Genomic Instability Analysis in DNA from Pap Test (2024) PDF

Pathology and diagnostics Prof. Sergio Marchini - Clinical pathology - Lesson 5 Genomic instability analysis in DNA from Papanicolau test 27/11/2024 - Group 8 (Giacomo Cecchin & Davide Beatrici) Nowadays, the focus has shifted to the idea that we could...

Pathology and diagnostics Prof. Sergio Marchini - Clinical pathology - Lesson 5 Genomic instability analysis in DNA from Papanicolau test 27/11/2024 - Group 8 (Giacomo Cecchin & Davide Beatrici) Nowadays, the focus has shifted to the idea that we could improve cancer treatment by enhancing our ability to perform early diagnoses. High-grade ovarian cancer is a very difficult disease to treat. You probably know very well the experience of the famous italian model chiara balti who suffers from ovarian cancer. This is an example of a high risk patient as she presented BRCA1 BRCA2 mutations and we failed to diagnose the tumor in the early stage. There are some international clinical trials where we tried using CA125, a serum biomarker, or imaging techniques in low-risk populations to achieve early diagnoses. The success rate of intercepting early symptoms was around 50%. We are currently not progressing well. Even in high-risk populations, where we know mutations in the BRCA1 or BRCA2 genes exist, prophylactic surgery is performed, but tumors can still develop. This brings me to the first point I want you to remember: ovarian cancer is a misleading term. When thinking about ovarian cancer, remember this image above. Though it is colorful and simplified, it describes the complexity of the disease. Ovarian cancer is a heterogeneous disease, comprising many different tumor cells. This heterogeneity is its hallmark. The term "ovarian cancer" is misleading because, even after removing ovaries and tubes, tumors can still develop elsewhere, as in the case of Bianca. For example, we know that for high-grade serous ovarian cancer, the site of origin is not on the ovary, but in the fimbrial ends. For mucinous tumors, the origin is probably in the GI tract. For clear cell, endometrioid, and low-grade serous tumors, the origin is likely from the endometrial tissues. This is why the recent classification from pathologists states that ovarian cancer originates from the ovary, the fimbria, and the peritoneum. In the case of this model, Bianca Balti, the probable site of 1 origin was the peritoneum. There are many different sites of origin, which is why biologists say "ovarian cancer" is a misleading term. This brings up the concept that it's quite difficult to perform an early diagnosis when we don't know the exact site of origin of the disease. The site of origin reflects the tumor behavior. High-grade serous tumors grow very quickly, while low-grade and mucinous tumors grow more slowly. The prognosis varies significantly depending on the stage of the disease. The stages of the disease—stage one, two, three, and four—are surgical parameters. This means that diagnosis is often made during primary surgery, not before. The tumor stage reflects the extent of the disease at the time of diagnosis. When the disease is diagnosed at stages 3 and 4, it means that the tumor mass is beyond the ovaries; it is in the peritoneum, the pelvis, or beyond the peritoneum. When the disease is in stage one, the prognosis is very good because the disease is confined to one or both ovaries. 2 We have clinical evidence showing that when we are able to detect this disease early, when it is limited to one or both ovaries, it is curable. Without metastasis, the disease is curable. The current treatment methods are highly effective. Surgery is curative for most women diagnosed at stage one. This means that after 10 years from the early diagnosis, these women are still alive. However, when the disease is diagnosed with multiple metastases in the abdominal cavity, the overall survival ranges from one to three years, with relapse occurring about 18 months after the end of chemotherapy. This is clear clinical evidence that if we could shift our ability to detect the early symptoms of the disease, we would be able to improve overall survival. But this is not as simple as it sounds, because the problem is that most of the symptoms women experience during the progression of the disease are nonspecific. There are many different symptoms that do not clearly indicate the presence of a growing disease in the ovaries. This tumor is often referred to as a silent killer because, in the vast majority of women, the disease is diagnosed with multiple synchronous lesions in the peritoneum when it is too late for effective treatment. From a molecular point of view, rather than a clinical one, it is important to remember that when you see many different lesions at the same time, they do not share the same genomic landscape. Most of them are completely different. This is extremely important from a therapeutic point of view. Because it is now clear in the oncology field of chemotherapy, the analysis of a single sample is not representative of the molecular portrait of the disease. If you are analyzing a single tumor biopsy from the ovary, it can be completely different in terms of driving mutations and actionable regions from other samples left in the abdominal cavity. This means you are treating a disease that is different from the one you have analyzed. It is now common knowledge that a single solid biopsy is not representative of the entire genomic landscape of the disease. This is why our current treatments are ineffective in improving survival. However, for ovarian cancer, for example, when the disease is limited to one or two ovaries, you are analyzing the same disease that you are treating, or what is left by the surgery in the abdominal cavity. This is extremely important. I want to emphasize the idea that if we could intercept the early symptoms of the disease, we would probably be able to improve patient survival. The question is, if we look at one of the most well-known patient associations in the US, the Ovarian Cancer Alliance, we see a flyer from this association. It highlights that there is no 3 early detection method. The Pap test will not detect ovarian cancer. It's not a suitable approach for ovarian cancer. We challenge this idea. Here are the data generated here at our research center, in collaboration with many clinicians in Italy, demonstrating a new route for early ovarian cancer diagnosis. These are two Kaplan-Meier curves based on two large international clinical trials with more than 10 years of follow-up. They clearly demonstrate that the currently available clinical methods for early diagnosis, which include serum analysis and imaging techniques, alone or in combination, are not effective. One study is UK-based, and the other is US-based. We see how many women were analyzed, and from the two curves, we see that they are almost identical. This means that by chance alone, we are not able to achieve early diagnosis. So, how can we deal with the silent killer nature of ovarian cancer? The question is how can we fulfill these requests from the clinic? How can translational research help clinicians address this unmet clinical need? To think of a new way, a new route for early diagnosis, we should consider the progress made in ovarian cancer research over the last few years. The first point, as discussed previously, was understanding the origin of ovarian cancer, which is the most common and significant isoform. The site of origin has now been well demonstrated to be in the fimbrial ends. It means that at the fimbrial ends, there are modifications in the cell composition. The first step is not shown in this cartoon because it was recently published in a paper a few months ago. There are many different epithelial cells, including ciliated and secretory ones, as well as some transition cells. The most 4 common cells are the secretory ones. Secretory cells, under the influence of certain transcription factors, can modify their genomes or transcriptomes and transition from secretory to ciliated cells. There is a transition phase where some transcription factors change over time, and these two cell populations interconvert within the fimbria. For reasons not completely known, in particular areas of these ends, secretory cells begin to overgrow. This thin layer of secretory cells is called the "scalp." The feature of these secretory cells is that they are largely unable to repair DNA damage, which is important. Anatomical and pathological analyses have shown that there are strings of secretory cells widespread in the fimbria, and from a molecular point of view, these cells are unable to repair DNA. The next step in tumor progression is the acquisition of a mutation in a driver gene, specifically a single point mutation in the p53 gene, which is the most commonly mutated gene in solid tumors. P53 is mutated in 99% of patients with high-grade serous ovarian cancer. All tumor cells diagnosed or relapsed at the end of life harbor the same point mutation in the p53 gene, at the same position. However, each patient has their own unique point mutation, resulting in a heterogeneous array of hotspot mutations at the molecular level. The acquisition of this mutation clearly changes the morphology of the cells, which can be analyzed using hematoxylin and eosin staining. This change in cell morphology is called p53 signature. These cells are genomically stable but harbor a mutation in the p53 gene and are unable to repair DNA. The acquisition of this somatic, not germline, mutation in the p53 gene begins to change the behavior of the cells, but they remain normal cells. It has been calculated that in the fimbria of a healthy woman, starting from age 20, there is an acquisition of a single p53 signature every 10 years. These mutations are widespread in the fimbria. The next step is that these cells, largely unable to repair DNA damage, start to acquire additional damages. The reasons for this are unknown and remain open questions awaiting answers. We also do not know the role of the surrounding microenvironment in losing control of these mutated cells. 5 Why? At some point in life, a single p53 signature becomes an in situ lesion, which is generally known as STIC (serous tubal intraepithelial carcinoma). STIC is the early molecular alteration that drives the progression to invasive carcinoma. As I mentioned earlier, high-grade serous ovarian cancer is a misleading term, and the site of origin is in the fimbria. Pathologically, what is generally recognized from a morphological point of view is the STIC lesion. Mutated cells in STIC are characterized by morphological changes in the nuclear-cytoplasmic ratio and their genomic activity, with the acquisition of gains and losses of various genomic materials. They are highly invasive. These in situ carcinoma cells eventually start to exfoliate and move, and when they do, they drop onto the ovarian surface, where they begin to grow into metastatic invasive carcinoma on the ovary or peritoneum. From a mathematical point of view, it has been calculated that the progression from a p53 signature to an invasive carcinoma takes about five to six years. This underscores the importance of early diagnosis. The question is, do we have time to perform early diagnosis? Yes, because if we can intercept the STIC, we would have six years to work on prevention. This is why women at high risk of developing ovarian cancer, such as those with BRCA1 and BRCA2 mutations, are often recommended to undergo prophylactic surgery around age 40 for BRCA1 or 35 for BRCA2 to remove the fimbria before any evidence of a STIC lesion appears. For high-risk women, the current method to achieve early diagnosis is prophylactic surgery, which aims to remove any potential STIC in the fimbria. In the low-risk population, this remains a question mark and is the focus of my talk today. The second pillar of the new route for early diagnosis is our ability to identify most of the genomic alterations that occur in STIC lesions. When the gatekeeper gene p53 is lost, there is a step-by-step acquisition of genomic imbalances that are not repaired by the cells. These are not random events. We now understand that the initial loss of genomic material is followed by whole genome duplication, leading to gains in specific chromosomal regions. All the point mutations you know or have studied as actionable lesions used for targeted therapy in patients are late events in ovarian cancer. Unlike other diseases with mutations in genes like Kras or Braf, ovarian cancer does not have early actionable lesions. In solid tumors characterized by p53 mutations, the early genomic events are not stochastic but occur in a deterministic manner, involving loss, whole genome doubling, and gain of genomic material. This means that at the molecular level, STIC or any other early genomic lesion is characterized by what we know as genomic bands across the chromosomes. We gain and lose genetic material in many different areas, resulting in complete alterations in genome ploidy and epigenetics. The third pillar of our hypothesis is the evidence of an anatomical corridor that connects the fimbria with the endocervix. This concept, referred to 6 as the "anatomical corridor," was observed by pathologists many years ago. Under physiological conditions, normal cells from the fimbria move backward into the endocervix area, where they can be collected by a Pap test. This is a normal cellular movement. However, we began to question whether pathological cells from STIC lesions, in addition to moving forward onto the ovarian surface, could also move backward into the endocervical canal and be collected there. Starting around 2015, several papers were published on this topic. It's important to note that NGS (Next Generation Sequencing) technology, which is now commonly used, was just beginning to be developed at that time, less than ten years ago. Consequently, high-coverage and high-sensitivity analyses were not yet possible. Early studies demonstrated that it was indeed possible to intercept pathogenic p53 mutations in Pap test smears, or more precisely, in the DNA purified from these smears. The focus on p53 is due to its role as an early pathogenic lesion present in all tumor cells. Studies conducted by Johns Hopkins University and other international centers focused on the time of diagnosis when there was already evidence of disease. They performed Pap tests during surgery and found that a year before surgery, when the disease was present, they could detect the pathogenic lesion in the Pap test. Why p53? Because it is the early pathogenic lesion present in all tumor cells. These studies, conducted by Johns Hopkins University and many international centers, focused on the time of diagnosis when there was evidence of disease. During surgery, they performed Pap test smears. Before surgery, the disease was present, allowing them to intercept the pathogenic lesion in the tumor and trace it back in the Pap test DNA to check for that mutation. However, when a tumor is diagnosed at a late stage, the anatomy is completely altered by the tumor mass, and the channel collapses. This means that the colleagues suggested that the Pap test is not a reliable source for analysis, as the detection rate was only about 30-35%, which is not acceptable for developing an early diagnostic test. In the US, where women frequently move and change hospitals and clinicians, it is challenging to obtain longitudinal samples from the same women. However, in Italy, our lifestyle often involves staying in the same town or moving to the countryside while maintaining long-term connections with the same clinicians. This stability led us to reason that if we could collect longitudinal samples from the same woman at different points in her history, we might increase the detection rate because, at those times, when there was no evidence of disease, the anatomical corridor was still functional. This project fascinated many clinicians across Italy, allowing us to collect FFPE tumor samples at the time of diagnosis and various Pap tests at different time points from the same woman. Please note that these Pap tests had been stored at room temperature for 5-10 years, making the material challenging to 7 work with. Despite this, we used two independent cohorts with more than 130 patients. The workflow is summarized in this cartoon. For each woman, we used NGS to identify the pathogenic p53 mutation in the tumor sample. Remember, p53 mutations are not common to all women and are highly heterogeneous across different tumors, so we needed to pinpoint the exact mutation in the tumor. We then employed digital droplet PCR, a high-performance PCR approach. If you're unfamiliar with this technique, imagine a 20-microliter buffer solution split into at least 20,000 droplets. Each droplet acts as a tiny tube, allowing for simultaneous PCR amplification and reading, thereby increasing sensitivity to detect mutations in a large amount of normal DNA with very few pathogenic molecules. This technique is widely used in hematology to track minimal residual disease and in infectious diseases to detect viruses like HIV in various human fluids. Using this approach, we performed experiments on the two cohorts. For each cohort, we split one tube into 20,000 droplets. This distribution of droplets allows us to identify clonal point mutations in the gene from tumor DNA, enabling us to trace the core mutation from the tumor sample. 8 These are the normal tissues, alright? This is the DNA from a healthy woman in which there were no mutant codes. This is a very specific and sensitive approach. Please take a look at this table. It might seem boring, but it is informative. Here we have the patient IDs and the mutations we identified in the tumor mass. As I mentioned before, each woman has her own p53 mutation in different sites. We collected Pap tests at different time points. Since this was not a clinical trial, we created temporal windows to ensure the Pap tests were comparable. In bold, you can see the analytic measure, which indicates how many molecules of the mutated p53 gene are present in the solution. For example, 49 months before the early diagnosis of ovarian cancer, we detected the same pathological mutation in the p53 genes as we found in the tumor. This is significant because it provides clear evidence that the anatomical corridor is a way to monitor what is happening in the fimbria. We can use molecular approaches to infer the molecular portrait of the fimbria. As we moved backwards in time during the follow-up, the detection rate increased from 35% reported in the literature to 65%. This significantly improved our ability to intercept early molecular events. Here is the first take-home message. Initially, we challenged the common opinion that the Pap test smear is not suitable for early diagnosis. Our approach suggested otherwise. However, we encountered the problem that p53 is not a good biomarker for diagnosis. Currently, we know that healthy tissue can harbor point mutations in many actionable genes, including p53, EGFR, KRAS, and BRAF. Healthy tissue is a mosaic of cells harboring mutations. Additionally, abnormal hematopoiesis can release DNA of abnormal cells into blood samples or tissues, making p53 a confounding factor. All the studies we performed were retrospective. We analyzed the tumor, selected the gene, and then checked with another technique. But prospectively, dealing with many different point mutations spread across the tumor gene is challenging. We must ensure we are identifying the correct pathogenic mutation, not a physiological mutation. Therefore, we need a different biomarker. While the path we followed was correct, we need a new biomarker to improve our approach. You can use a very low coverage approach to intercept these abnormalities. Normal DNA does not tolerate any gain or loss. So you have two completely different configurations: “mountain” and “landscape”. Therefore you can distinguish between normal and mutated DNA due to the discrepancy in their genomic landscape. 9 In the image above, the background you don’t see here represents the normal state. So what you see are the imbalances compared to this normal state, which is represented by a flat baseline. They are easy to identify, and this is why I suggest using a comparison between landscapes and mountains. Specifically, this type of abnormality is common across all lesions in a woman with high-grade serous ovarian cancer. What does this mean? It suggests that these abnormalities, when shared across synchronous lesions, are early events in tumor progression, as they represent the root causes of tumor development. This is important for us because we are searching for an early molecular portrait that can reveal these early changes. In fact, when we analyze the molecular profile of a STIC lesion, you may be seeing this type of analysis for the first time. This is a fibrated analysis using specific stains, showing a stretch of cells with high staining. These cells are morphologically distinct from the surrounding tissue and are considered a 'stick' or focal area. This is a single stretch of cells that are featured by p53 mutation and genomic imbalance. This morphological feature can be distinguished through genomic sequencing, revealing a genomic pattern similar to what we identified previously. So, we now have the anatomical, the biomarkers, and the genomic data to help us better understand the lesion. We then hypothesized that the Pap test might not be a suitable tool for early diagnosis of ovarian cancer. To test this hypothesis, we conducted a study in Italy, collaborating with multiple clinical centers. We collected Pap tests from 62 women who were considered healthy at different time points in their lives, with no evidence of disease. Please recall that ovarian cancer is often referred to as a "silent killer" and typically takes around six years to progress into invasive carcinoma. We collected these Pap tests when the women were healthy, and also collected formalin-fixed paraffin-embedded (FFPE) tissue samples at the time of diagnosis. This provided us with a clinical timeline, where we had both healthy samples from women with no disease for the past 5–6 years, and samples from when they were diagnosed with ovarian cancer. At the same time we collected pap tests from 77 women with no evidence of disease in the last 5-6 years. These women formed our control group, and this is the scope of our study. From this collection of 62 women, we selected Pap tests and FFPE samples from different 10 time points, prior to the diagnosis. This allowed us to perform the analysis, which led to the development of the EVA test (Early Ovarian Cancer Test), designed to assess the risk of developing ovarian cancer. The study used low-pass whole-genome sequencing, a technology I’ve already discussed. I won't go into too much detail, but please refer to this diagram for an overview. In this cartoon, you can see the normal genomic coverage used for our assay 200x. The difference in coverage across specific genomic regions is significant. From a bioinformatics perspective, we applied algorithms available online that are now commonly used for non-invasive prenatal testing (NIPT). The shallow whole-genome sequencing approach we are using now will, in the near future, be integrated into NIPT assays. Currently, most of these tests are performed using microarray technology, which is outdated, expensive, and labor-intensive. Nowadays, our guideline is to introduce, or better yet, to change in the near future, the array technology for NIFTY analysis with low-pass whole genome sequencing. The algorithms we used were those that are commercially available for this approach. This is a really interesting finding because it represents the first evidence from a single woman in which we analyzed the PAP test and the FFP profile using low-pass sequencing. So, here is the FFP, and you can easily follow this cartoon: these are the chromosomes, this is the ploidy, and you have a scatter plot showing the ploidy. This is not a flat situation; you can see many regions of gain and loss. This is a common profile, or rather, a typical copy number profile of high-grade serous ovarian cancer. Now, please take a look at this PAP test. We call this a match test. The samples are from the same woman, collected seven, six, and four years before the diagnosis. In the seven-year sample, there were seven regions of interest, marked in orange or red. Are you seeing the Orange. These regions are shared between the tumor and earlier time points, like the one on chromosome 11, T11.1. This region is present in the tumor and appears across all time points, 11 which strengthens our confidence that, even seven years before the diagnosis, we could identify pathogenic regions in the PAP test DNA. Thus, the anatomical corridor is working, and our ability to detect pathogenic lesions is confirmed. But please take a look at these plots. There are many other different regions, and they are widespread. I would describe them as heterogeneous. Some regions are private to a single time point, while others are lost in later time points, or acquired and not conserved. This is a classical situation. For example, when you are training a patient cohort and looking at the liquid biopsy results of this patient, tracking tumor evolution in plasma, you see regions that are gained, acquired, lost, or private to a single time point. There are very few common regions, and most are private. This reinforces the message I emphasized earlier: heterogeneity is at the root of tumor evolution. This means that even seven, four, or six years before diagnosis, the lesion (the invasive carcinoma) is already heterogeneous at the beginning of the disease. This presents a challenge when working on a project to develop a test for early diagnosis. So, if this holds true for 53 women, how can we evaluate the copy number alterations? Somatic copy number alterations seem to be an interesting biomarker, but we face the same problem of heterogeneity. This is an idea we developed in August while we were still working on this project: we should move towards a tumor-informed approach, adopting a tumor-agnostic strategy. This means we should focus not just on individual genomic lesions, but on the overall amount of genomic lesions across the genome. The idea behind our binary partition was to focus not on the single lesion, but to analyze the entire landscape. Using this approach, we introduced the concept of CPA (Copy Number Profile Anomaly). The concept is simple yet powerful: instead of focusing on the specific sites where gains or losses occur, we count the number of regions with abnormal copy numbers. This gives us a summary number that indicates how many regions in the DNA are abnormal. In healthy DNA, there are no gains or losses, or if there are, we might miss them due to insufficient coverage. Please remember that we are working with 0.5x coverage, which is very low-pass sequencing. While there may be regions with gains or losses in fragile sites in healthy DNA, these are often missed. 12  However, these events are typically physiological. In contrast, in tumors, such abnormalities are not physiological, we have a large amount of alterations. And this is exactly what we tried to perform and finally we came out with this score. In collaboration with bioinformaticians and staticians we developed a paradigm in which we observed the distribution of the cpa value. The distribution of the cpa score in women who developed cancer is statistically different from the distribution of cpa score among women who are still healthy. And we identified three different areas. A green zone with a cpa from 0 to 0.5. A grey zone with a cpa score between 0.5 and 0.6. And finally a red zone with a cpa score greater than 0.6 in which all the pap tests of all the pre high grade women was identified. These are the performances of our test, the sensitivity and specificity is interesting. This is the first attempt to achieve early diagnosis but interestingly when we analyzed all longitudinal samples from the different cohorts we observed that the pap test successfully identified all of the women in the high risk group. As pap tests are very commonly used this could be introduced in the future as a routine clinical exam especially for the high risk women. So that ideally the treatment will be based on these results rather than exclusively on the age of the patients. As there is no evidence supporting that 40-45 years of age is the correct time for prophylactic surgery. Therefore, the objective would be to understand if we can delay the prophylactic surgery in the high risk population. This approach would be beneficial because prophylactic surgery is followed by a lot of aspecific effects such as hormone imbalances, cardiac problems, hypertension and bone demineralization which have a negative impact on the life of the patient. Therefore, the possibility of having a very rapid and cheap clinical essay able to help the clinicians and women in their decision making has a great potential to be extremely beneficial. 13