Germany for a Summer: My time with the ants

My summer research began with me driving down to the tiny town of Rensselaerville, NY in search of ants.  There, at the idyllic Huyck Nature Preserve, I met up with my advisor and some of my labmates, who had arrived a week earlier from the Johannes Gutenberg Universitat in Mainz,Germany. Each day we would hike out into the woods of the preserve and search for black fungus-covered felled branches,  which our species of interest, Temnothorax longispinosus, loves to inhabit. After collecting armloads of the sticks, we would sit down and crack open the sticks with pocket knives. Upon finding a T. longispinosus colony crawling around inside a stick, we’d tap the colony out into plastic bags for storage.

Searching for T. longispinosus out in the field.

Searching for T. longispinosus out in the field.

Just a fraction of the 1000 colonies we collected at the Huyck.

Just a fraction of the 1000 colonies we collected at the Huyck.

After collecting close to 1000 colonies, we stuffed the bags into several suitcases (amazingly, the colonies all survived) and flew across the Atlantic. Upon arriving at the university with our tens of thousands of guests, we began a several week process of unpacking, cataloguing, and feeding the colonies in preparation for our experiments–the downside of using organisms collected in the field! I began my project by labelling my ants with colored wire based on their caste. T. longispinosus are particularly small ants, so it was quite a challenge to slip loops of colored wire around their abdomens–the first few took about 45mins each. But after a few days, I was able to “wire” over 30 ants per hour.

Deep in focus while wiring ants.

Deep in focus while wiring ants.

Then at last came the actual experiment. I fed half of my 40 colonies dsRNA fragments designed to knock down expression of a gene hypothesized to be essential for maintenance of the ants’ circadian rhythms. No one had looked at this gene in ants before, only a Drosophila homologue. My advisor and I were specifically curious in what effect knocking down this gene would have on foraging behavior, since an earlier expression analysis had found the gene to be overexpressed in foragers relative to brood carers. For comparison’s sake, we instead fed nonsense RNA to the other half of the colonies.

 

After a week-long knockdown period, I began videorecording colony behavior. Colonies were placed in arenas and filmed for 24 hours, with alternating light-dark conditions in the filming area to simulate day and night. Ants can’t see red light, so we used red LEDs to illuminate the arenas without disrupting the ants’ perception of day and night.

 

I then took to analyzing foraging behavior in the videorecordings. With 480 hours of footage to review, this was quite the task. But I was able to finish just in the nick of time, obtaining the final results a day in a half before I was scheduled to present my findings to the lab.

 

What did I find? Somewhat unsurprisingly, I found that  knocking down this gene indeed appears to disrupt the circadian rhythm of foragers, causing them to abandon their normal diurnal foraging schedule and forage equally throughout the day and night. The gene does not, however, appear to more generally affect the decision to forage, since we found no difference between control and knockdown colonies in foraging levels over the entire 24hr period.

 

With confirmation that this gene is essential for the normal circadian rhythm (at least in foragers) in T. longispinosus, the door is now open for some deeper research questions. Is this gene’s circadian-related function limited to foraging, or does it affect activity in general, for all castes of ants? And does the ant circadian rhythm respond to perception of time of day, light conditions, or both?
Working on this project in the context of a different culture made the experience especially stimulating for me. It was difficult adapting to the new environment at times, but the challenge made for a good deal of personal growth. I’m returning to Haverford feeling rejuvenated (though jet-lagged!), excited to dig into my senior research. No more ants for me, probably, but I appreciated the opportunity to dip my toes into research with a behavioral focus. Thanks so much to the KINSC Summer Scholars program for funding me this summer.

The view of Mainz across the Rhine.

The view of Mainz across the Rhine.

Ten weeks with alpha-synuclein

I spent my summer working with alpha-synuclein, a somewhat poorly-understood protein found in the presynaptic terminals of human neurons. Its suggested functions include chaperoning SNARE complexes, regulating vesicle size in the presynaptic terminal and the release of neurotransmitters into the synapse.

The purpose of my research this summer was to use a thiocyanate vibrational probe group at several sites in relevant regions of alpha-synuclein to better characterize its structure. This builds on previous work by Alice Vienneau ’12 and Dan Konstantinovsky ’16, who had inserted the SCN probe group at 8 sites of interest. This process relies on site-directed mutagenesis to replace the original codon with a codon for cysteine, then expressing this monocysteine mutant and cyanylating to attach the CN moiety. The thiocyanate probe group has been the subject of previous and ongoing research in the Londergan lab and can be used in IR spectroscopy to report on whether the site is buried in a lipid bilayer, such as a vesicle membrane, or exposed to the aqueous solvent.

Recent research has discovered that codon 136 (TAC) of the alpha-synuclein plasmid used, which is typically translated into a tyrosine, can be mistranslated in Escherica coli to incorporate a cysteine; this happens roughly 20% of the time. This additional cysteine would likely be cyanylated along with the desired mutant cysteine at the probe-group site and could interfere with the IR spectrum results. By changing the codon to a degenerate codon for tyrosine (TAC to TAT), we can eliminate this problem, as E. coli will not express this as a cysteine.

The product of the site-directed mutagenesis reaction is transformed into bacteria and grown on a plate. Since it's unreliable, we tend to sequence three colonies at a time to check for the desired mutation.

The product of the site-directed mutagenesis reaction is transformed into bacteria and grown on a plate. Since it’s unreliable, we tend to sequence three colonies at a time to check for the desired mutation.

Consequently, my teammates and I spent part of our time in the lab using site-directed mutagenesis to alter the 8 original monocysteine mutants to incorporate the Y136tat change as well. Site-directed mutagenesis uses oligonucleotide primers of roughly 50 bp that contain the desired mutation—generally just one or two base pairs’ difference—but match the template in the surrounding base pairs. PCR creates the product DNA. This is a finicky process that requires a good bit of patience—some of our plasmids worked on the first try, but others required sequencing a dozen colonies before we saw the desired mutation. During the process we created three additional monocysteine mutant plasmids to study new sites. All the mutant alpha-synuclein plasmids were expressed in cultures of E. coli and induced with IPTG, since our plasmids use some of the same machinery as the lac operon.

Purification of the cell lysate took the form of a separation on the basis of charge (anion-exchange chromatography), followed by separation on the basis of size (size-exclusion chromatography). We encountered a major hiccup here: UV spectra of purified protein samples revealed very strong absorbance at 260 nm, which indicates the presence of nucleic acids. It seemed that DNA was eluting with alpha-synuclein on both columns. We experimented with salt cuts and DNases to try to eliminate the nucleotides, but the solution was to replace the DEAE anion-exchange column we were using with a similar column—a Q FF anion-exchange column. SDS-PAGE and UV spectra have suggested that this column separates the DNA from the protein.

Protein UV absorbance should peak at 280 nm (right vertical red marker), but this spectrum clearly shows the peak at 260 nm (left marker). This indicates the presence of nucleotides.

Protein UV absorbance should peak at 280 nm (right vertical red marker), but this spectrum clearly shows the peak at 260 nm (left marker). This indicates the presence of nucleotides.

With a reliable protocol finally in hand, the next step is to express and purify all the new, Y136tat mutants. These will then be cyanylated to create the SCN probe group at each of the 11 desired sites and studied with IR spectroscopy. We’re hoping to characterize the conformations alpha-synuclein takes when exposed to various sizes of lipid systems and to learn more about which regions and residues are involved in its aggregation process.

Designing a more efficient synthesis of a long-forgotten molecule of great potential

With the growing and essentially unregulated use of antibiotics for all sorts of ailments, that of antibiotic resistance is a serious and understated problem in today’s society. There are few antibiotics capable of dealing with multi-drug-resistant (or “super”-) bacteria, and it is now generally accepted that new, more potent antibiotics need to be discovered if we want to put an end to this problem. This is where my story begins.

My story is about a molecule discovered back in 1980 in Japan, extracted from a soil bacterium, S. lavendulae. It was discovered as an anticomplement agent (i.e., one acting on the complement or immune system to repress responses) and aptly named complestatin, although its structure was not yet known at the time. In 1989 its structure was finally elucidated and found to consist of a string of seven amino acids including several “unnatural” or non-canonical amino acids (see Figure 1)––hence, complestatin was a heptapeptide and belonged to a class of natural products known as the Non-Ribosomal Peptides (NRPs), produced independently of the ribosome by a mega-enzyme assembly known as a Non-Ribosomal Peptide Synthetase (NRPS). In the 90s it was then found that complestatin possessed a number of other interesting and useful properties including anti-HIV and neurotoxicity inhibitor properties. Recently, it has been found that it is also a very potent antibiotic, equaling if not surpassing vancomycin in potency––vancomycin being the last-resort antibiotic used in cases of infection by multi-drug-resistant bacteria. Clearly, complestatin is a very interesting molecule!

Published structure of complestatin/chloropeptin II

Figure 1: Published structure of complestatin/chloropeptin II

In an effort to presumably advance the molecule into clinical trials, several organic synthesis research groups attempted its total synthesis, but success was unthinkable to start with because of two cross-links between the amino acid side-chains (a biaryl and biaryl-ether bond) that are critical for the molecule’s three-dimensional shape and bioactivity, yet very challenging to install. As a result, the synthesis (which was successfully achieved) required more than 30 steps––presumably taking several years to achieve––and resulted in no greater than 5% yield. It should be obvious to anybody that this is not a good way to proceed, especially if the molecule is to be synthesized in large enough quantities for clinical trials and, later, channeling into the drug market.

In an effort to improve the synthesis of complestatin, in the Charkoudian lab we are currently working on a so-called “chemoenzymatic” synthetic route for the synthesis of complestatin. The “chemo” part refers to the chemistry behind the synthesis that takes advantage of a technique called Solid-Phase Peptide Synthesis (SPPS). In SPPS, micron-to-millimeter-sized beads visible to the naked eye are loaded with an amino acid. Following a washing step (during which the amino acid is retained on the solid matrix and all excess reagents are washed off) and a deprotection step to get rid of a protecting group (which prevents the added amino acid from reacting with itself), the beads are washed and loaded with another amino acid. The cycle of addition, washing, deprotection and washing is repeated until the desired number and types of amino acids are loaded, and the peptide is subsequently cleaved to yield the final product. The “enzymatic” part of the name instead refers to the use of two different enzymes to install the two critical bonds referred to earlier. The DNA encoding the enzymes is taken directly from the genome of the bacterium that produces complestatin (S. lavendulae) and heterologously transcribed and translated in E. coli. Following their expression and purification, the enzymes are added to the linear version of complestatin, itself bound to a PCP7-X didomain responsible for recruiting the enzymes to the peptide and allowing for cyclization to occur sequentially; this mimics what happens in the NRPS when the peptide is cyclized as it is bound to the 7th and last module of the enzyme assembly (hence the “7” in the PCP7-X nomenclature). All species, including the cyclized products, are analyzed by NMR and LC/MS to confirm their identity.

The project was started about 3 years ago by Niki von Krusenstiern and was taken up by Joshua Bulos and then myself. This summer was very productive. I was able to continue the synthesis of the peptides, investigating two different resins: the Wang Resin and the Dawson Resin. The Dawson Resin has the advantage that following activation of the linker, the peptide is cleaved from the resin directly as a thioester conjugate, ready for transesterification with CoA and then loading onto the PCP7-X didomain. Instead, the peptide is cleaved from the Wang Resin in its free carboxylic acid form, requiring an extra thioesterification (and purification) step. While we attempted the Dawson Resin synthesis over and over again, using different protocols that were supposed to work very well, we were never able to obtain peptide in sufficiently high yield or purity. However, the Wang Resin continued to yield peptide in extremely high yield (> 70%) and purity (> 95%); see Figure 2.

Untitled presentation

Figure 2: UV and mass spectrum of peptide made via Wang Resin

The extra thioesterification step in the case of the Wang Resin however means that the peptide’s N-terminus needs to be protected with an acetyl cap to prevent the cleaved peptide from forming dimers when its C-terminus is primed for nucleophilic attack preceding the addition of a thiol. In the past year, this so-called “capping” step has given Josh and I quite a bit of issues because the N-terminus is not the only part of the peptide that reacts with the capping reagent: the side-chain hydroxyl groups also get capped. This is clearly an issue because the hydroxyl groups are spatially very close to the bonds that we are trying to form! We have tried deprotecting the peptide in a mixture of 2:1 Methanol:Water at pH 9––although the deprotection works efficiently, the purification requires separating the peptide from the salts, resulting in over 70% product loss. This summer I found that using 1.1 equivalents of capping reagent instead of 2 equivalents as are often used results in > 95% fully deprotected peptide, that is therefore pure enough to proceed.

This summer I was also successful in the subsequent thioesterification step. As shown in Figure 3 below, I was able to create the thioester conjugate in high purity by incubating the peptide with a carboxylic acid-activating reagent (PyBOP) and the relevant thiol (thiophenol) in a basic environment. This was a great success because Josh and I had great trouble making the thioester conjugate. It is likely that the issue was not so much that we never made the compound as the fact that we could not detect it: as Niki had already found out before us, the thioester does not ionize well by chemical ionization (CI), but it ionizes much better by the softer technique known as electrospray ionization (ESI), which however requires sending the sample out for analysis. I also proceeded to (1) converting the peptide to its CoA conjugate and (2) loading it onto the PCP7-X didomain and (3) carrying out turnover studies with ComJ. However, the results are difficult to interpret decisively because the peptide is too dilute, and so the mass peak tends to merge with baseline noise. This upcoming semester I will work on optimizing the turnover protocol in order to obtain more concentrated peptide from the turnover reaction, either by upping the scale or concentrating the elute.

Figure 3: ESI-MS spectrum of thiophenol conjugate

Overall, I am very happy with my progress this summer and hope for even better progress this coming semester. I would like to thank the KINSC for providing me with a great scholarship and all the necessary equipment to carry out my research smoothly. I would like to thank Lou, my advisor, for helping me with my research even on maternity leave, and the entire lab for support. Lastly, I would like to thank Niki who started the project, and Joshua Bulos who worked on this project with me for over one year before his graduation this past May.

The Inside Scoop on Alpha-synuclein and Ntail/XD Proteins

Alpha-synuclein is a small, neuronal protein that is abundant at the presynaptic membrane and is estimated to play a role in the maintenance of synaptic vesicles. Alpha-synuclein is known to aggregate into fibrils called Lewy bodies, which is often noticed in patients with Parkinson’s disease. It is for this reason that aS is often studied. aS is also an intrinsically disordered protein, which means that unlike proteins whose function is dictated by a fixed secondary structure, aS will assume a partial conformation once bound. Since it is such a dynamic protein, it is more difficult to conclude upon its binding behavior and its role at the presynaptic membrane.

NMR structure of alpha-synuclein bound to SDS micelles (1XQ8) with eight original variant sites highlighted (blue is solvent exposed, orange is lipid bound)

NMR structure of alpha-synuclein bound to SDS micelles (1XQ8) with eight original variant sites highlighted (blue is solvent exposed, orange is lipid bound)

One way to ascertain the binding behavior of aS is to use vibrational spectroscopy. The Londergan lab uses site-specific vibrational spectroscopy to study the conformations and binding dynamics of intrinsically disordered proteins. These conclusions can be made because various functional probes such as thiocyanate or nitriles have spectra that differ depending on its environment. For example, the spectra for the thiocyanate probe will red-shift to a frequency below 2160 cm-1 if it is in a water-excluded environment, and it will blue-shift to ~2163 cm-1 if it is in a hydrogen binding environment (Konstantinovsky, 2016). Therefore using this technique it can be estimated at what locations aS is lipid-bound or solvent exposed.

IR of lipid-bound thiocyanate probe (red) versus solvent exposed probe (black) (Figure from Dan K.'s thesis)

IR of lipid-bound thiocyanate probe (red) versus solvent exposed probe (black) (Figure from Dan K.’s thesis)

The aS project was originally started by Haverford graduates Alice Vienneau and Daniel Konstantinovsky, and this summer, Kavita Shroff, Franklin Kostas, and myself were handed over the torch! Dan and Alice looked at eight sites over the length of the protein and found much agreement with published NMR studies. However, work from collaborator, Dr. David Eliezer at Weiss Cornell Medical School, demonstrated that aS is most commonly N-terminally acetylated in Parkinson’s patients, as well as others without the disease (Fauvet, 2012). Our goal for the summer was to express, purify and cyanylate N-terminally acetylated aS for eight of the original sites, as well as for three more additional sites.

In order to insert the thiocyanate probe into aS, the desired codon of interest is mutated to a cysteine by site-directed mutagenesis. Our plan of action was to first perform site-directed mutagenesis (SDM) on the original eight variant plasmids to correct a codon at the 139th position to be a tyrosine rather than a cysteine. Next we used SDM to mutate the wild-type plasmid to put a cysteine at the three new sites of interest. After confirmation by sequencing that we had indeed transformed the desired plasmids (credit goes to Kavita and Franklin for all of the SDM troubleshooting!), we were able to continue forward with expression and purification. For purification, we went about adapting the procedure from the Eliezer lab, which involved ammonium sulfate cuts rather than BOG detergent purification. Due to equipment differences, we did our best (with much guidance from Casey) to adapt the procedure, but in the end it was a much less straightforward process than anticipated.

While the aS project was in full force lead by Franklin and Kavita, I also had the ability to work with Ashley Guzman on the Nipah project (thanks Biochemistry SuperLab for teaching me how to express proteins). The Ntail protein of the Nipah virus is intrinsically disordered and assumes an ordered conformation after binding to XD to form the viral replicative complex (Hess, 2013). The Londergan lab used the thiocyanate probe to better understand the binding dynamics between the XD and Ntail proteins. Haverford graduates, Sara Hess, Rebecca Wai, and Renee King were able to compile IR scans for eleven sites along the Ntail protein. However, in order to better understand the differing bandwidths and shapes of peaks, the strength of the XD-Ntail interaction has to be known. Many attempts were made using ITC or isothermal calorimetry to determine the Kd (binding constant), but the process was not getting great results due to (believed) aggregation. Moving forward the next step was to attempt a pull-down experiment proposed by collaborator, Dr. Sonia Longhi. The Ntail protein has a histidine tag and therefore can be isolated on a nickel resin column. The pull-down experiment consists of binding the Ntail portion to the resin and then applying a XD-eGFP construct (kindly provided by the Longhi lab), which should bind to the Ntail. The fluorescence of the eGFP in the new complex can be used to determine the binding constant. We are still in method development, but plan to continue work into the fall to solidify a final procedure.

Working with GFP! This is during the expression of the XD-eGFP construct

Working with GFP! This is during the expression of the XD-eGFP construct

While I was working on the Nipah project, Franklin and Kavita were able to solidify a procedure to purify N-terminally acetylated aS. Our plan moving forward is to purify all of the expressed aS variants. Following cyanylation, which adds the thiocyanate probe to the mutated cysteine location, our plan is to use IR to look at the N-terminally acetylated aS in SDS, POPC/A and POPC/S/E lipid systems and compare our results with those obtained previously.

I have to thank the Frances Velay Science Fellowship Program for funding my opportunity this summer, as well as Professor Casey Londergan for all of his guidance and support (as well as for allowing me to join his lab)! I also have to thank the entire Londergan and Charkoudian labs for being incredible lab mates, and for all of their help troubleshooting and teaching me this summer. Last but not least, shout out to Kavita, Franklin, and Ashley for being amazing project partners!

References:

Fauvet, B.; Fares, M-B.; Samuel, F.; Dikiy, I.; Tandon, A.; Eliezer, D.; Lashuel, H. A. Journal of Biological Chemistry 2012, 287 (34), 28243–28262.

Hess, S. Cyanylated Cysteine Used to Examine the Ntail/XD Bound Complex of the Nipah Virus. Haverford College, 2013.

Konstantinovsky, D. Characterizing the Structural Distribution of Lipid-Bound Alpha-Synuclein by Site-Specific Thiocyanate Infrared Probe Groups and Molecular Dynamics Simulations. Haverford College, 2016.

Spectrophotometric Titrations

This summer my research focused on the use of spectrophotometric titrations to determine equilibrium binding constants between the metal complex Co[DIG3tren]2+ and various anions. DIG3tren (Tri-N’,N’’-diisopropylguanidine tren) is a ligand synthesized in the Scarrow Lab characterized by three guanidine based groups connected to a central tren.  The tripodal, tetradentate nature of DIG3tren leaves an apical binding sight that various anions bind to with differing strengths based on the polarity, shape, and solvent interactions of each unique anion.  The UV-Vis absorption spectra of a Co[DIG3tren]2+ solution changes as anions bind to the available apical sight giving the basis for the subsequent analyses used to determine the equilibrium binding constants.

An example fit of the molar absorptivities.  The red outline is the Co[DIG3tren] complex and the green outline is the Co[DIG3tren]SCN  coumpound

An example fit of the molar absorptivities. The red outline is the Co[DIG3tren] complex and the green outline is the Co[DIG3tren]SCN coumpound

 An Ocean Optics spectrophotometer was used to capture thousands of UV-Vis spectra over the course of a titration that would last between 1-3 hours.  The collected spectra were organized in IGOR (a data analysis environment similar to Origin) before being componentized with the aid of a macro written in Igor’s programming language.  Principle component analysis, a process familiar to those with linear algebra experience, breaks data sets such as the collected spectra into linearly independent variable.  The linearly independent variables in the case of the spectrophotometric titrations were component spectra the total number of which was assumed equal to the total number of contributing species in solution. The component spectra are given unique weights at each moment in time, which correspond to their relative contributions to the observed spectra.  A nonlinear, least-squares refinement algorithm, also written in Igor’s programming language is used to fit both the coefficient weights and molar absorptivities of the absorbing species.  This fitting function is dependent upon the binding constant of the equilibrium (commonly abbreviated K), and by adjusting K the mathematically best fit can be reached.

During the beginning of the summer the procedure used (slow, continuous addition of titrant throughout the duration of the titration) was similar to that used during the research done in previous years.  Unfortunately, the data collected, similar to the data from previously completed experiments, was imperfectly fit using the nonlinear, least-squares refinement program.  The simplest explanation for the variance between the fits and observed data was imperfect mixing, a consequence of the fact that a true equilibrium between the metal complex and the anion is not reached instantaneously.  The problem was solved by modulating the addition of titrant with an Arduino microprocessor and a DC switched power supply.  A five second burst of titrant could be added before allowing for a minute of mixing of the solutions.  Between 90 and 150 discrete additions were made this way per titration and an IGOR script was written to find the addition points among the collected spectra and average the absorptions after equilibrium was reached.  The averaged absorption spectra constituted a smaller but more accurate data set to componentize and fit.

Previous DIG3tren equilibrium research had revealed K to have fair amount of solvent dependence, a likely consequence of solvent interactions.  Little research was done, however, on the temperature dependence of the equilibrium constant which was the primary focus of this summer’s research.  All of the titrations performed this summer used 200-proof ethanol as a solvent but varied the temperature over a range of -20oC to 55oC.  The equilibrium constants could then be plotted in the form log(K) vs 1/T to produce a van ‘t Hoff plot.  Because the resulting van ‘t Hoff plot revealed a linear temperature dependence the change in entropy and enthalpy could be pulled off of the slope and intercept of a line of best fit.

A van 't Hoff plot produced by plotting log(K) over 1/Temp.  Change of entropy and enthalpy can be pulled off of the line of best fit.

A van ‘t Hoff plot produced by plotting log(K) over 1/Temp. Change of entropy and enthalpy can be pulled off of the line of best fit.

The summer was long enough for three anions (acetate, thiocyanante, and fluoride) to be thoroughly investigated, each posing its own challenges.  For instance, acetate bound fairly weakly to the metal complex requiring upwards of thirty equivalents to be added in order to achieve a set of data that would lead to reproducible and accurate fits, while thiocyanante tended to have precipitation problems if used at the concentrations required for the acetate titrations.  In the end, however, reproducible equilibrium constants for a range of temperatures were found to three significant figures for all three anions.  Fluoride was the only anion that proved difficult to demonstrate linear equilibrium temperature dependence, with such behavior only being observed in the -5oC to 25oC temperature range.

An example fit of component coefficients.  The equilibrium constant for a given titration is a parameter of the fit.

An example fit of component coefficients. The equilibrium constant for a given titration is a parameter of the fit.

The simple equilibrium expression taught in general chemistry courses, unfortunately, relies on several key assumptions that weaken its ability to match observational data.  Debye-Huckel theory provides a more powerful and accurate model, and thus, the original nonlinear, least-squares refinement program was adapted using Debye-Huckel theory to provide a more accurate estimation of the equilibrium constants.  Other parameters, such as relative concentration and titrant addition rate, could also be refined to lead more accurate fits.

A piece of the code used in fitting the molar absorptivities of the absorbing species.

A piece of the code used in fitting the molar absorptivities of the absorbing species.

Hopefully, there will be time to test more anions during the school year leading to a more complete data base of equilibrium constants and a better understanding of the anion properties that lead to stronger or weaker binding.

Conquering Conclusions Through Zebrafish Genetic Research

In everyday life, people like to jump to conclusions. For instance, that guy who cut you off on the highway, total jerk; that girl who failed her test, really dumb; and that guy with really long flowy hair under a baseball cap, totally a lax player. With this thinking applied to biological research, that should mean when a genetic screen for a behavioral mutant yields a specific genetic mutation, they’re automatically linked, right? Well, like the prior misguided conclusions, this thought process is way off.

The Big Questions

In biological research conclusions made between mutated genes and various physical outcomes, or phenotypes, cannot be made so quickly. Many biological systems are controlled by more than one gene and sometimes are affected by some other biological function which may also be controlled by various genes. The possibilities are endless in terms of what can really cause these phenotypes. If this is the case, and we don’t get answers immediate from these screens, why do we have them and why are the results deemed so important?

In the case of my research this summer, with Haverford Professor Roshan Jain, the results of these screens allowed us to answer a larger question in biology: How do genes affect behavior?

First off, why do we care? In many cases the “behaviors” which we are talking about in humans are ones which are impairing, like schizophrenia and epilepsy.

So what do we know already that we need to embellish? Well, we (as a scientific community) know that RNA is transcribed from DNA and is then translated into protein. Both on the RNA and protein level, it is understood that cells in neurons and the neural circuits they are a part of are directly affected, depending on the genetic products which were transcribed and translated. Even some of the most minor genetic changes can have a huge effect on behavioral neural circuits. Specifically, the Jain lab and I this summer looked into the acoustic startle circuit and changes which cause learning defects in a specific behavior called habituation.

Ignorance is Bliss

Habituation is a simple form of non-associative learning which occurs when organisms decrease the number of responses or cease to respond to non-consequential stimuli. This type of behavior is very important in terms of research because its implications in survival are critical, and can represent a split second action making the difference between life and death. This summer, we specifically observed this behavior in larval zebrafish (Figure 1) due to ease and accessibility. 

A 30 day old larval zebrafish.

Figure 1. larval zebrafish (age: 30 hrs)

In their larval stage, zebrafish are transparent, making it easy to observe developmental structure as well the general circuitry of neural systems; making observational data simple to collect. Additionally, it is very easy to perform experiments on a large scale using zebrafish as a model organism, allowing mass testing and fewer potential errors due to the size of the experimental pool. Using zebrafish, habituation is also very easy to track on a large scale and can be defined using few factors. 

The habituation mutant which I focused my research on this summer, ignorance is bliss (72FAGA), was identified in a forward genetic screen performed by Professor Jain as well as several members of the Granato lab at the University of Pennsylvania. ignorance is bliss has a mutation of a candidate gene known as ap2s1, a gene controlling the sigma subunit of the AP-2 Complex which targets membrane proteins for endocytosis, a biological trafficking process, and is believed to be necessary in decision making as well as habituation.

Solidifying Findings

Since we can’t jump to conclusions with these findings, a major question which was necessary to address before making any connections/assumptions was “How do we know this specific genetic mutation is what is causing abnormal habituation?” This question was answered using a technique known as a complementation cross. The goal of this test is to cross two organisms, zebrafish here, which have mutations in the candidate gene in question, for our purposes ap2s1. Once the cross is complete, the larval offspring of the cross are genotyped and behaviorally tested to determine whether or not they are mutants for the habituation phenotype and contain two differently mutated ap2s1 alleles from the mutant parents. This is the desired outcome will show that that the specific genetic mutation observed in the genetic screen is truly associated with the mutant habituation phenotype. This is called a failure to complement. A generic outline of this process is shown below (Figure 2).

Figure 2. complementation cross outline

As seen above, the mutant which was generated for the cross to complement with 72FAGA (ignorance is bliss) was an ap2s1-CRISPR mutant, where a targeted insertion/deletion was put into the candidate gene.

Each of these offspring were behaviorally tested using a 16 well plate attached to an amplifier, which repeatedly exposed the larvae to acoustic stimuli, triggering a response. Depending on how the responses changed over time, habituation was calculated. Once behavioral data was collected, the larvae were used as DNA for genotyping. Genotyping revealed whether the tested larvae were heterozygous for one of the mutant ap2s1 alleles, wild type with regular ap2s1 copies, or double mutant trans-heterozygotes for both the CRISPR and 72FAGA ap2s1 genes. This was done through restriction enzyme digest and polymerase chain reaction. The samples were run out on a gel and the bands were observed to make distinctions between larvae. Examples of these digests are seen in Figure 3.

Figure 3. gel results of restriction enzyme digest for 72FAGA and ap2s1-CRISPR mutations – In the left image the restriction site was located in the 72FAGA specific mutation, so the double bars represent the presence of a mutated ap2s1 allele. In the left image, the additional bar represents uncut DNA in CRISPR mutants. The cut site for this specific gel’s DNA was located within the wild type ap2s1 gene, so a CRISPR carrier would have both cut and uncut DNA.

The Results

Figure 4. results of the complementation cross

The graph in Figure 4 shows the combined results from the genotyped larvae and the behavior which was tested prior to genotyping. This was especially exciting because the graph reflects exactly what I wanted to see. While wild type behavior (>50% habituation) was seen in the hets and wild type larvae, mutant behavior (<50% habituation) was seen in some of the 72FAGA mutants (expected as a control) as well as in the trans-hets. This means that the fish failed to complement and ap2s1 is in fact the mutated gene associated with a lack of habituation. With this information at the end of the summer, it is now possible for the Jain lab to raise these fish up for attempted rescue of wild type behavior when they are older, utilizing what we know about them genetically already and our ability to test them.

My Thoughts on the Summer

I know this blog post is fairly dry, but my overall experience with the Jain lab was not. I very much went into this summer thinking that I was done with biology, and this was my last ditch effort to see if I was wrong. I’m a current math major, who frankly has been disenchanted with my experience with biology thus far at Haverford. However, this summer rekindled a lot of the love which I had for the subject. I found myself being very inquisitive in lab because I was genuinely interested in the topics of research and was deeply engaged in figuring out what years’ worth of research would yield. What I found was that there is a tremendous amount of work which goes into research, from asking the big “Why?” question, to getting down into the gritty details, and eventually churning out a result and a publication. Yes, it’s grueling, and yes, it’s challenging, but the results are worth it, and the feeling of pride/accomplishment is unlike anything else. I think that experiencing Professor Jain’s research at the end of its 4-5 year time span really allowed me to be excited by the outcomes of it all, and become more motivated to ask my own big questions. Honestly, figuring things out and realizing that you have accomplished something is one of the most exciting and rewarding events I’ve experienced while in college.

Although I may stay a math major, that doesn’t mean that I won’t keep up my biology studies. In fact, this summer may have even moved me to produce a math bio related senior thesis, and merge my two academic loves: a subject I’ve always loved, and another which I’ve luckily and thankfully become reacquainted with.

Calreticulin and California

Image

“Mom, what if there is a cancer treatment that we already have but didn’t know that it treats cancer.  Like, what if we just injected lemon juice into the blood of all cancer patients and that helped them.  What if it was just that simple, lemon juice, and we don’t even know it yet”    

I obviously didn’t understand the catastrophic effects of blood acidosis at age ten. A huge thank you to my poor parents, who enduring years of “what if” questions so that I may one day bother my postdoctoral mentor, Kristopher Marjon, with them.  

At the beginning of the summer I had the incredible opportunity to travel to Stanford University through the generosity of the Frances Velay Science Fellowship Program.  My goals for the summer included learning new scientific techniques and skills, contributing relevant scientific data to the field of immunotherapy and cancer research, as well as establishing a professional relationship with a renowned research institution.  The Weissman laboratory not only has historic legacy as a beacon of early stem cell research, but continues to be internationally competitive in the scientific fields of pathology, immunology, immunotherapy, and cancer research.  I am so grateful for this experience that was made possible by so many of those supportive of me at home, Haverford College, Stanford University, and the committee board of the Francis Velay Fellowship.

Saying that I’ve learned quite a deal this summer is a vast understatement of my time at Stanford University. The data that I have accumulated contributes to an ongoing research project that aims to better understand the cellular trafficking of a protein known as calreticulin.  It has become increasingly important to understand the location, signaling, trafficking, and possible transfer of calreticulin from one cell to another because it has been shown to play an important role in maintenance of a cell’s life cycle and has the potential to be incorporated in cancer-immunotherapy.

In the never-ending struggle for homeostasis, there is a constant uphill battle for cells, tissues, and organs to maintain control over the many distinct and complex pathways and mechanisms.  Specifically, there is a delicate balancing act between pro-phagocytic and anti-phagocytic signals, which serve to either to signal or sequester alerts to the innate immune system to phagocytose (engulf or destroy) another cell.  The maintenance of these signals allows for regular and healthy programmed cell removal when a specific cell is either dysfunctional or no longer useful. If a cell is able to evade or manipulate these progressions, it runs the chance of developing into a cancer.

I think of a cancerous cell as a cell that is no longer apart of the organism’s “self”.  For a cell to be apart of an organism, it must cooperate.  A cell that is apart of the self must obey the signaling pathways and inherent mechanisms possessed by the organism, it must behave according to the systems of the body enforced upon it. A cell that belongs to the “self” must operate as the smallest part of “the whole” and must dedicate its entire being to helping the tissue, organ, and system remain healthy.  Cancer cells do not do this, in fact, they individually select for destructive behaviors and compete against other non-malignant cells for resources.  Not only can cancer recklessly proliferate within an organism, they are somehow able to do so under-the-nose of the organism’s immune system (Chao, 2011.  The Weissman laboratory asks: What proteins or cellular interactions are normally in place to prevent cancer growth, how does cancer circumnavigate these processes to proliferate, and what can be done to reverse this process? These are just a few of the many questions the Weissman laboratory is looking to answer.

To answer some of the questions about how a cancer proliferates in an organism, it is important to first start with its means of communication. At Stanford University, there is no better protein to start with than calreticulin (CRT). CRT is most commonly recognized as a endoplasmic reticulum chaperone protein where it aids in the folding of other proteins produced by the cell. However, in a unique twist of events, CRT is sometimes trafficked to the surface of the cell and takes on the role of a cell surface protein. On the cell membrane, it acts as an “eat me” pro-phagocytic signal and recruits the help of the organism’s innate immune system to clear the cell that it decorates (Chao, 2010). This process of cell removal is a clever way for the cell to internally recognize that it needs to be cleared, or, as we will explore, transfer CRT to a cell that is in need of being cleared. The integrity of our body and our health depends on balancing “eat me” and “don’t eat me” signals to direct programmed cell removal. This allows for healthy cells to remain in the tissue to resume their normal function and, alternatively, destroying cells that have the potential to do harm.

This pathway is dependent on the fact that CRT is able to leave its job in the ER and take on a new role at the cell surface, however, the mechanism by which CRT is trafficked to the surface of the cell is not well understood. Even further, it has been suggested that CRT can be used by certain cells of the immune system to decorate other cells, allowing for the accumulation of CRT on the target cell’s surface to act as an indicator to the immune system to aid in its clearance of other cells. In order to understand the role of CRT in programmed cell removal we used a thioglycollate peritonitis model. Thioglycollate is used as an irritant and is injected into the peritoneal cavity of mice, which results in inflammation and induces recruitment of immune cells to the peritoneal cavity. In a study by Lagasse and Weissman in 1994,

Figure 1. Kinetics of neutrophils and macrophages in the peritoneum after thioglycollate injection. These data demonstrate the percentage of reactive cells in the peritoneal cavity at time 0 hours (the point of injection).  Over the next four hours, the reactive neutrophil population (the first responders of the immune system that react quickly to an infection)  climbs from 0% to ~65%, until about hour 5, but then falls steadily from ~65% to 0% over the next 19 hours.  The reactive macrophage population at time 0h is relatively high because of the resident macrophage population, which is the predominant immune cell that is within the peritoneal cavity. After injection with thioglycollate the peritoneal cavity is overwhelmed by neutrophils and therefore the macrophage population drops dramatically by 4 hours.  Then, from time point 4 hours to 24 hours, the reactive macrophage population climbs from about ~10% to ~60% as recruited monocytes mature into macrophages which is occurring at the same time the neutrophil population is falling by being phagocytosed by the macrophages. Figure adapted from original publication (Lagasse et al 1994).

Figure 1. Kinetics of neutrophils and macrophages in the peritoneum after thioglycollate injection.                             This data demonstrates the percentage of reactive cells in the peritoneal cavity at time 0 hours (the point of injection). Over the next four hours, the reactive neutrophil population (the first responders of the immune system that react quickly to an infection) climbs from 0% to ~65%, until about hour 5, but then falls steadily from ~65% to 0% over the next 19 hours. The reactive macrophage population at time 0h is relatively high because of the resident macrophage population, which is the predominant immune cell that is within the peritoneal cavity. After injection with thioglycollate the peritoneal cavity is overwhelmed by neutrophils and therefore the macrophage population drops dramatically by 4 hours. Then, from time point 4 hours to 24 hours, the reactive macrophage population climbs from about ~10% to ~60% as recruited monocytes mature into macrophages which is occurring at the same time the neutrophil population is falling by being phagocytosed by the macrophages. Figure adapted from original publication (Lagasse et al 1994).

it was demonstrated that the percentage of reactive neutrophil and macrophage populations in the peritoneal cavity of a mice injected with thioglycollate (an agent to produce peritonitis, inflammation of the peritoneal cavity) had similar kinetics. This finding was partially expected as neutrophils are the first responders during an inflammatory insult but are typically clear by the macrophage populations because they undergo apoptosis rapidly. The fundamental finding from this study is that mice overexpressing BCL2 in the neutrophil population (bcl-2) had similar clearance rates compared to control mice within the peritoneal cavity that was independent of apoptosis. Lagasse’s findings are critical in establishing a clear inverse relationship between the two cell populations, and indicated that there was a mechanism or switch that occurred that was independent of apoptosis that induced the removal of the neutrophils from the peritoneal cavity. These findings were some of the first demonstrations of program cell removal and have led the Weissman Laboratory to their most recent area of interest. Unknown to the research team at that time was the role of calreticulin and other molecules that mediate “eat me” and “don’t eat me” signals to the macrophages.

Revisited by Stanford University in 2016, a similar relationship emerges between cell surface levels of CRT and the time course of this experiment.  Quantified flow cytometry, Figure 2 shows CRT surface levels from macrophages (the red line) drops dramatically from a mean fluorescent intensity (MFI) of 1000 to ~500, while the MFI of surface CRT on

Figure 2. Cell surface levels of calreticulin on thioglycollate recruited neutrophils and macrophages. Neutrophils over expressing BCL2 (PMNs) and macrophages (MAC) were isolated from the peritoneum at indicated time points after thioglycollate injection. Cells were stained for surface markers to distinguish between PMNs and macrophages and analyzed for cell surface levels of calreticulin by flow cytometry.

Figure 2. Cell surface levels of calreticulin on thioglycollate recruited neutrophils and macrophages.                                Neutrophils over expressing BCL2 (PMNs) and macrophages (MAC) were isolated from the peritoneum at indicated time points after thioglycollate injection. Cells were stained for surface markers to distinguish between PMNs and macrophages and analyzed for cell surface levels of calreticulin by flow cytometry.

the neutrophils over expressing BCL2  (the black line) rises from about 300 to 1000. In summary, these two experiments shows that the dramatic rise of the neutrophil population at the site of infection is followed by the rise in macrophages, which then phagocytose the neutrophils.  And, in addition, the neutrophils arrive at the site with relatively low cell surface CRT levels but accumulate a high cell surface level of CRT right before they are engulfed by the macrophages.  Conversely, the macrophages arrive with high levels of cell surface CRT but have much lower levels by the time they start the phagocytosis process.  Not only does this data imply an association between CRT transfer from the macrophage to the neutrophil, but it also points to CRT as being an important pro-phagocytotic signal independent of apoptosis.

At the beginning of my time at Stanford, all of these experiments had been previously performed in an in-vivo model. While this is an excellent model to utilize and aid in our understanding of  the role of calreticulin in programmed cell removal, it does not allow for further genetic manipulation or convenient means of interrogating and investigating of the different cell populations and trafficking of calreticulin. For these reasons, my summer project was focused on recapitulating the in-vivo findings by using laboratory cell lines.  This would allow the Weissman laboratory to interrogate at multiple levels and have higher fidelity in understanding the role of calreticulin in programed cell removal. The cells that I incorporated into my research were immortalized mouse macrophage cells, known as J774, forced to over express CRT tagged with mCherry. mCherry is a fluorescent protein that can be utilized to indicate where the cellular localization of the protein (in this case calreticulin) is located. Forcing higher levels of CRT fused to the mCherry protein creates a built-in tracking device for CRT. So, wherever mCherry was observed and expressed in the cell, it could be assumed that CRT was located there as well. Instead of neutrophils, the target cells that were used in my assays consisted of an immortalized human colon cancer cell line known as SW620. The SW620 cell line gave a surface to which the CRT-mCherry was able to bind when being co-incubated with the J774-CRT mCherry cells. My assays were able to be quantified for total and surface levels of CRT by flow cytometry, location of CRT and other proteins of interest with immunofluorescence imaging, and possible soluble CRT in the supernatant of our co-incubation experiments by enzyme-linked immunosorbent assay (ELISA).

Figure 3. Transwell assay. Target cells are placed on the top well separated from the bottom chamber by a porous (0.4 μm) filter. The bottom chamber contains the macrophages.

Figure 3. Transwell assay.                        Target cells are placed on the top well separated from the bottom chamber by a porous (0.4 μm) filter. The bottom chamber contains the macrophages.

The transwell assays that were performed consisted of the two populations of cells, J774 CRT mCherry macrophages and SW620 colon cancer cells, configured like the image in Figure 3.  The J774 CRT mCherry cells populated the bottom of the wells (shown in blue) and the SW620 cells sat on top of the filter (shown in red).  Both cell populations were separated from each other by a 0.4 um filter through which only the supernatant bathing the cells and small proteins, such as CRT, could pass. Since it has been previously suggested that macrophages can decorate target cells with CRT, I was primarily looking for a transfer of CRT-mCherry from our J774-CRT mCherry macrophages to the SW620 cancer cells. Figure 4 shows an averaged trend of CRT on the cell surface of the cell populations. There is a trend showing a decrease in MFI for CRT-PE from J774 CRT mCherry macrophages alone or co-cultured with the SW620 cancer cells (solid red bar) to

Figure 4. Calreticulin transfer from macrophages to target cells. Macrophages were cultured on the bottom well, expressing calreticulin fused to mCherry, in the presence or absence of SW620 cells. Cells were harvested from the wells then calreticulin was detected by assessing presence by mCherry signal or by antibodies against calreticulin and analyzed by flow cytometry.

Figure 4. Calreticulin transfer from macrophages to target cells.                    Macrophages were cultured on the bottom well, expressing calreticulin fused to mCherry, in the presence or absence of SW620 cells. Cells were harvested from the wells then calreticulin was detected by assessing presence by mCherry signal or by antibodies against calreticulin and analyzed by flow cytometry.

J774 CRT mCherry macrophages that were co-cultured with the SW620 cancer cells (red/black patterned bar). Additionally, there is a trend showing an increase in MFI for CRT-PE in surface CRT from SW620 cancer cells not co-cultured with J774 CRT mCherry macrophages (solid grey bar) compared to the SW620 cancer cells co-culture with the J774 CRT mCherry macrophages (grey/black patterned bar). This suggests that our control J774 CRT mCherry macrophages have a greater amount of surface CRT compared to the J774 CRT mCherry macrophages co-cultured with the SW620s, supporting the hypothesis that the J774 CRT mCherry macrophages can transfer their CRT.  Additionally, the control SW620 cancer cells, on average, have a lower amount of surface CRT than the SW620 cancer cells co-cultured with J774 CRT mCherry macrophages, suggesting that they are able to receive CRT from the J774 CRT mCherry macrophages.

We also wanted to determine if there was a difference in the calreticulin that was released into the medium during the coculture. Figure 5 demonstrates the levels of soluble CRT in the

Figure 5 Soluble calreticulin is generated by macrophages. Macrophages were cocultured alone or in the presence of SW620 cells. Supernatant was harvested and tested for the presence of calreticulin by ELISA.

Figure 5 Soluble calreticulin is generated by macrophages.            Macrophages were cocultured alone or in the presence of SW620 cells. Supernatant was harvested and tested for the presence of calreticulin by ELISA.

supernatant from these transwell experiments, which were determined by ELISA.  Measured in nano grams per milliliter, J774 CRT mCherry macrophages that were not co-cultured with SW620 cancer cells show a high level of soluble CRT in the supernatant, ~9.5-10 ng CRT/mL, while SW620 cancer cells not co-cultured with J774 CRT mCherry macrophages produce ~0.25-0.5 ng CRT/mL, a negligible amount.  The difference of CRT in the supernatant between the experimental co-culture group and the control is accounted for by the transfer of CRT from the from the J774 CRT mCherry macrophages to the SW620 cancer cells. These data also highlight that macrophages readily make calreticulin and secrete it into their local environment.

Assessment of CRT localization was also an important part of my research project in the Weissman Laboratory.  While CRT is an ER protein anchored by its KDEL ER retention sequence, it uniquely has the capability to travel to the cell surface.  For most cells, it has been estimated that up to 95% of the cell’s CRT levels are contained within the cell, affording little on the surface. Therefore, I also sought out to determine the cellular localization of calreticulin and potential interactions within and on the cell in both primary cells as well as in cell lines that I incorporated in my research projects. Previous observations demonstrate that CRT is expressed at very low levels, if at all, in neutrophils and at a much higher level in macrophages. In an effort to investigate previous findings by the laboratory I used human primary macrophages (hMacrophage) and neutrophils (hPMN). hMacrophages and hPMN cells were stained with a rabbit polyclonal antibody against CRT and a biotinylated ligand that stains modified sugars.  Staining for CRT aimed to identify cellular localization of this protein, while staining for modified sugars aids in determining all possible binding sites for CRT. Additionally, all cells were stained with DAPI to observe the nucleus.

Figure 6A shows a robust staining for CRT in hMacrophages within the cell, most likely located within the ER and vesicles.  Figure 6B shows staining for modified sugars on the hMacrophages, revealing that a majority of possible binding sites for CRT are located on what appears to be the surface of the cell.  There is little localization of possible CRT binding sites within the cell, opposite of what we see from the CRT staining.

Figure 6. Calreticulin and MSS stain on human macrophage cells. Human macrophages generated from peripheral blood mononuclear cells were stained fix, permeabilized and stained for Calreticulin (A) or a modified sugar stain, MSS, (B). Nucleus was counterstained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 6. Calreticulin and MSS stain on human macrophage cells.         Human macrophages generated from peripheral blood mononuclear cells were stained fix, permeabilized and stained for Calreticulin (A) or a modified sugar stain, MSS, (B). Nucleus was counterstained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

While the same staining platform was applied to hPMN, there were very different results.  In contrast to the hMacrophages, there is a lack of CRT signaling as seen in Figure 7A and the signal for the modified sugar staining (MSS) appears quite robust in Figure 7B.  This is an incredibly interesting finding since it further supports and validates the CRT’s path of cellular trafficking and transfer from a macrophage cell to the target neutrophil cell.  The results of CRT and modified sugar staining from the hMacrophage and the hPMN cells show that the hMacrophage cells have an abundance of CRT and a deficit of CRT binding sites, whereas the hPMN (the hopeful CRT target cell) has an abundance of CRT binding sites but extremely low CRT levels.  These immunofluorescence images confirm what previous research has shown

Figure 7. Calreticulin and MSS staining on human neutrophil cells. Human neutrophils were isolated from peripheral blood and fixed and stained for Calreticulin (A) and MSS (B) and nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 7. Calreticulin and MSS staining on human neutrophil cells.       Human neutrophils were isolated from peripheral blood and fixed and stained for Calreticulin (A) and MSS (B) and nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 8. BTK localization in macrophages. Peritoneal macrophages were isolated and fixed and stained for BTK and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 8. BTK localization in macrophages.                                               Peritoneal macrophages were isolated and fixed and stained for BTK and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Still, another question remains: How does the cell traffic CRT from the ER to the cell surface?  Prior research done by the Weissman laboratory suggests that bruton’s tyrosine kinase (Btk) aids in the trafficking of CRT to the cell surface.  Immunofluorescence staining of Btk on macrophages derived from mice have a diffuse staining profile as demonstrated in Figure 8.

To assess the relationship between Btk and CRT, a Proximity Ligation Assay (PLA) was used.  This assay has incredible sensitivity because it is a fluorescent assay in which fluorescence will only be observed if two proteins of interest are in close proximity to each other, or otherwise co-localized. This is accomplished by using primary antibodies against CRT and Btk and then applying probes that have oligonucleotide chains (sequences of DNA). If the two proteins are approximately 30-40 nm apart, the oligonucleotide chains interact to form a single stranded DNA ring, which is then replicated through rolling cycle amplification and a pile of DNA that has specific repeating sequences is formed.  Fluorescent tags can attach to these repeating sequences, creating a signal that is strong enough to detected and quantified through immunofluorescence imaging.

Figure 9 demonstrates a Btk and CRT staining with PLA.  The points of red punctate on Figure 9B indicate areas that are positive for protein-protein interaction.  It is interesting that most of these points appear to occur at a distance from the nucleus, suggesting that most of the interaction of CRT and Btk occurs close to the cell surface.  There is a relatively low level of interaction signal even for an experimental group.  However, this might be due to the fact that resident peritoneal macrophage cells derived from mice were used for this assay.  If these peritoneal macrophages were stimulated with lipopolysaccharide (LPS) or from mice that were injected with thioglycollate to induce peritonitis, more punctate indicating interaction between Btk-CRT may be expected.

Figure 9. Colocalization of BTK and Calreticulin occurs in unstimulated macrophages.                                                                                                                                Proximity ligation assay was performed to determine if BTK and Calreticulin colocalize together in unstimulated macrophages. Cells were fixed and permeabilized and PLA assay was carried out as described by the manufacture. Cells were incubated with secondary antibodies only (A) or with antibodies against BTK and CRT (B). Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

However, just observing our proteins of interest with immunofluorescence imaging isn’t enough to determine their location.  In order to gather more definitive evidence, the last piece of my project at Stanford University attempted to optimize staining for CRT and modified sugars against different biomarkers.  The purpose of this experiment was to take a biomarker for the cell surface, vesicles, ER, or other organelle and cross stain it with CRT or our modified sugar stain (MSS). If there is overlap between the two stains, such as the CRT protein and a cell surface biomarker, we can be confident that our images are really displaying CRT inhabiting the cell surface.  Many of the stainings produced spectacular images, such as that in Figure 10.  J774 parental macrophages were stained with a panel containing rabbit polyclonal antibody against CRT, phalloidin which stains actin which aids in deciphering what is intracellular, and DAPI to visualize the nuclei.  In the unstimulated macrophage cells, it is apparent that most of the CRT is within the cell.  There is, however, some cross-over between the phalloidin stain and the CRT stain and some staining for CRT outside of the phalloidin stain suggesting staining at the surface of the cell, confirming that CRT is observed on the cell surface.

Figure 10. CRT cellular localization in Macrophages. Macrophages were fixed, permeabilized and stained for Calreticulin, phalloidin and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 10. CRT cellular localization in Macrophages.                               Macrophages were fixed, permeabilized and stained for Calreticulin, phalloidin and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

A similar stain was done for possible CRT binding sites by staining J774 parental cells with our modified sugar stain (MSS), phalloidin, and DAPI in Figure 11.  Some interesting differences arise between this staining panel and the staining that was performed against the CRT protein.  There is a robust stain for our modified sugar stain (MSS) within the cell, what seems to be marking many vesicles and the ER. Additionally, there are more MSS punctate that are overlapping with the phalloidin stain and outside of it. This suggests that there are many possible CRT binding sites within the cell as well as on the cell surface.  However, the catch-all clause for the MSS is that, while it picks up on all possible CRT binding sites, it also a lot of other binding sites that many other proteins besides CRT can also bind to.

Figure 11. Calreticulin cellular localization in macrophages. Macrophages were fixed, permeabilized and stained for Calreticulin, phalloidin and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

Figure 11. Calreticulin cellular localization in macrophages.             Macrophages were fixed, permeabilized and stained for Calreticulin, phalloidin and the nucleus was stained with DAPI. Images were acquired using a LSM710 Ziess confocal microscope and all images were examined with a 63x oil immersed objective.

As my my project winds to an end, I concluded that calreticulin is, in fact, a very tricky protein to study.  My attempt to recapitulate findings of prior experiments done in the Weissman lab proved difficult when we did not see as robust of a transfer of CRT from our J774 CRT mCherry macrophages to the SW620 cancer cells.  While this was disappointing, there are many reasons as to why this may not have been the best model to use. Both cells lines are immortalized, meaning that they have a much longer life span than most-derived cells, calling in question the normality of their pathways and mechanisms. Additionally, the cell lines used where mismatching species; the J774 CRT mCherry macrophages being an immortalized mouse cell line, while the SW620 cancer cells are an immortalized human cell line. Prior experiments used derived macrophages and neutrophils, while we were used two cancerous cell lines. All of these factors could have played a part in why the CRT transfer signal was not as robust as earlier experiments had shown, but, even the suggestion of transfer from these experiments gave insight into CRT’s trafficking abilities.

Interestingly, CRT was readily released by the J774 CRT mCherry macrophages into the supernatant bathing the in co-culture. This sparked many questions, such as, whether or not cell density was a factor (for either the macrophage or the cancer cells) to be taken into consideration while assessing secreted CRT.  If, in future experiments, cell density proves to be a factor, then there would be an interest in investigating the possible mechanisms by which macrophages can assess their own population’s density and release appropriate levels of CRT.  Or, conversely, how they may assess their target cell’s density and make important decisions on how much CRT to release per target cell. It is also unknown at this point whether secreted CRT is in anyway modified compared to CRT adherent to the ER or the cell surface, or, if our secreted mCherry-CRT actually retains the mCherry fusion protein.  All of these details are important is understanding the nuisances of CRT export and transfer.

Aside from learning new microscopy techniques on some incredible equipment, I had success with assessing interesting patterns of CRT localization on primary human and mouse macrophages and neutrophils. It was great to know that what we were observing on the microscope held true to prior research and data when macrophages were observed with a high levels CRT protein and low levels of CRT binding site content, and neutrophils possessed a low CRT protein content and high levels of possible CRT binding sites. The assays for antibody staining experiments produced a large number of remarkable photos that visualize the relationship between a macrophage cell and a potential target, really breathing life into the CRT dependent relationship built between these two cell populations. These images were, and will continue to be, supported by cross examining with known, multiple biomarkers along with selective staining for the CRT protein or possible CRT binding sites to create the most definitive panel for CRT localization.

While assessment of co-localization of CRT and Btk by using the PLA has many more stages of optimization to undergo, it was incredible to quantify the closeness of these two proteins interactions.  Being able to assess their co-localiztion in an unstimulated cell population, and still able to get a signal, only makes me more curious to know how much more CRT would co-localize with Btk under stimulated or stressed conditions.

It saddens me to conclude my time at Stanford University, yet I am excited for the future directions of this project! In the coming months I am hoping to hear back on additional projects such as the growth of different biomarker staining for CRT, MSS, and Btk, or PLA assays performed on primary cells or cells treated with LPS.  I know the Weissman laboratory will continue to work towards answering the questions that they originally posed, as well as other questions that I have contributed along the way, such as: What innate mechanism does the cell possess in order to know to send CRT to the surface of the cell? What happens to the CRT’s KDEL signal once it leaves the ER?  Does CRT gain new modifications or a new homing sequence in order to be exported? Are there other proteins or pathways involved in the trafficking of CRT? These questions, along with many others, are in amazing hands as I leave Stanford University to rejoin my fellow Haverford students.

I would like to give a final thank you to all who made my time at Stanford University a possibility.  Thank you to Irv Weissman and his entire lab for being so welcoming; selflessly inviting undergraduate student researchers to get a taste an elite and competitive laboratory experience.  Thank you to Dr. Kristopher Marjon, my post-doctorate mentor, who I tortured day-in and day-out and who should be rewarded for his patience and compassion. Thank you to my parents, grandparents, and other extended family who helped support me on this really incredible trip, as well as friends Marie Vastola and Yvonne Wilson.  Thank you to Rachel Hoang and Tim Chaya, as well as the rest of the Haverford College KINSC faculty and department for investing an incredible amount of time and emphasis on the importance of summer research and creating the opportunities for all the summer KINSC scholars. And finally, thank you to the Frances Velay Summer Scholarship board of trustees, from whom I was very fortunate to receive funding which covered the expenses that would have prevented me from experiencing Stanford University.  Their dedication to supplying young women with the tools needed to grow and cultivate a career in STEM and research is inspiring, as is their mission to continually hold the life of Frances Velay in memory.

DSC04594

Chao, M. P., Jaiswal, S
., Weissman-Tsukamoto, R., Alizadeh, A. A., Gentles, A. J., Volkmer, J., . . . Weissman, I. L. (2010). Calreticulin Is the Dominant Pro-Phagocytic Signal on Multiple Human Cancers and Is Counterbalanced by CD47. Science Translational Medicine, 2(63). doi:10.1126/scitranslmed.3001375

Chao, M. P., Majeti, R., & Weissman, I. L. (2011). Programmed cell removal: A new obstacle in the road to developing cancer. Nature Reviews Cancer Nat Rev Cancer, 58-67. doi:10.1038/nrc3171

Feng, M., Chen, J. Y., Weissman-Tsukamoto, R., Volkmer, J., Ho, P. Y., Mckenna, K. M., . .  Weissman, I. L. (2015). Macrophages eat cancer cells using their own calreticulin as a guide: Roles of TLR and Btk. Proceedings of the National Academy of Sciences Proc Natl Acad Sci USA, 112(7), 2145-2150. doi:10.1073/pnas.1424907112

Lagasse, E. (1994). Bcl-2 inhibits apoptosis of neutrophils but not their engulfment bymacrophages. Journal of Experimental Medicine, 179(3), 1047-1052. doi:10.1084/jem.179.3.1047

 

 

 

Transcriptional Profiling of CASK

I spent this summer in San Diego because I went to Chicago in the fall. I attended the Society for Neuroscience Conference with Laura Been in the Psychology department to present my previous research and to talk with other presenters. To give you a sense of scale, the city nearest to where I grew up is Wheeling, West Virginia with a population of 28,000. This conference had some 29,00 people, and I was a little intimidated. To make the conference easier to navigate, I emailed some labs ahead of time to talk to at the meeting. One of the scientists I met with was Francesca Telese of the Rosenfeld lab. The Rosenfeld lab is an epigenetics lab. Essentially, this means they study how gene expression is regulated in cells and what this means for cell and organism function. She had recently developed a mouse model for CASK, a protein implicated in severe intellectual disabilities, and invited me to work with her over the summer.

I’ve been asked not to describe some of the details of my project, so instead I’m going to describe the methods we used to create a story of how CASK functions. 

Most of the known functions of CASK take place at the synapses on the cell membrane. However, CASK is also known to be translocated to the nucleus. Post-translational modifications, small peptides added to a protein’s structure, could signal CASK to change locations in the cell. To model if this is how CASK is translocated we used cell culture. TransfectingT293 cells, an immortalized cell line from the human kidney, makes these cells over expressed CASK and other proteins of interest in addition to any endogenous expression. More specifically, transfection involves getting the cell to take up a plasmid of circularized DNA containing the gene of interest and expressing it as if the foreign DNA were part of the cell’s own genome. It may seem strange to use treated human kidney cells to study to a protein important for neuronal function, but many human cells have a similar biological context for facilitating protein-protein interaction. To detect if CASK was modified we first used antibodies to isolate CASK from the cells and then a second antibody to see if the modification was present on CASK. This technique is called Western Blotting. I used computer models of CASK and modification prediction software to support our tentative results from cell culture.

Transfection Control

We also transfected T293 cells with GFP to check for overexpression before beginning the Western Blots.

The T293 cells were also made to overexpress proteins that might explain how CASK interacts with genome once it’s in the nucleus. The structure of CASK isn’t suited for DNA binding, and so we think another protein that can bind DNA is also bound to CASK to facilitate CASK’s influence on gene expression. To test this, we first have to identify candidate proteins that could facilitate CASK binding DNA using tissue culture, similar to the process described above. Then we can confirm this using brain tissue samples for ChIP. ChIP stands for chromatin immunoprecipitation. Essentially this technique could identify what parts of the genome CASK and a potential facilitator are bound to. If CASK and the candidate protein bind to the same places on the genome, it’s more likely that our candidate is mediating CASK’s interaction.

Metal Beads with Antibodies U

Part of ChIP is isolating a protein-chromatin complex. This protein-chromatin complex is bound to very small magnetic beads, which can be easily washed to remove unspecific binding.

Knowing what parts of the genome CASK interacts with is useful but doesn’t necessarily mean it changes gene expression. To look for these changes, we used protein extracted from the CASK knock-out mice brains to test whether levels of other proteins change with CASK expression. However, this technique requires candidate proteins to be selected for testing, so we also isolated RNA, a precursor to protein. In addition to testing one gene RNA at a time, the RNA extracted was sent for RNA-seq, which generates a genome-wide view of RNA expression changes due to CASK knock-out. RNA-seq uses deep-sequencing to quantifying how much of gene RNA is present in an tissue sample.

RNA Extraction UCSD

A crucial step in isolating RNA is separating the sample into organic(pink) and aqueous(clear) layers using a differential extraction. The water soluble RNA rises above the organic impurities.

Describing how a protein translocates to the nucleus, how it interacts with other proteins and changes the expression of the genome is ambitious and by no means completed in 10 weeks. Still, seeing bits and pieces of how the story is put together was a valuable experience. 

SFGs Developing Quiescent Properties

Over the summer, I worked with Desika Narayanan, the head of Haverford’s Astronomy Department, and a fellow peer, Jarren Jennings, in learning about how early, active galaxies in the universe may have evolved into currently non-active galaxies. We tackle this problem by using model and simulations we generate from the programming language Python.

Many galaxies in the early universe were once very active at producing stars. These types of galaxies are called star-forming galaxies (SFGs). However, now in their place we see inactive, quiescent galaxies. Although the quiescent galaxies now take the place of the SFGs, we cannot make the assumption that the SFGs became these quiescent galaxies. We are not sure what caused the large disappearance of SFGs, and as such, astronomers have to develop tools in order to identify how SFGs evolve over time and may have become these quiescent galaxies. The most common approach is to make computer simulations of galaxies that have similar properties to SFGs and have them all run in the models. The simulation can be observed at different points in time to see how the properties of the model galaxies have changed as it ages.

Galaxies emit light in different wavelengths. When you make a plot of the intensity of the light compared to its wavelength, there is a peak in the optical wavelength and a second peak in the longer wavelengths. This is because the gas and dust in galaxies absorbs the optical light from the stars and then re-emits them at longer wavelengths. If one wants to fully observe how a galaxy appears, they would need to look at that galaxy’s shape from multiple wavelengths. By taking multiple snapshots of a SFG simulation at a particular wavelength one can chart how the galaxy evolves over time.

Doing this for hundreds of galaxies over different time scales is not feasibly for an individual, which is why computers are necessary for handling these tasks. However, computers do not automatically know what type of object is in an image. A computer is not as skilled at distinguishing objects in a photo like our human eyes are capable of doing. As such, I have to program the computer for it to know what an object looks like by having it go through certain procedures when it’s looking at a photo.

A computer script has the computer measure the brightness of each pixel in an object. If there is a spike in brightness from one pixel to another, that is an indication that the brighter pixel is an object. This is because the background of images in astronomy is empty space. So, almost any bright pixel in astronomical photos is likely an object. If there are a large number of bright pixels clustered together, then that must mean those clusters are objects in an image. Since our research is specifically looking at a galaxy simulation, then all objects must be galaxies. The computer then creates contours around all the clusters of pixels in order to outline their shape. By plotting the image as a contour, I can have a way to detect the object in the image by changing the contour strength. For example, if I want to have a loose definition of what defines a galaxy, I can lower the contour strength to include more gas clouds. If I want to say a galaxy is strictly its bright center, then I can increase the contour strength. Lastly, the computer then finds the area of those contours by finding the contour’s vertices coordinates and, using Green’s Theorem from calculus, finds the area of the galaxy.

plot_164_085con_164

If I were to plot the change in area over time, then I could make some assumptions about what happened just from the graph. If I see that the area says constant for the most part, but temporarily shrinks in size and then reverts to its original area that is an indication that the galaxy may have temporarily separated/dispersed and then recombined. If the area were to greatly increase in size and remains that increased size, then that probably indicates another galaxy collided with the original one and merged with it. Combining both of these ideas together, I could see from the plot when the galaxy is primarily intact, collides with another galaxy, disperses in size, and then recombines. As shown below, to the left is a GIF of a galaxy as it evolves over time in the .44 micron wavelength, while to the right is a graph of the object’s snapshot number (corresponds to time evolution of the galaxy) plotted against its area:

viewd.44.1 axis

For the SFGs our group is modeling, when you observe the area of as a function of time, the area teds to stay constant at lower wavelengths. However, in the higher wavelengths, which are what the galaxies primarily re-emit light in, the area increases over time up to a point where they start shrinking and decreasing in size. This is pictured below with all the different wavelengths layered into one graph:

at

What this means is that in longer wavelengths, the SFG starts to develop some of the properties of a quiescent galaxy. The theory is that as there is less star formation in an SFG, the galaxy should begin to shrink over time until it becomes fully quiescent. However, this SFG is only doing this for some small portion of time. A future research project should continue to observe how the SFGs in the simulation evolve at timescales closer to our present epoch to see if they do become fully quiescent.

Soil Bacteria in the BiCo

This summer, I had the opportunity to perform research as part of a BiCo collaboration. My time last year in the lab of Professor Lou Charkoudian gave me the opportunity to work with her collaborator, Dr. Monica Chander, at Bryn Mawr. The Chander lab studies the soil bacterium Streptomyces coelicolor, mostly investigating how it synthesizes the antibiotic Actinhorodin.

Coming from Lou’s lab, I thought my summer would be spent running analyses on the metabolites produced by S. coelicolor, as I had been trained for in the spring. But I soon learned my first lesson of the summer: performing scientific research often means doing whatever is needed to get quality data. The Chander lab has a lot of different and exciting projects going on, and this summer I got to assist on all of them.

The least expected element of my work was, perhaps, what came from our microscopy experiments. This was a new line of experimentation in the lab, so Dr. Michelle Kanther of Bryn Mawr kindly helped us optimize our methods. That was the second thing I learned this summer: much of research is spent developing experimental methods. For our microscopy project, I accumulated pages of scribbled notes while hunched over a computer in the dark room where the imaging took place. But our efforts paid off. We made protocols for preparing bacterial cells and taking pictures quickly and efficiently.

Using our new protocols, we took images of the live (green) and dead (red) regions of our bacterial colonies.

Using our new protocols, we took images of the live (green) and dead (red) regions of our bacterial colonies.

I still did perform my expected chemical analyses this summer. But here, too, unexpected problems in our experiments gave me new insights into research.  It essentially boils down to ‘the devil is in the details’.  Many of our experiments failed this summer, particularly early on. The Streptomyces wouldn’t grow, or would become contaminated, or we would get a result completely contrary to what we had seen before.  As we kept going, I learned how crucial careful organization and preparation are for an experiment. With a good, solid plan to go from, it becomes easier to tackle the unexpected complications that inevitably arise when you explore a completely new area.

At the end of the day, my collaborative work at Haverford and Bryn Mawr proved to be a fulfilling research experience, and all of its unexpected lessons showed me how much I enjoy science research. I have always thought I would want to spend my life working at the lab bench, and this summer intensified that dream and brought it closer to reality.