Undergraduate Research

I successfully defended my Honors Senior Research Thesis! Check out my thesis here: Deciphering the mechanisms of dedifferentiated liposarcoma resistance to targeted therapy.

Project Title: DEVELOPING AND CHARACTERIZING MODELS OF RESISTANCE TO CYCLIN-DEPENDENT-KINASE 4 (CDK4) INHIBITION IN THE CDK4 AMPLIFIED CANCER, DEDIFFERENTIATED LIPOSARCOMA

Project Mentor: James Chen – Internal Medicine

ABSTRACT

Introduction/Background:

Dedifferentiated liposarcomas (DDLPS) are malignant adipocytic cancers characterized by an amplification of the cell cycle regulatory gene, CDK4. Palbociclib, a CDK4/6-inhibitor, is rapidly becoming a standard of care treatment; however, innate and acquired resistance to the drug limits efficacy in patients with DDLPS. Molecular mechanisms and biomarkers of palbociclib resistance have been posited in non-mesenchymal tumors, but are poorly understood in DDLPS. To this end, we developed palbociclib-resistant DDLPS cell lines derived from several patient DDLPS samples to better characterize these phenotypic and genomic differences.

Methods:

Three human CDK4 amplified DDLPS cell lines (224A, 246, 863) and one liposarcoma control cell line without CDK4 amplification and RB loss (LiSa2) were brought into culture. Developing resistant cells were intermittently treated at palbociclib’s IC₂₅ and surviving cells were allowed to recover before retreatment. Parental lines were grown as controls in tandem with the developing resistant lines to compare molecular changes in response to palbociclib. Cell viability was measured by an XTT assay and protein expression was measured using whole cell lysates for Western blotting.

Results:

Parental IC₅₀ of the three CDK4 amplified DDLPS cell lines ranged from 8-12 µM and the cell line without CDK4 amplification had an IC₅₀ of 12 µM. Resistant cell strains demonstrated minimal changes in the measured IC50. Initial response to palbociclib decreased expression of cell cycle proteins, CDK4 and pRb, and increased Cyclin E, CDK2, and CDK6. After 6 months of treatment, a statistically significant decrease in Cyclin E, CDK2, and pRb, was observed in the resistant cell lines and no changes were seen in CDK4, Cyclin D, and p16. (all p<0.01).

Conclusion:

Here we present the development of novel, palbociclib resistant, DDLPS cell lines. While the initial response to combat palbociclib treatment increased Cyclin E and CDK2, these G1/S proteins are reduced in resistant lines. Resistant cell lines demonstrated decreased protein expression of Cyclin E, CDK2, and pRb. Clinically, DDLPS patients that exhibit innate dysregulation of RB1 or G1/S phase regulators may be more likely to demonstrate resistance to palbociclib. Next-generation sequencing of DNA and RNA expression in parent and resistant lines are underway.

This summer, 24 CSPC/PS-ON fellows and I learned how to use R-studio to analyze a gene data set to understand breast cancer metastasis. At the end of our 10-week internship, we participated in a hackathon-style project. My partner and I had three hours to come up with a solution to enzalutamide-resistant prostate cancer.

The following infographic covers the basics of our proposal. This project was an opportunity to creatively think of an approach to treat drug resistance, consider the limitations of different methods, and develop scientific communication skills, not to come up with solid answers.

Click the image to see it move!

INTRODUCTION

Pancreatic cancer remains one of the deadliest solid malignancies, with a 5-year survival rate of approximately 10% in the USA. Patients often do not show symptoms until cancer reaches an advanced stage where cancer can aggressively metastasize (travel to different parts of the body)–the mechanism of which remains unknown. Venous invasion occurs when cancer cells gain access to the veins and is a promising explanation for the development of liver metastases and thus the rapid progression of disease in most patients.

BACKGROUND

The pancreas is located behind the stomach and underneath the liver. It aids in digestion and blood sugar regulation. Pancreatic cancer develops in stages including the formation of precancerous pancreatic intraepithelial neoplasia (PanIN) which forms as a result of genetic mutations impacting cell-cycle progression, cell division, and cell growth in ductal cells. These mutated cells can multiply and travel through the ductal system until it potentially develops into cancerous pancreatic ductal adenocarcinoma (PDAC).

AIM

This project consists of further developing a new imaging method of tumors in 3D at single-cell resolution named CODA–to be validated and tested by studying venous invasion in pancreatic cancer. CODA is based on novel registration and deep learning algorithms. 

Why is 3D imaging important to study venous invasion? As of now, we’ve only been able to look at veins in 2D cross-sections. 

• A: If my vein is horizontal, and I cut the biopsy horizontally, then I will see the whole vein. Great!
• B: What if the vein is vertical? If I cut the biopsy horizontally, then I will see a dot–just a section of the vein
• C: Okay, let’s say I can see a dot. Well that means 1 dot == 1 vein. So 2 dots == 2 veins, right? Actually, veins can branch off. So 2 dots could mean we have 1 vein with a branching point
• Conclusion: you cannot determine vein position or branching without looking at the whole thing in 3D. As of now, we’ve only been able to see veins in 2D, so reconstructing full tissues after they’ve been sliced up is really neat

METHODS

1. We (not me, but someone else) took a biopsy of a patient’s pancreas and sliced it into multiple horizontal sections every 4 μm. Then they stained it so different pancreas cells are stained different shades of white, pink, blue, and red so it is easier to identify them
2. Manually annotate tissue components of pancreas tumors
   1. I spent my first three weeks circling different components of pancreatic tissue like fat, blood vessels, and ducts to help the computer identify patterns and manually determine what type of cells the tumor image had. “Hey, Computer, these white circles are fat. So if you see one of those white circles, identify that as fat”
3. Determine % composition of each component in normal and cancer samples
   1. Once the computer identifies the components in the whole image, I want to know how much of the pancreas sample is made of cancer cells or other components
   2. In the heatmap below, I’ve analyzed normal pancreas (no cancer), PanIN (precancerous), and PDAC (pancreas sample with cancer). We can see that as the cancer develops, acini decreases and collagen increases
   3. The pancreas is mostly acini. Those are the cells that produce enzymes for digestion. So the cancer may actually kill those important cells… not good!
   4. Collagen is like the support structure that mainly surrounds blood vessels (and your muscles/skin/bones/tendons). Collagen tends to increase as cancer develops and we believe that stronger, denser collagen provides a shield to protect cancer and allow it to grow without immune cells attacking it
4. Determine how components may change throughout the tissue
   1. Let’s just confirm that collagen increases where cancer develops. Because these numbers are great, but what if collagen was increasing in a corner of the pancreas and had nothing to do with developing cancer.
   2. This graph plots the % composition of each slice. We can see that as PDAC increases (orange) (take my word, PDAC develops near the 3-4 z-plane, it’s just very minuscule in comparison to the rest of the graph), collagen (pink) also increases and acini (purple) decreases
   3. 
5. Determine how the density of tissue components change
   1. From Figure 3.5 (blue heatmap), we can see that acini decreases and collagen increases as PDAC develops. But how do we know that these changes are directly related to the formation of PDAC? If we see a large increase in collagen surrounding PDAC and a decrease in acini around PDAC, we can conclude that PDAC is the reason for acini and collagen changes.
   2. To do this, we can calculate the density of tissue components as a function of distance from PDAC. Density of Collagen = #pixels of Collagen within a certain distance from PDAC / # total pixels within a certain distance from PDAC
   3. In the graph below, the x-axis is the distance from PDAC and the y-axis is the density. When we’re close to PDAC (x=0), collagen is high and acini are low which matches our previous observations. As PDAC develops, collagen increases, and acini decreases.
6. Regenerate 3D video of tissue from multiple 2D slides
   1. This is nice and all, but let’s see the cool stuff. The blue are ducts that are normal in the pancreas, the red are blood vessels, and the yellow is the cancer
   2. Click the image!
7. Locate venous invasion
   1. In another tumor sample, we found venous invasion
   2. When I bring up venous invasion, most people think the cancer will make its way into the vein and travel to another place. BUT the cancer will actually get comfy, enter the vein, and stick to the interior, replicate, and take up the ENTIRE vein. Therefore, cutting off circulation through that vein. It can still travel through the vein and exit to another place
   3. The blue is cancer and red is the vein. The cancer completely occludes the vein
   4. Click the image!
8. Calculate blood vessel and PDAC length 
   1. Using MATLAB, I was able to calculate lengths of the blood vessel. Notice how the vein branches off instead of remaining a singular tube. We can calculate the length of each of the extensions. If I can calculate the length of the blood vessel and the length of the cancer (PDAC), then I can see if PDAC extends throughout the entire vein. In MATLAB, this figure is measured in pixels, so I included both pixels and mm lengths.
9. Calculate PDAC occlusion
   1. Now we know that PDAC can travel a significant length inside the vein, does PDAC take up the entire space inside? or does it only touch the inside vein walls and leave an open tube through the center? The amount of space PDAC takes up inside of the vein is defined as occlusion.
   2. To determine occlusion, I can use a single slice of the total volume, calculate the area of the PDAC and then calculate the area of the PDAC and the open space inside the PDAC (this is called lumen). For example, if PDAC looked like a straw inside the vein, its area would only be the outer area of the straw and the lumen would be the open space. That is a low occlusion. BUT if PDAC looked like a wooden dowel inside the vein, its area would be the entire circle, and there would be no lumen. The occlusion would be 1.
   3. Once I calculate the occlusion for each z-plane of the 3D volume, I calculated the average occlusion to be 0.83

FUTURE WORK

Although it is evident that pancreatic cancer can travel through the vascular system, determining factors that inhibit or promote movement will aid in preventing metastasis. Early-stage surgically resected pancreatic cancers often lead to metastasis, but if we can prevent metastasis by changing the environment surrounding cancer and its means of travel, early intervention can be a promising treatment. These findings can also help to stop developing metastasis and attack cancer at locations such as the bloodstream if that is its main source of travel. Concern also rises over the method of travel for noninvasive PanIN through the vascular system and how that can lead to the implementation of several precancerous spots in various organs prior to PDAC formation.