Although many of the chemicals listed in this article are proprietary names and the exact chemical compounds are unknown, it does give some insight into the chemicals used during the process of well injection.

Rahm, B. G., Bates, J. T., Bertoia, L. R., Galford, A. E., Yoxtheimer, D. A., & Riha, S. J. (2013). Wastewater management and Marcellus Shale gas development: Trends, drivers, and planning implications. Journal of Environmental Management, 120, 105-113.
The authors summarize the economic benefits of hydraulic fracturing and note that interest is increasing internationally and that until unconventional (hydraulic fracturing) gas mining was implemented in the United States, gas production was on the decline. The authors also state that while hydraulic fracturing has many economic benefits, it also has environmental negatives associated with fracking wastewater. The authors investigated how the wastewater management process has changed over time with regards to hydraulic fracturing fluids. Data for this study were obtained from Pennsylvania Department of Environmental Protection (PADEP). Overall, the volume of wastewater resulting from fracking in the Marcellus shale increased roughly 5 fold from 2008 to 2011 due to increased well drilling. However, much of the volume has shifted its destination from being treated and returned to groundwater systems to being reused by well operators for new wells. Disposal of wastewater by injecting it into wells and leaving it there (injection disposal) also increased. Well drilling tended to decrease as the prices of natural gas decreased from $4 per million cubic feet from 2010 to $2 in 2012. Wastewater was treated or reused in over 70 counties and the travel distance for wastewater disposal decreased by 30% most likely associated with increased treatment infrastructure.

Warner, N. R., Christie, C. A., Jackson, R. B., & Vengosh, A. (2013). Impacts of shale gas wastewater disposal on water quality in Western Pennsylvania. Environmental Science & Technology, 47(20), 11849-11857.
The authors analyzed effluent discharged from a brine treatment facility near the Marcellus shale in Pennsylvania and sought to understand the short and long-term environmental effects of unconventional (hydraulic fracturing) gas drilling. Effluent (water initially leaving the treatment facility) samples were collected and also surface water samples from both upstream and downstream from the treatment facility. Samples were collected over the course of 2 years (2010 to 2012). The authors found major salt elements were up to 6,700 times higher concentrations in effluent discharge compared to upstream river sites. The treatment plant was found to contribute 78% of downstream salt (chloride) flux, mean annual salt concentration multiplied by total annual effluent discharge volume. Barium and radium elements (both radionuclides) were reduced by 99% following water treatment. However, sludge produced during the treatment process would theoretically contain radium levels higher than safe disposal regulations allowed in the U.S. Furthermore, although radium discharge was significantly decreased, much of the radium that was discharged would accumulate in sediments near the discharge site and amplified by increased salt levels. These levels run the risk of causing bioaccumulation in benthic invertebrates, fish, and even plants.

Hsieh, P. A., & Bredehoeft, J. D. (1981). A reservoir analysis of the Denver earthquakes: A case of induced seismicity. Journal of Geophysical Research: Solid Earth (1978-2012), 86(B2), 903-920.
Analysis of data from 1960’s injection of wastewater fluids by U.S. Army Corps of Engineers at the Rocky Mountain Arsenal in Denver, Colorado. Author concludes that earthquakes were caused by the pressure build-up of fluid injection. Some earthquakes exceeded magnitudes of 3 or 4 on the Richter scale. Earthquakes continued after fluid injection stopped and caused structural damage in the Denver area.

The Evolution of Eusociality in Insects

Epigenetics in Social Insects: A New Direction for Understanding the Evolution of Castes

Originally written April, 2012 by Bryan White

Article 1 Source: https://www.hindawi.com/journals/gri/2012/609810

Epigenetics is a new field of biology that deals with an only recently discovered method of DNA inactivation called DNA methylation. DNA methylation is the process in which sections of DNA are methylated and primarily occurs on cytosines, although they could occur on any nucleotide. In this paper, the current state and understanding of DNA methylation and how it relates to the development and evolution of insect castes (particularly in the eusocial insect groups) is reviewed. Methylation is not the only possible epigenetic mechanism. DNA acetylation (the addition of acetyl molecules) is also possible, as well as ubiquitination (the addition of the ubiquitin protein).  However, DNA methylation is probably the most common. The end result of DNA methylation is the existence of a secondary language on top of the DNA language that can be modified by environmental factors, can be passed on to the next generation, and influence the development of offspring. DNA methylation can also have an evolutionary affect by increasing the rate of mutations in genes that are methylated for multiple generations, for genes that are inactivated can accumulate stop codons and other deleterious mutations. Based on this, the authors hypothesize that DNA methylation is potentially the primary method for caste selection in eusocial insects.

Epigenetics brings a whole new aspect to the table for understanding how castes evolved, and how castes are regulated (should a larva develop into a queen or worker?) in eusocial systems. In hymenopteran eusocial species, there is typically a vast amount of physical diversity amongst castes (workers, soldiers, queens and male drones), and workers have found it hard to explain this diversity using only genetic methods. This is largely due to the fact that it is well known that the development and selection of what a larva will develop into is environmentally based, but scientists do not have a clear idea of exactly how that developmental “decision” is enforced. Epigenetics stands as a good explanation for how environmental factors can influence larval development, and the authors suggest this probably carried out by the presence of DNA methylation genes such as DNMT3, coincidentally which Drosophila is lacking and so was thought unimportant. The direct connection between the expression of DNMT3 and the genes that are methylated is a new, expanding area of research.

Another one of the difficulties in understanding the evolution of eusociality has been trying to explain its evolution in terms of kin selection, specifically that haplodiploid species exhibit on average 75% more genetic relatedness of sisters than other species. The benefits of a haplodiploidy system as an example of kin selection theory were that it provided a strict means for both the regulation of sexual dimorphism (males are made up of only the queen’s genome) and suggested some involvement in the development of castes. However, epigenetics and DNA methylation offers a much better explanation for the existence of both large amounts of sexual dimorphism and phenotypic plasticity. DNA methylation has been found in many eusocial hymenopteran species, as well as primitively social hymenopterans, suggesting that DNA methylation is both a heavily conserved trait and is correlated to sociality, phenotypic plasticity and sexual dimorphism. Better understanding the phylogenetic location of insect groups that make use of DNA methylation can probably elucidate the question as to whether or not DNA methylation is the sole (or primary) source of caste determination.

The authors also attempt to lay out a conceptual framework for future studies, however I found their model unclear. What the authors seem to suggest is that eusociality is correlated with DNA methylation, but not a requirement. They do, however, do a good job outlining the specific areas of DNA methylation that need to be explored and understood to eliminate other possible explanations for the correlation between DNA methylation and eusociality, such as understanding the mechanistic effects that DNA methylation has on gene splicing and whether or not it is possible for eusocial insects to exhibit caste differentiation without DNA methylation genes.

Article 2 Source: http://www.nature.com/nature/journal/v466/n7310/full/nature09205.html

In this article the researchers hypothesize that up until this date, all progress on kin selection theory has largely been abstract in nature and not provided any concrete evidence for the theory. They argue that, in order for kin selection theory to be fulfilled in an empirical system, several stringent conditions must be met.

First, all interactions that are measured must be “additive and pairwise”, that is, they must only affect the pair of individuals involved in the interaction. This means that synergistic effects, such as the simultaneous cooperation of more than two individuals, are unable to be measured or incorporated into any mathematical model of kin selection.

Second, they argue that kin selection theory can only be applied to a very limited subset of population structures due to the requirement of global updating of interactions wherein global updating is the idea that any two individuals are competing uniformly for reproduction regardless of their geographic proximity to each other.

Third, they argue that if these two requirements are met, and they can only be met in some limited, artificial world, then when these requirements are met that the organismal interactions within that aforementioned world are also acting according to the conditions of natural selection theory, and that kin selection theory does not provide any additional biological information.

Finally, the authors also argue that the apparent simplicity of kin selection theory compared to that of natural selection theory is an illusion. Since the primary component of kin selection theory is the calculation of inclusive fitness, and the calculation of inclusive fitness requires the state of “all individuals whose fitness is affected by an action, not only those whose payoff is changed” to be known, then in effect kin selection theory is requiring the same information to be known as natural selection theory the state of all individuals affected rather than only those whose payoff (fitness) is increased.

In order to overcome the limitations imposed by kin selection theory, the authors propose a general, multi-level model of natural selection theory using only the general principals of population genetics. This model is used to explain how eusociality might evolve in five distinct evolutionary stages.

First, an organism must reach a state where there are clear groups within a population. Groups typically form around resources, nest sites, when parents and offspring stay together, or when flocks go to known breeding grounds.

Second, these groups begin to accumulate traits, otherwise known as pre-adaptations, that will increase the overall cohesion and cooperation of these groups. One such pre-adaptation is when a parent places large numbers of paralyzed prey around her eggs so that when the eggs hatch they will have a food source readily available, and then she moves on to create another nest. The next step towards eusociality would be for the parent to stay near the nest and guard the eggs until they are hatched. However, at this stage, the offspring will still leave the nest and so will the parent –  there is still dispersion.

Third is the evolution of clearly eusocial alleles, that is, traits that enforce the primary traits of eusociality. The key traits here are for individuals to stay in the nest instead of dispersing, and then other cooperative pre-adaptations can come into play.

Fourth is probably what can be called the optimization stage in which these eusocial alleles can be selected upon to reinforce the nest/colony structure.

Fifth is the final phase and selection now operates on the colonies instead of the individual organisms, and the evolution of more derived traits such as castes (workers/soldiers), fungal farming, aphid farming, and other highly cooperative activities. Here the authors have outlined the framework through which future studies can be conducted, most likely which will be a combination of behavioral ecology and phylogenetics. My criticisms of this paper can only be restricted to the authors’ use of the words “primitive” and “advanced”, which are common misnomers in evolutionary biology. A better term should be less derived or more derived, in reference to the ancestral state. For instance, the caste system of most ants is more derived compared to the loose grouping structure of some wasps.

Comparing structural and functional elements of orthologous HSP70s in the fission yeast Schizosaccharomyces pombe and the budding yeast Saccharomyces cerevisiae

This is a research article I did on the heat shock proteins of two species of yeast in 2013.


Seventy-kD Heat Shock Protein (HSP70) is a multigene family of proteins that is important for cellular stress response and survival (Lindquist 1988). The HSP70 proteins are approximately 70 kDa in size and are highly conserved across all three domains of life (Eukaryotes, Prokaryotes, and Archaea). These genes are either constitutively expressed or heat inducible (Lindquist and Craig 1988). HSP70s are a family of ATPases that contain an N-terminal Adenosine Triphosphatase domain (aka. nucleotide binding domain, NBD), a substrate binding domain (SBD), and a C-terminal domain of varying length. These proteins are involved in the transport of proteins across membranes as well as protein folding in a cell (Hartl and Hayer-Harlt 2002). HSP70s’ role in protein folding is important in cell survival during heat shock stress. Higher temperatures can lead to protein misfolding and subsequent aggregation within the cell. HSP70s bind denatured or abnormal proteins via the exposed hydrophobic regions to prevent aggregation (Finley et al. 1984). Binding of these proteins also facilitates refolding into the proper conformation (Wegele et al. 2004). The structure and function of HSP70s are well studied in the yeast Saccharomyces cerevisiae.

Saccaromyces cerevisiae is a single-celled fungus that is used in applications such as beer brewing and bread making. Because the organism has important commercial uses, it has been subject to extensive study. S. cerevisiae has also been used as a model organism to study the function and structure of eukaryotic cells. Like other organisms, S. cerevisiae contains many HSP70 genes. There have been a total of 14 HSP70 genes discovered that are grouped by sub-cellular location. SSA1-4, SSB1-2 and SSE1-2 are HSP70s that reside in the cytosol (Lindquist and Craig 1988; Mukai et al. 1993); SSC1, SSQ and ECM10 are mitochondrial HSP70s (Voos et al 2002); and Kar2 and LHS1 reside in the endoplasmic reticulum (Normington et al 1989; Saris et al 1997). Although some hsp70 genes are heat-inducible, not all in the family share the same expression profile. Previous studies have shown that SSA2 expression is not temperature based, while the SSB proteins had decreased expression when temperature was increased (Craig et al. 1985).

Schizosaccharomyces pombe is a basal member of the fungi phylum Ascomycota, as are the rest of the subphylum Taphrinomycotina (Ebersberger 2012), although Taphrinomycotina may not be a monophyletic grouping (Schoch 2009). Unlike other Ascomycotes who reproduce by producing ascospores, Schizosaccharomyces divide by medial fission (Nurse 1976), hence Schizosaccharomyces are known as the “fission yeasts”. S. pombe was originally used as a component in the traditional African sorghum beer “pito” in Ghana (N’guessan 2011) and was not used as a scientific model organism until 1950 (Leupold). Previously, S. pombe has been used as a model organism for various genetic studies (Mitchison 1970, Gutz 1974, Beach and Nurse 1981, Hagan and Hyams 1988, Matsuyama 2006, Kim 2010), although it has not been used as widely as a model organism as S. cerevisiae due to its lack of easily controllable gene expression methods (Zilio 2012), although some recent progress has been made on developing effective methods of gene control that do not induce cellular stress (Zilio 2012).

The complete genome of S. cerevisiae was published in 1996 by Goffeau et al, and the genome of S. pombe was published by Wood et al. in 2002. Interestingly, the genome of S. cerevisiae is marked by a whole genome duplication event that led to the duplication of many genes (Wolfe 1997, Kellis 2004). Comparisons of S. pombe and S. cerevisiae are vital in understanding how HSP70 maintain similar functionality across great timespans as these two species likely diverged around 425 million years (Berbee et al. 2007). Interestingly, these two species have drastically different genomic arrangements. S. cerevisiae maintains 16 chromosomes and only 250 introns, S. pombe maintains only 3 chromosomes, but thousands of introns, which suggests these two species are experiencing very different selection pressures on a genomic-scale. If these two species have been experiencing different genomic-scale selective pressures, we would expect that HSP70s might have been shuffled around and undergone significant sequence divergence, yet still remained functionally the same. Specifically, we hypothesized the following: 1.) The presence/absence of regulatory elements (HSE, intron/exon) has been unchanged. 2.) Amino acid sequences have diverged significantly (~5%). 3.) Presence/absence of signal peptides or transmembrane proteins remained unchanged. 4.) Local gene neighborhood synteny has been lost. 5.) 3-dimensional structures have remained unchanged. 5.) The nucleotide binding site has remained functionally unchanged in the lhs1 HSP70 orthologs.


Sequence collection and ortholog detection

Orthologous sequences were detected by first obtaining known HSP70 sequences from S. cerevisiae S288 and then searching known yeast sequences against the S. pombe genome using BLASTp. We considered proteins that had greater than 25% identity to each search sequence to be potential orthologous sequences. We retrieved those potential orthologs and constructed a preliminary tree to make sure that all possible orthologs had been found in S. pombe. Orthologs that had not yet been found in S. pombe but were found in S. cerevisiae were then searched against the genomic sequence of S. pombe using the corresponding S. cerevisiae protein sequence in tBLASTn.

Phylogenetic analysis

S. cerevisiae and S. pombe protein sequences were aligned online using the MAFFT program (Katoh 2013) withG-INS-i parameters in order to achieve an optimal global alignment. Ortholog detection was done using the neighbor joining method and bootstrap method with pairwise p-distances of the amino acid (AA) sequences in the MEGA5.1 program (Felsenstein 1985; Saitou and Nei 1987; Tamura 2011). The best distance model for the protein data set was determined by using ProtTest 3.2 (Darriba 2011) and modeled AA trees were inferred using PhyML 3.0 and MEGA 5.1.

Following AA based tree drawing, the AA alignment was converted codon-by-codon to an aligned, genomic CDS alignment using a lookup table so that the resultant nucleotide alignment matched the AA-based alignment. The best nucleotide model for this data set was determined using jModelTest (Guindon and Gascuel 2003; Posada 2008). Following model selection, maximum likelihood phylogeny was inferred using the PhyML 3.1 standalone version (Guindon 2010) with 100 bootstrap replicates. The amino-acid based tree and nucleotide-based tree were compared for topological differences.

Exon-Intron Analysis

The NCBI gene database for S. cerevisiae and S. pombe was used to identify exons and introns within hsp70 genes. Genes with introns were analyzed using SPIDEY software to determine position of splice points in genomic DNA and intron phases.

Regulatory elements analysis

TransFac Match program was used to search for the presence of Heat Shock Elements (HSE) 1000 base pair (bp) upstream of each hsp70 coding sequence. A matrix and core match score above 85% was used to represent strong evidence for a HSE with thenucleotide motif nGAAnnTTCnnGAAn.The sequences GAAnnTTC or TTCnnGAA were also considered for possible HSEs.

Synteny Analysis

Gene maps of each hsp70 were obtained through NCBI database. Maps were used to compare the orientation of each gene in their respective genomes. Conservation of neighboring genes were also compared.

Protein characterization

SignalP 4.1 server software was used to predict possible signal peptides in HSP70 amino acid sequences using eukaryotic organism setting. TMHMM 2.0 server software was used to predict whether the HSP70 proteins were membrane bound by detecting the presence of transmembrane helices. Presence of conserved protein domains in HSP70s were searched for using Superfamily HMM library and genome assignments server version 1.75.
Nucleotide binding site analysis

In order to determine if HSP70 proteins were functionally different or merely differed in their nucleotide sequences, the 3-dimensional structures of a subset (the lhs1 genes) were analyzed. The nucleotide binding site (NBS), a conserved site in the nucleotide binding domain (NBD) found in all hsp70 genes, was targeted. This region is responsible for binding ATP which allows the opening of the substrate binding domain (SBD), and so its function should remain highly conserved through time, particularly the NBS site which itself is a cleft within the NBD where ATP binds (Liu 2007).


Orthologous sequence detection

We found a total of 14 hsp70 family proteins in S. cerevisiae, but only found 8 orthologous sequences in S. pombe (Table 1). The ssa gene group was marked by a paralogous duplicate in S. cerevisiae not found in S. pombe (Figure 1), as well as in the ssb, sse, and mitochondrial groups (Figure 1). All gene names will be referred to as shown in Figure 1. For each of the cellular regions that Hsp70 proteins were found in S. cerevisiae, S. pombe also had at least one ortholog. However, S. cerevisiae produced a markedly greater number of paralogs, although it maintains the same number of orthologs.

Phylogenetic analysis

The best model for the amino acid data set was found to be LG+I+F+G, although a JTT based NJ tree was also inferred for comparison. GTR+I+G was found to be the best model for the genomic CDS nucleotide data set. In the AA based data sets, several notable topological differences between NJ, p-distance based tree and the NJ, JTT distance based AA data set, as well as between both NJ methods and the maximum-likelihood based LG+I+F+G based tree, although these differences were not usually supported with strong bootstrap support. There were no topological differences between the maximum-likelihood, AA based tree and the maximum-likelihood, nucleotide based tree.

Differences between phylogenetic methods were present but subtle, and not usually strongly supported (greater than 95% support) by bootstrap analysis. The fact that the maximum-likelihood trees inferred using the best amino-acid model (LG+I+F+G) and the best nucleotide model (GTR+I+G) produced identical topologies suggests that these trees are correctly representing the evolutionary history of these heat-shock proteins.

Intron analysis of hsp70s

NCBI gene database revealed no introns within S. cerevisiae hsp70 genes. This was also the case in S. pombe except for pdr13-pombe, which contains two exons and one intron. SPIDEY analysis of pdr13-pombe revealed a 126 nucleotide phase two intron between positions 125-126 of the mRNA. This is compared to the ortholog ssz1-yeast in S. cerevisiae which contains no intron (Figure 3). The intron of pdr13-pombe is a phase 2 intron located fairly close to the start codon of the mRNA between positions 125-126.

Regulatory elements analysis

TransFac analysis revealed full HSE regions containing the GAAnnTTCnnGAA sequence in ssa1-yeast, kar2-yeast, ssa1-pombe, and ssa2-pombe genes. Both ssa1 genes from S. cerevisiae and S. pombe contained HSEs approximately 300 bp upstream of the start codon (Figure 4). All other hsp70 genes contained no HSEs or a Partial HSEs with either a GAAnnTTC or TTCnnGAA motif. The ssa2-yeast, ssa4-yeast, ssb2-yeast, sse1-yeast, sse2-yeast, bip-pombe, ssc1-pombe, pdr13-pombe, and pss1-pombe genes all contain partial HSEs. Ecm10-yeast, lhs1-yeast, ssa3-yeast, ssb1-yeast, ssc1-yeast, ssz1-yeast, ssq1-yeast, lhs1-pombe, and sks2-pombe contained no HSEs. The mitochondrial ssc1-pombe contains a partial HSE approximately 200 bp upstream of the start codon. This is compared to the ortholog ecm10-yeast which contains no HSE. Another example of this pattern is seen in the intron-containing pdr13-pombe and ssz1-yeast genes. The endoplasmic reticulum kar2-yeast was found to have a full HSE compared to its ortholog bip-pombe. The sse2-yeast and pss1-pombe both contain two partial HSEs (Figure 4).

Synteny Analysis

Comparison of gene maps for ssa1 and ssa2 orthologs revealed few syntenic relationships (Figure 5). Though ssa1 genes showed the same orientation, ssa2 genes were found to be in opposite orientation. There were no similarities in the neighboring genes for between ssa1 and ssa2 orthologs. The pattern of no syntenic relationships was observed for the other hsp70 genes (data not shown).

Protein Characterization

Lhs1-yeast, Lhs1-pombe, and Bip-pombe were predicted to have signal peptides located in their amino-terminus region. The Bip-pombe ortholog Kar2-yeast was not predicted to have any signal peptides (Figure 6). SignalP software identified possible cleavage sites for Lhs1-yeast and Lhs1-pombe between amino acid position 20-21 and position 21-22 respectively. Cleavage site for Bip-pombe was predicted to be between position 24 and 25. All other HSP70 sequences were not predicted to have a signal peptide. Bip-pombe, Kar2-yeast, Lhs1-pombe, and Lhs1-yeast sequences contained possible transmembrane regions in their N-terminus regions (Figure 7). Bip-pombe had transmembrane helices at position 7-24. Kar2-yeast and Lhs1-yeast both contained helices at position 7-29. Lhs1-pombe contained helices near the N-terminus region with a 0.57 possibility score. All other proteins were predicted to be non-membrane bound. Superfamily web database search revealed conserved ATPase domains at the N-terminus of ssa1-yeast (Figure 8). Ssa1-yeast also contained a HSP70 domain as well as the C-terminal HSP70 domain. This pattern was seen in nearly all other HSP70s. The notable exceptions were Lhs1-yeast and Lhs1-pombe. Lhs1-yeast lacked the HSP70 domain while Lhs1-pombe was missing both the HSP70 and HSP70 C-terminal sub-domain.

Nucleotide binding site analysis

We found that overall, the structure of LHS1 in both S. cerevisiae and S. pombe were nearly identical, and most visual differences were located on the loop portions of the amino acid sequence (Figure 9). When looking at differences between the NBS in both species, we found did differ in its amino acid composition, but structurally the two sites were also nearly identical. The differences in amino acid sequence may result in functional differences between these two proteins as several amino acids frequently changed in charge and polarity (Figure 10) between these two species.


S. cerevisiae and S. pombe are two fungi separated by millions of years of evolution as well as a complete genome rearrangement, and exhibit markedly different life histories. S. cerevisiae is a budding yeast and so it is likely under less selective pressure for rapid DNA replication and maintains 16 chromosomes. Conversely, S. pombe is likely under great selective pressure for its DNA to rapidly condense to chromosomes, replicate, and then quickly separate to the poles so that fission can occur, and so maintains only 3 chromosomes. This replication strategy can leave S. pombe vulnerable during its replication process, whereas S. cerevisiae is relatively unaffected. We anticipated that these different selective pressures might have caused large differences in the location of hsp70s in both of these species genomes, but the actual proteins themselves maintained structural and functional similarities.

When we compared the homologous HSP70s of S. pombe to S. cerevisiae, we found that S. cerevisiae usually contained more hsp70 genes than S. pombe. In some cases, for instance, in the mitochondrial based hsp70s, the divergences between both paralogs and orthologs are high. It is likely that these mitochondrial HSP70s represent ancient origins. However, it’s uncertain whether or not these genes arose before the genome duplication in S. cerevisiae, as we cannot determine whether or not they were lost in S. pombe or gained in S. cerevisiae from the data presented here. We can say though that since the divergences between these mitochondrial HSP70s are high, that it’s not likely they are undergoing concerted evolution. Conversely, for example, in the SSA complex of HSP70s, paralogous pairs of HSP70s appear to be undergoing concerted evolution, as paralogs always exhibit almost zero sequence divergence, although we cannot eliminate the possibility that each of these paralogous pairs arose only recently with the data presented here. In order to better answer these questions more fungal species must be added to the protein family tree of orthologs.

We found that, after modeling the evolutionary distances, the HSP70s of these two species had undergone large amounts of amino acid sequence divergence. In some cases, as in the LHS1 orthologous pair, the pairwise distance exceeds 100%. These findings suggest that these genes have undergone large amounts of mutations in the same locations (multiple hits), and that it is possible these distances have undergone saturation so that no more additional phylogenetic information might be garnered from comparing these sequences. This suggests that even though HSP70 genes are conserved across all three domains of life, they might actually be poor phylogenetic markers.

After determining that large amounts of sequence divergence had occurred, we then investigated whether or not these significant changes in amino acid sequence were associated with actual functional changes in these proteins. We did not find any significant differences in intron/exon structure save for pdr13-pombe. The presence of a single intron in pdr13-pombe could be attributed to S. pombe having nearly 20 times the number of introns compared to S. cerevisiae. One could reasonably expect that the large number of introns would increase the possibility of gaining introns in hsp70 genes even if the ancestral state was to not have introns. It is also possible that the intron corresponds to the regulatory sequence for a neighboring gene due to the compact nature of the S. pombe genome. Again, the addition of more fungal species to this analysis is necessary to truly assess the ancestral state of introns/exons in these two species and decide whether or not the similarity in the current state is due to convergent evolution or homology.

Analysis of the HSEs in each gene revealed surprising results. The ssa1 gene orthologs contained full HSEs which contradict the literature stating they are not heat-inducible. Ssa2-yeast is not heat-inducible as well. However, ssa2-pombe contains a full HSE. It is interesting to note that while the homologs ssa3-yeast and ssa4-yeast contain partial HSEs, they have been found experimentally to be heat inducible nevertheless (Boorstein and Craig, 1990; Werner-Washburne et al, 1989). It is possible that ssa3-yeast and ssa4-yeast retained its heat-inducible nature by having a sequence that was still sufficient enough for binding. It is also suggested that due to S. cerevisiae having 4 ssa homologs, ssa1-yeast and ssa2-yeast might have lost its heat-inducible ability over time due to the other genes being able to compensate for this loss. It is possible that ssa1-pombe and ssa2-pombe are indeed heat-inducible since they lack a third and fourth protein. The retention of the heat-inducible nature would be important since S. pombe does not have a third and fourth gene to compensate for such a loss. This would also suggest that while the S. pombe genes are more closely related to their S. cerevisiae orthologs, functionally they may be more closely related to SSA3-yeast and SSA4-yeast. Another thing to note is that the HSE sequence motif was positioned within 300 bp upstream of the coding sequence in genes that had HSEs. This suggests that the position of HSEs is evolutionary conserved between the two fungi, though the actual HSE sequence motif might have diverged. Overall, we found the presence or absence of HSEs to be highly conserved through time which we believe supports our hypothesis that these proteins are exhibiting similar functions.

When we analyzed the position of signal peptides, we found that all of the cytosolic and mitochondrial HSP70s lacked a signal peptide sequence. The ER proteins Bip-pombe, Lhs1-yeast, and Lhs1-pombe were predicted to have a signal peptide sequence in the N-terminus region followed by a cleavage site. This motif is consistent with other ER proteins as well as the experimental data (Baxter et. al, 1996). It is interesting to note that SignalP did not predict a signal peptide sequence for the Kar2-yeast protein even though it was present in the S. pombe ortholog. The presence of positive signal peptide scores within the first 40 amino acids of Kar2-yeast would suggest a possible sequence (Figure 6). Although the software could not confidently predict a signal peptide, Kar2-yeast has been experimentally verified to contain a signal peptide that is cleaved off after transport into the ER (Normington et al, 1989). This discrepancy could be explained by the presence of an extra 15 amino acids within the signal peptide region of Kar2-yeast that is not found in bip-pombe. The addition of these amino acids could be the reason why SignalP software did not accurately predict Kar2-yeast to have a signal peptide.

The TMHMM software detected transmembrane regions within the N-terminus of the ER proteins. Since these transmembrane helices are located within the signal peptide, it is suggested that these regions could help in the transport of the protein to the ER. None of the cytosolic or mitochondrial HSP70s were predicted to have transmembrane helices. Though prediction of protein characteristics using web software was accurate for almost all of the proteins, these results underscore the importance of experimentally verifying these predictions as well.

We also analyzed the 3-dimensional structure of these HSP70s, and found that for the most part, these proteins were structurally identical, although we did not conduct an electrostatic surface analysis which might detect more subtle changes in the structure of these proteins. We did notice that the lhs1 gene in S. cerevisiae and S. pombe, which exhibited the highest sequence divergence between orthologs, did also diverge structurally both from the remainder the HSP70 family proteins analyzed here as well as from each other. This is not surprising given that analysis of conserved protein domains showed identical patterns across all proteins analyzed save for the LHS1 protein group. However, when we focused on the nucleotide binding site of the lhs1 gene, we did not see any obvious structural change there, although we did note potentially important changes in amino acids (i.e. some amino acids shifted from negative to positive charge, hydrophobic to non-polar, etc). Most of the structural changes were observed in the loops, rather than helices or sheets of the proteins, but there is a great need for electrostatic potential analysis of these proteins to more accurately predict whether or not they might be functionally different.

Our studies have shown that the HSP70s in both S. cerevisiae and S. pombe are divergent at a genomic level, but highly conserved at the functional/structural protein level. Some of the discrepancies found in our analysis (e.g. HSE elements and signal peptides) underscore the importance of experimentally verifying each protein. Many of the S. pombe sequences used in this study were inferred from the S. pombe genome sequencing project. It is possible that different regulatory characteristics or novel protein functions may be discovered as future experimentation is done on S. pombe HSP70s. In conclusion, we believe that while these proteins are encountering extremely high rates of mutation and shuffling throughout their respective genomes, they are also experiencing equally strong purifying selection which acts on those mutations to maintain conserved structure and function in these sequences.


Literature Cited/Figures available upon request.

Oceans Around the World – The Sunda and Sahul Shelves

Notes on several key papers regarding biodiversity hotspots in and around the Sunda Shelf.

Article 1: Crandall 2008

Vicariance patterns as a result of Pleistocene sea-level changes in the Sunda Shelf area should be present in both invertebrates and their ectosymbionts. Highly variable results across many different studies have spurred the authors to explore a more closely linked hypothesis: That patterns of genetic variation found in marine invertebrates (in this case, two seastars) should closely match that of their ectosymbionts (a mollusk and crustacean). Most of the four species did show at least some genetic structure, but it was not concordant across species, with each species displaying a different pattern of range expansion most likely due to differences in dispersal, and adult survivability.

Source: https://www.ncbi.nlm.nih.gov/pubmed/19067797

Map of Sunda and Sahul.

Map of Sunda and Sahul. CC 3.0 By Maximilian Dörrbecker (Chumwa). Wikimedia Commons.

Article 2: Crandall 2012

Sea-level changes during the end of the Last Glacial Maximum (LGM) should correspond closely with population range expansions of marine species. Prior to sea-level rises, the Sunda shelf and neighboring shelves were well above the ocean. Beginning roughly 20,000 years ago sea-levels began to rise rapidly, covering the Sunda shelf under water and facilitating the rapid expansion of marine species into this new habitat. The authors suggest that since the genetic signal of this sea-level rise is present in so many species, this event can be used as a means of calibrating the heterogeneous rate of mutation rates of lineages through time, that is, that younger lineages tend to have higher mutation rates. The authors proclaim strong support for the idea of time dependency of molecular clocks. This is an important understanding because correlating the time of geologic events with species/population events is a critical aspect of marine phylogeography.


Article 3: Kraus 2012

Here the authors investigate a genus of freshwater crab, Parathelphusa, for its historical biogeographic distributions in the Sunda region and the relation of those distributions to Pleistocene sea-level changes. The authors suggest that if Pleistocene-aged sea-level changes are responsible for the diversification of Parathelphusa clades throughout the Sunda region, then the rate of speciation should have greatly increased during that time. However, the authors find that most clades have Miocene or Pliocene origins, all with origins from Borneo although some speciation events did occur during the end of the Pleistocene, although rarely and via sporadic dispersal events as there have been no recent land-bridge connections.


Undergraduate Thesis: Effects of Artificial Moonlight on the Foraging Behavior of Mojave Desert Rodents

Effects of Artificial Moonlight on the Foraging Behavior of Mojave Desert Rodents

An Undergraduate Thesis Project
By Bryan White


Desert rodent communities are extremely diverse, which has prompted researchers to ask how so many species can coexist on similar, limited resources. Differences in foraging preferences associated with predator avoidance may contribute to coexistence. I determined how the foraging behavior of Mojave Desert rodents, especially pocket mice (Chaetodipus), were influenced by the increase in perceived predation risk associated with moonlight, which I simulated using artificial illumination. Millet (6.00 g) was mixed into trays filled with 2 L of pre-sifted sand. Seed trays were placed at stations located at different distances (2-82 m) from Coleman camping lanterns, and in either open or shrub microhabitats, so that rodents could choose to forage from resource patches with different levels of perceived risk. I also live-trapped rodents to identify likely foragers near the lanterns, and to determine the diversity and abundance of rodents in the area. Background illumination levels were recorded with using a Lux meter. I predicted that the amount of seeds removed would be highest at seed trays farthest from lanterns and under shrubs, and lowest at stations closest to lanterns and in open microhabitats. Surprisingly, I found no effect of distance, microhabitat, or illumination level on the amount of seeds removed by rodents.

Merriam's Kangaroo Rat

Merriam’s Kangaroo Rat. Photo CC 4.0 By Bcexp. Wikimedia Commons.


Desert rodents are intriguing animals because of their ability to survive in extremely harsh climates where resources are limited. Desert rodent communities are extremely diverse, which has prompted researchers to ask how so many species can coexist on the same resources, seeds (Brown 1988). In the Mojave Desert, for example, six different genera (Ammospermophilus, Chaetodipus, Dipodomys, Neotoma, Onychomys, Perognathus, Peromyscus), representing three rodent families (Heteromyidae, Muridae, Sciuridae) can all be captured in roughly the same area (Stevens et al. 2009). Most explanations suggest that animals reduce competition via resource partitioning, but differences in predator avoidance abilities may also contribute to coexistence (Kotler 1984).

It is widely accepted that desert rodents differ in their microhabitat preferences, and that these preferences reflect differences in the ability of rodents to detect and avoid predators, including owls, mammalian carnivores, and snakes. For example, quadrupedal rodents such as pocket mice (Chaetodipus, Perognathus), tend to forage in the cover of large shrubs, whereas bipedal rodents such as kangaroo rats (Dipodomys) are often found in open microhabitats between shrubs (Kotler 1984). Kangaroo rats are adapted to forage in open microhabitats in which there is little cover from visual predators (Thompson 1982; Kotler 1984). These adaptations include hopping locomotion, the ability to hear very low frequency sounds (1-3 kHz), and dorsally located eyes that should aid in spotting predators (Thompson 1982; Kotler 1984). Predation rates on rodents by owls are higher both in open microhabitats and (in a separate experiment) during periods of full moon, when levels of illumination might make movements more conspicuous (Kotler 1988). In contrast, pocket mice lack these morphological specializations, but presumably can move more efficiently beneath the denser shrub canopy (Thompson 1982). Interestingly, owls and rattlesnakes, the two most important rodent predators in the Mojave Desert, may have different effects on microhabitat use by rodents. Owls directly affect the perception of risk by desert rodents (Kotler 1988), which is higher in open microhabitats (Brown et al. 1988). The presence of rattlesnakes, which tend to hide near shrubs to wait for prey and to avoid being eaten themselves, decreased foraging of kangaroo rats in shrub microhabitats, although only during summer, when snakes are active (Bouskila 1995).

Mojave desert rattlesnake

Mojave Green Rattlesnake. Predator of Kangaroo rats. Public Domain By Lvthn13. Wikimedia Commons.

The response of rodents to predation risk has traditionally been measured in 2 ways: analysis of microhabitat characteristics at locations where rodents are captured, and foraging experiments to estimate differences in seed removal rates associated with different microhabitats. Optimal foraging theory states that animals will forage in an area until the costs of continued foraging, including perceived predation risk, outweigh the benefits (Morris 1997). The giving-up density (GUD), the density of seeds remaining in an artificial seed patch after a foraging bout, reflects this quitting harvest rate, and thus provides an index of an animal’s perceived risk and foraging costs associated with particular microhabitats (Brown et al. 1988). Both methods have been used to study how moonlight intensity affects rodent activity. Kotler (1984) found that increased illumination, simulated by camping lanterns, decreased captures in open microhabitats for some species and shifted habitat use for others in the Great Basin. Others have reported that bright moonlight reduces overall rodent activity aboveground (Brown et al. 1988).

I modified Kotler’s (1984) approach to investigate the possible effects of artificial illumination on foraging behavior of rodents in shrub and open microhabitats in the Mojave Desert. Rather than studying shifts in captures in different microhabitats, I measured seed removal rates in artificial seed trays placed at different distances from an artificial light source, a gas-powered Coleman lantern. I also quantified variation in light levels at varying distances from the lantern and in the open and beneath shrubs to understand better how actual light levels differ between these microhabitats. I predicted that seed removal rates would be lower (and GUDs higher) in trays close to the lanterns, where illumination was greatest, than at trays where there was only natural light. I also expected that rodents would remove relatively more seeds from trays beneath shrubs than in open microhabitats, especially near the lanterns, where the difference in illumination would be greatest.


My study site was conducted approximately 5 km NW of the Desert Studies Center, Zzyzx, California, during the months of June and July 2010. The site was a broadly sloping bajada at the base of an alluvial fan. Vegetation consisted mostly of creosote bush (Larrea tridentata), burrobush (Ambrosia dumosa) and desert holly (Atriplex hymenelytra), with scattered forbs and grasses. The substrate was a mixture of medium-size to small rocks and gravel, with some sandy washes. Shrub microhabitats were considered to be any shrub that appeared large enough to provide adequate canopy cover over a seed tray. Open microhabitats were locations that were at least 1 m from shrubs.

Creosote bush, Mojave Desert

Creosote bush, Mojave Desert. Public Domain By Klokeid. Wikimedia Commons.

To determine which rodent species were present at my site and foraging in seed trays, I set large Sherman live traps on the night prior to foraging trials. Traps were baited with commercial bird seed that had been microwaved for 5 min to prevent germination. During June, 30 traps were placed at same locations as the seed trays. During July, trapping was done in a 7 x 7 grid (49 traps separated by 10 m) in the area where seed trays were placed.

Artificial seed trays were houseplant saucers (6 cm deep and 32.5 cm in diameter) buried so that edges were flush with the ground. When set, each tray contained 6.00 g of millet mixed in 2 L of pre-sifted, fine sand. In June, I placed trays 2, 12 and 22 m points along 2 parallel lines extending out from a central Coleman (Dual Fuel) camping lantern. This was repeated 3 times at 50-m intervals, for a total of 30 trays. In July, I placed 32 trays from (2 m to 72 m at 10 m increments) in four different directions extending out from two centrally-placed lanterns. Two lanterns were used in the second design in an attempt to increase illumination levels. At a given location, seed trays were randomly assigned to be either in a shrub or open microhabitat. One additional tray was covered with hardware cloth to prevent foraging and was used as a control to quantify changes in weight of seeds due to moisture overnight (Stapp and Lindquist 2007).

Trays were set out at dusk. I allowed rodents to forage in seed trays for approximately 4 h. The remaining seeds and sand were collected from trays and taken back to the Desert Studies Center lab, where the sand was sifted to remove seeds. The seed was cleared of debris and weighed using a precision scale to estimate the amount of seeds removed. Seed trays were considered to have been foraged if the amount of seed removed differed by 2% of control trays from that night.

At the beginning of each foraging trial, I measured light intensity using an Extech 401036 light meter. Illumination was measured by placing the light meter on the seed tray so that the light receptor faced straight up. This measured the ambient light in the area, as opposed to the relative intensity that a rodent may perceive being emitted from a light source (either the moon and stars or lantern). I assumed that rodents look to their immediate surroundings, rather than some distant light source, to decide the relative risk of the potential foraging patches. Trials were conducted under similar background moon conditions (waxing gibbous).

All statistical analyses were conducted in Microsoft Excel (2007) Data Analysis Toolpack and Minitab 15 (Minitab Inc. 2007).


Based on a total trapping effort of 64-trap-nights over the 2 trapping sessions, pocket mice Chaetodipus (17 individuals of 2 species, C. penicellatus, C. formosus, that were not distinguished) were the most abundant rodents, followed by Merriam’s kangaroo rat (D. merriami, 6 individuals) and the desert woodrat (Neotoma lepida, 1 individual). I therefore assumed that most trays were visited by pocket mice.

A total of 62 seed trays were set out during the 2 trials. During the June trials, 3 of 30 trays (1 shrub, 2 open, 1 spilled) were considered to have not been foraged, whereas in July, more than 2/3 (22/32) of the trays were not foraged (10 shrub, 12 open, 1 spilled). Only results from seed trays that were considered foraged were included in the analysis. Combining across all distances and trials, there was no significant difference in the amount of seeds removed in shrub and open microhabitats during June or July (Fig.1; P > 0.05). Combining both trials, the amount of seed removed was not related to distance from the lanterns in either open (Fig. 2; R2 = 0.007, P = 0.745, DF = 15) or shrub (Fig. 2; R2 = 0.15, P = 0.0936, DF = 18) microhabitats.

Illumination levels were highest in seeds trays near the lanterns, but declined considerably by 12 m from the lanterns (Fig. 3). Shrub and open trays were exposed to similar light levels.

Surprisingly, illumination did not influence seed removal rate in the way that I predicted. At all levels of illumination and in both shrub and open microhabitats, rodents consumed most of the seeds in the trays (Fig. 4). In fact, the amount of seeds removed was lowest and most variable at the lowest light intensities.


I found no evidence to support my hypothesis that rodents would spend more time foraging in darker, shrub microhabitats, where risk of predation would presumably be lower. This was particularly surprising because pocket mice were the most common rodents I captured at my study sites and probably were responsible for most of the foraging in seed trays. Pocket mice are quadrupedal and generally prefer the cover of shrubs (Kotler 1988), and therefore would be expected to be sensitive to predation risk. My results differ from those of Kotler (1984), who found, based on live-trapping, that rodents, including quadrupedal species, increased their use of shrub microhabitats in the presence of artificial illumination. However, Kotler (1984) also found that seed enrichments increased the use of the open microhabitat by kangaroo rats. This suggests that, in my study, while pocket mice may have focused their foraging efforts on removing seeds from under shrubs, kangaroo rats may have been opportunistically foraging in brighter areas, and due to its larger body size and bipedal locomotion, consistently removed large amounts of seeds from those trays.

The fact that rodents ate nearly all the seeds in seed trays at all distances from the lanterns and irrespective of illumination levels suggests that the bright light associated with the lanterns did not deter them from foraging. In fact, rodents removed the smallest amounts of seed in trays at the darkest light levels, including some beneath shrubs (Fig. 4). This suggests that factors other than illumination and microhabitat influenced foraging behavior at these low light levels. It also suggests that rodents can find a large amount of dispersed seed (6 g) in a relatively short time. It is possible that multiple rodents visited a given tray, but I was not able to determine the number of rodents using each tray.

If I were to repeat this study, I would increase the number of replicate seed trays and use the same experimental design throughout. I also would keep a record of whether there are tracks in trays as an index of foraging. Another way to improve my experimental design would be to video record seed trays to know how many animals and of which species visited a seed tray during a foraging bout.

Literature Cited

Bouskila, A. (1995). Interactions between predation risk and competition: A field study of kangaroo rats and snakes. Ecology, 76(1), 165-178.

Brown, J. S., Kotler, B. P., Smith, R. J., & Wirthz, W. O.,II. (1988). The effects of owl predation on the foraging behavior of heteromyid rodents. Oecologia, 76(3), 408-415.

Kotler, B. P. (1984). Risk of predation and the structure of desert rodent communities. Ecology, 65(3), 689-701.

Kotler, B. P. (1988). Environmental heterogeneity and the coexistence of desert rodents. Annual Review of Ecology and Systematics, 19, 281-307.

Morris, D. W. (1997). Optimally foraging deer mice in prairie mosaics: A test of habitat theory and absence of landscape effects. Oikos, 80(1), 31-42.

Stevens, R. D., & Tello, J. S. (2009). Micro- and macrohabitat associations in mojave desert rodent communities. Journal of Mammalogy, 90(2), 388-403.

Stapp, P. and Lindquist, M (2007). Roadside foraging by kangaroo rats in a grazed short-grass prairie landscape. Western North American Naturalist, 67(3), 368-377.

Thompson, S. D. (1982). Microhabitat utilization and foraging behavior of bipedal and quadrupedal heteromyid rodents. Ecology, 63(5), 1303-1312.