Things you learn from fish that can’t

Biology research can be a shot in the dark in the early stages.

While this may sound like an “unscientific” approach to research, research begun without a hypothesis can lead in a promising direction and reveal new information. One example is a genetic screen. There are two types of genetic screens, forward and reverse. A reverse genetic screen involves observing the effect of mutating a known gene, moving in reverse from gene to phenotype. A forward genetic screen is a powerful tool for determining what genes are responsible for a known phenotype. Both of these types of genetic screen allow you to elucidate relationships between phenotypes and genotypes that you could never predict. Thus research such as these genetic screens, which are not hypothesis-based, saves the time and effort of making many hypotheses in the wrong direction and also helps to reduce bias. This type of research is extremely useful for the groundwork on complicated biological processes, such as genetic disease.

My research this summer started with a forward genetic screen in the Granato Lab long before I arrived. The Granato Lab at the University of Pennsylvania works with zebrafish, focusing on behavior and development. Although you might not consider fish to be complicated or intelligent animals, zebrafish and other fish can participate in a simple form of learning called habituation. This form of learning is considered “simple” and is seen in most animals, however, the process of habituation is not fully understood which makes it a great candidate for a genetic screen. The Granato Lab sought to understand the basis of habituation in zebrafish by performing a forward genetic screen and investigating promising mutations affecting habituation. The forward genetic screen provided many candidate genes that may be involved in the process of habituation in zebrafish.


Abby Frerotte (another Haverford student) injecting zebrafish embryos for CRISPR, the method used to generate the mutant zebrafish.

In this type of research, some results are more promising than others. One particularly promising gene was the focus of my summer: hip14. There are several reasons this gene was promising for further research. Hip14 is a gene that encodes a palmitoyl transferase that physically interacts with Huntingtin. The relationship between hip14 and Huntingtin points to a connection with learning and with other animals. The protein Huntingtin is produced in many animals, including humans. In humans, Huntingtin protein is implicated in Huntington’s disease, a debilitating genetic disease. The connection between the gene identified in zebrafish and this human illness goes even deeper: zebrafish with mutations of hip14 are unable to habituate to auditory stimuli, and Huntington’s disease patients also have auditory processing deficits.

IMG_9860 IMG_9864

The hip14 mutant zebrafish. (Fish are surprisingly hard to photograph)

This genetic work was done before the summer, and a line of hip14 heterozygous mutant zebrafish was raised in the lab. Fast-forward to now: my boss, Jessica Nelson, was a visiting professor for SuperLab and reached out to me and another student to run a drug screen for hip14 in zebrafish. Our aim for the summer was to find drugs that restored homozygous mutants’ ability to habituate. Our drug screen was partially hypothesis based. For the beginning part of the internship, Abby and I researched work already done with habituation in zebrafish to identify candidate genes known to be somehow connected to hip14 and therefore affected by a mutation of that gene. We specifically focused on palmitoylation substrates of hip14. Then, we delved further into research to find drugs that were known to affect those candidate genes. Thus we made a short list of drugs that would possibly restore habituation to the mutant zebrafish.

The rest of my summer consisted of testing each of those drugs on homozygous mutant zebrafish larvae. This first involved setting up matings between the heterozygous mutant zebrafish to give us a group of offspring, which were both mutant and non-mutant (siblings). One subset of these mixed offspring would receive the drug, and the other subset would not. Therefore we had four categories: mutants treated with drug, siblings treated with drug, mutant control group, and sibling control group. Then we would test their ability to habituate. Have you ever been to an aquarium and noticed that when you tap on the glass the fish do nothing? The normal response to an auditory stimulus is for a fish to quickly swim away. Zebrafish do this in a very characterized manner, forming a perfect curve as they reflexively contract the muscles on one side of their bodies to swim away. These are called short-latency c-bends. Eventually, after many close-together stimuli, the fish will stop responding this way; they are now habituated. This is measured in the lab by an amplifier producing a tapping sound and a high-speed camera recording the movement of the fish after the sound is made. We can then use tracking software to track many, many fish and test whether they change their behavior in response to the sound over time.

A 16-well plate where the fish are held in water while they receive tap stimuli and are recorded by a high-speed camera from above.

A 16-well plate where the fish are held in water while they receive tap stimuli and are recorded by a high-speed camera from above.

Over the summer we varied the drugs that we used, and the concentrations and treatment times for the drugs, searching for a combination that helped the hip14 mutant zebrafish habituate. This involved data analysis where we graphed their response to the sound over time, watching as to whether the mutants ever stopped responding. We then matched our tracking results with the genotypes of the fish, mutant or sibling, and treatment group, drug or no drug. In this way during testing, tracking, and data analysis, we had only a rough idea of who were mutants and who were siblings. It was only after data analysis that the results were revealed by these categories. I am writing this blog post at the end of testing and analysis for our last candidate drug of the summer. Unfortunately, none of the drugs we tested this summer showed any significant restorative effect for the mutant zebrafish.

Screen Shot 2016-07-29 at 11.41.06 AM

An example of a null hypothesis: Indirubin (drug) failed to significantly increase mutants’ habituation (in red) as compared to controls (in pink).

Results that go against hypotheses or that show a null hypothesis are often pushed under the rug in discussions about research. Many researchers even shy away from publishing negative results or inconclusive results. This practice leads to the loss of scientific knowledge. In my quest to reflect on my research this summer I have been exploring the merits of failed hypotheses and non-hypothesis-driven research. The beginning of this research was not hypothesis-based and led to a very promising connection between a mutant deficient in habituation and Huntingtin. However, the protein encoded by hip14, which physically interacts with Huntingtin, is involved in a complicated network of other genes and proteins. The drugs that we tested this summer came out of a hypothesis: drugs affecting genes that are palmitoylation substrates of the protein encoded by hip14, the former known to increase habituation, would rescue hip14 mutant’s ability to habituate. This sounds complicated, but the short explanation is that these drugs were meant to be positive controls, they were meant to provide a baseline for more “risky” drugs. We assumed because they were connected with hip14 and were shown to increase habituation that some of them would have a positive effect on habituation during our screen. Our negative results do not necessarily point to a failure in our hypothesis as much as the difficulty of predicting these complicated, interwoven pathways of genes and proteins.

However, these results start to elucidate what is not involved in this habituation pathway in zebrafish. Why did our results go so against previous findings in habituation again and again? Is there a missing link or a completely different pathway that is also involved? These are interesting questions that are raised by research that goes against a hypothesis. More than anything this summer has taught me that no matter how “sure” a hypothesis seems, the results will surprise you.

Gas Clouds in Space: the Smith Cloud and it’s Magnetic Field

Space.  It is “to place as eternity is to time” (Joseph Joubert).  It has been called the “Final Frontier” (Star Trek), a “miracle” (Walt Whitman), and according to Lady Violet Bonham Carter, a British politician during the 1920-1940s, ”space is no place for a person of breeding.”

No matter what we call the area surrounding Earth, it is something that fascinates, moves, and puzzles astronomers and non-scientists alike.  It is vastly empty, with galaxies and clouds of dust and gas filling in the gaps.  There is so much we do not know from the Big Bang to our own Sun.  However, everyday thousands of astronomers find an aspect to focus on and we are slowly piecing together the physics and mysteries of space around us, and Earth’s place among it all.

This summer, I have focused on one small piece of the larger puzzle that is in the universe.  To help you understand what I am doing, I will first describe a galaxy in more detail that usual.  Most people assume that a galaxy is just a disk of stars with planets orbiting around them; however, galaxies are so much more!  There is dust and gas, dark matter, nebulas (giant clouds that produce and act as nurseries for new stars), and supernovae (exploding stars) between every solar system.  The dust and gas help form new stars as gravity slowly pushes the particles together.  Outside the disk of stars and planets, floating around the galaxy (think of a sphere with the galaxy cutting the sphere in half) are giant clouds of dust and gas and small galaxies, called dwarf galaxies. (See Fig 1)

They are part of the galaxy, either having been ejected from the host galaxy or pulled in by the galaxies gravitational pull.  It is from these gas clouds that we can learn about how stars form and how galaxies eject gas.


Fig 1 – Milky Way Structure

I am focusing on one gas cloud in particular for my summer research; it is called the Smith Cloud, a large cloud mainly composed of hydrogen, discovered by Gail Smith in 1963. Some basic facts:

  • Mass: 1 million solar masses1 (1 million times the mass of the Sun)
  • Actual Size: 11,000 light-years long and 2,500 light-years across.1
  • Size From Earth: About 30x greater than the full moon
  • Distance from Earth: Between 36,000 light-years and 45,000 light-years1
  • Speed: 600,000 mph toward the Milky Way2

Fig 2 – Path that the Smith Cloud is taking. Credits: NASA/ESA/A

One of the main questions being asked about the Smith Cloud is where did it come from?  Andrew Fox, an astronomer at Space Telescope Science, answered the question earlier this year when he discovered that the cloud most likely came from the Milky Way and traveled through part of the disk of the galaxy before being ejected from the disk (See Fig 2).

Astronomers have been able to measure the amount of hydrogen that the cloud has and how much should have been lost as the cloud traveled through Milky Way due to forces exerted from both the galaxy and cloud.  From computer simulations, the cloud should have lost more hydrogen to the galactic disk than observed.

How the cloud survived almost intact falling through the galaxy is unknown, though the cloud’s magnetic field is one suggestion for the cloud’s survival. In 2013, my advisor Alex Hill and his collaboration decided to test the hypothesis that a magnetic field helped keep the cloud together.  A magnetic field had not been detected                                                       around the cloud and they found one!

The main question is how to detect a magnetic field.  The short answer is that you look at the Faraday rotation of a hydrogen point source located far beyond the Smith Cloud. The longer and easier to understand answer is that Alex used a survey of a sky taken from the Very Large Array, a radio telescope in NM, which looked at very distant bright galaxies, so far away that they look like points.  The radio waves from selected bright radio sources travel through the Smith Cloud to reach us. As the waves pass through the cloud, the Smith Cloud’s magnetic field changes the polarization, or direction that the waves are traveling.  The amount the wave’s direction changes depends on the strength of the magnetic field. By measuring the change in the sources’ radio waves, we measure the strength and direction of the magnetic field. This measurement is called a Faraday rotation measure (RM).


Fig 3 – RM map of the Smith Cloud.                                             Credit: Hill, et. al, 2013

Alex used the RMs found from this survey and mapped them onto an image of the Smith Cloud (see Fig 3). The RMs are either positive or negative, which tells us the direction of the magnetic field.

  • Red Circle: positive magnetic field, towards us
  • Blue Circle: negative magnetic field, away from us

From this map, Alex discovered a magnetic field of ~8 uG and proposed that the field is cylindrical around the cloud.3 However, he needed more data to confirm and bulk up his conclusions.  Alex used the VLA and looked at ~3000 point sources around the Smith Cloud.

And this is where I came in.

I was given this dataset of 2850 point sources to analyze.  Using various computer scripts that I have written or borrowed, I am going through each source and finding its Rotation Measure.   In order to reduce the data, I used a program called CASA, or Common Astronomy Software Applications.  It is a very frustrating program, as anything can cause it to crash or not work!  Luckily, I used it during the past school year and last summer, and was able to avoid a lot of issues.


Fig 4 – Image of a point source

The data reduction process starts with taking out noise and RFI.  RFI is radio frequency interference, which is emitted from anything electronic, from your cellphone to your microwave.  This emission interferes with the radio waves from space and has to be removed before doing any science with the data. After the data was reduced and all that was left was a data cube of each source (See Fig 4),which is essentially many images of the source at different frequencies stacked together, I ran the image through a program called RMCLEAN which finds the RM.  I ran this program with IDL, or Interactive Data Language which is a programming language, which increased my knowledge of IDL.

RMCLEAN can take anywhere from 30 seconds to several minutes to run for each source.  Also, I have to look at an image of each source to make sure that the source is visible.  A little over half of the images do not actually show a point source. This has posed a serious problem during the summer as I had to go through every source and either discard it or keep it.  Luckily, I was able to write a script to do this and was able to go through all the sources in about an hour. Finding the RM took up the most time this summer as I had to go through most of the sources by hand.


Fig 5 – Intensity vs Rotation Measure.

Part of the script I was using created a plot of Intensity vs RM for each source (See Fig 5) which allowed me to find the RM.  This is one of the best examples, as it has a very high signal (peak) to noise (everything surrounding it).  However, not all of the RMs looked this nice and it was incredibly difficult sometimes to discern where the RM peak was amidst the noise.   The maximum of the main peak of the plot is the rotation measure. The value is either negative or positive. After finding all the RMs of the data – I found around ~930 sources after discarding the ones without point sources – I mapped them onto an image of the Smith Cloud in order to see how the magnetic field changes/is amplified. (See Fig 6)


Fig 6 – RM Map with old data along with new data. New RMs are red/blue, old RMs are pink/cyan


Fig 7 – The Green squares and circles are the clumps of RMs that I have grouped together.

I am working on finding the magnetic field of clumps of RMs. Hopefully, by the end of the summer I will have magnetic fields for these parts of the magnetic field (See Fig 7).  I am doing this with a script I have written that allows me to group together RMs near each other, draw a box/circle around them, average them together, and find their magnetic field.

Now, I will be refining the groups and calculating more accurate magnetic fields, as the ones I have found can be better! As I continue this project into my senior year as my thesis, I will learn more about the Smith Cloud and its magnetic field.  This will help confirm the existence of a magnetic field and give a better understanding to how the Smith Cloud remained intact.


  1. Lockman, et. al, 2008.
  2. Fox, et. al, 2016.
  3. Hill, et. al, 2013.

The Effects of Varying the Sequence of a Semenogelin Derived Peptide on Hydrogelation

During the summer, I worked as a research assistant in Prof. Åkerfeldt’s chemistry lab in Haverford College. We studied a peptide sequence, referred to as the Sg-peptide, which represents a portion of semenogelin (Sg), a protein found in semen.  From earlier experiments, we have shown that it forms a hydrogel. A hydrogel is a substance consisting of a network of polymer chains that acts as a scaffold and traps large quantities of water molecules in between its chains. Examples of hydrogels that are used in everyday applications include diapers and contact lenses.

We are interested in studying the properties of the Sg-peptide to learn more about how it forms a hydrogel, which includes testing the Sg-peptide with various conditions such as temperature, pH and salts concentration.

Because of its natural origin, we believe that Sg-peptides will have many potential applications in the medical field, such as being a drug delivery vehicle in which drugs are encapsulated in the gel, injected, and then slowly released within the body as the hydrogel dissolves.  In particular, we are interested in the sequences SgII 0, SgI 0 (Sg I and SgII refer to two different forms of the Semenogelin protein) and a series of truncated sequences within SgI 0 (Figure 1). The longest sequences contain 13 amino acid residues.

Figure 1.  The amino acid sequences of SgI and SgII-peptides 0 expressed in one letter codes and truncated versions of SgI 0.

Figure 1. The amino acid sequences of SgI and SgII-peptides 0 expressed in one letter codes and truncated versions of SgI 0.

The amino acid residues in the sequence are arranged such that they alternate between being hydrophobic and hydrophilic. This allows them to form a structure called a beta-sheet, which can come together to form long fibrils, which then cross-link to trap water molecules. We also speculate that a particular amino residue histidine may also be a major contributor to hydrogelation (formation of hydrogel).

Working alongside two other Haverford students on this project, my role in the research team has been to synthesize some of the truncated sequences.  All peptides were made by solid phase methods.  The tripeptide, sequence 10, was made manually in which I carried out each step of the synthesis in a plastic tube, while other peptides (0, 1, 7, 8, 9, 11) were made using a machine (referred to as an automated peptide synthesizer). I learned to cleave these peptides from the solid support and purify them using a method called high-performance liquid chromatography (HPLC), which involves using an instrument that pumps two different solvents through a column in which the peptide gets separated from side-products (Figure 2).  On the analytical side, I have learned to use a mass spectrometer, which determines the mass of the peptide, both in crude and purified form (Figure 3).  I have also conducted gelation experiments of both crude and purified peptides at different concentrations and pH.

Figure 2. HPLC trace of SgI 0 with a retention time of 21.5 min using a gradient of X-Y water and 90% acetonitrile in water with the buffer 0.1% trifluoroacetic acid.  The shoulder before the main peak was removed whereas the shoulder right after the main peak was shown to also be SgI 0, eluting later due to aggregation.

Figure 2. HPLC trace of SgI 0 with a retention time of 21.5 min using a gradient of X-Y water and 90% acetonitrile in water with the buffer 0.1% trifluoroacetic acid. The shoulder before the main peak was removed whereas the shoulder right after the main peak was shown to also be SgI 0, eluting later due to aggregation.

Figure 3. Mass Spectrometry Chromatograph of pure Sequence 9.  Expected mass: 574  Observed mass 574  (M+H+) with the dimer (2M+H+) also observed at 1147.

Figure 3. Mass Spectrometry Chromatograph of pure Sequence 9. Expected mass: 574 Observed mass 574 (M+H+) with the dimer (2M+H+) also observed at 1147.

Through trial and error, optimal procedures have been found for preparing each peptide sequence by solid phase methods and for purifying them using HPLC. Sequence 10 (a three-residue peptide) required a C8 column and the strong denaturant guanidine thiocyanate; Sequence 9 and sequence 11 used a C8 column; and SgI 0 KD peptide required a C18 column and the addition of the base NaOH in order to change the sample pH from 3 to 9-10, which more readily dissolved the peptide. Using the 433A peptide synthesizer, sequence 1, 7, 8, 9, 11 and multiple sequence 10 batches were found to be successfully synthesized.

The amount of progress made in just 10 weeks is outstanding and I’ve acquire many lab skills that I’m confident will greatly aid me in my organic chemistry lab course. I would like to thank my instructor Karin, for her amazing support and guidance and my colleagues Brett and Abby for their help. I would like to thank Bash who was always available to offer help and suggestions. And lastly, thank you to the KINSC for granting me this great scholarship and opportunity to conduct research in an organic synthesis lab.