CNP practice Flashcards

Question

Are dormant cancer cells the same as cancer stem cells?

Answer 1

No. They can partially overlap but are different. CSC are a subpopulation of cancer cells with increased renewal capacity and the ability to recapitulate the heterogeneity found in primary tumors. The primary feature of the dormant cells/persisters is the strong reduction in proliferation and metabolic activity leading to drug tolerance. That being said they are not exclusive: some cancer stem cells can enter a dormant state, and some dormant cells may display stem-like features, but one does not imply the other.

Answer 2

What we can say from the data is that persister cells are not necessarily enriched for cancer stem cell features. In our CRC DTP models, we looked specifically for stemness markers via both gene signatures and IHC, and found no significant enrichment. To be clear, I cannot exclude that a small fraction of resident CSCs contributes to the eventual regrowth phase. However, our data demonstrates that the survival of the persister state itself does not require a CSC identity.

Answer 3

Breast: CD44+ / CD24 – or low CRC: CD44+ / CD133+ / CD26+ ; Lgr5+

Answer 4

-Clinical Relevance: A mastectomy does not equate to a 0% risk in situ. Roughly 5% to 10% of patients still experience a local recurrence (=in the same breast or chest wall). -Experimental Feasibility: The primary site is accessible more easily and provides a high-biomass window into the DTP state. While disseminated cells (DTCs) in the bone marrow are clinically critical, they are extremely rare and difficult to isolate. So for an ambitious, high-resolution multi-omics project, using the primary site makes sense and allows to uncover the fundamental mechanisms of survival that can still be relevant for distant metastases.

Answer 5

First, it is important to acknowledge that there is no universal clinical consensus on the exact cutoff to classify a relapse as 'dormancy-driven'. However, in clinical practice, a relapse is strongly considered a consequence of dormancy when it occurs after a Pathological Complete Response (pCR) and a Disease-Free Interval (DFI) of at least a year. But after pCR: - 10-20% of patients will experience relapse - TNBC: about 70% occurring 1-3 years after initial treatment - Luminal: early relapse almost non-existent after pCR, ~50% between 1-5 years and ~50% 5-20 years after initial treatment

Answer 6

First, it is important to acknowledge that there is no universal clinical consensus on the exact cutoff to classify a relapse as 'dormancy-driven'. However, in clinical practice, a relapse is strongly considered a consequence of dormancy when it occurs after a Pathological Complete Response (pCR) and a Disease-Free Interval (DFI) of at least a year. But after pCR: - 10-20% of patients will experience relapse - In most cases the peak of relapse will occur between 1-5 years, with some cancers (called indolent like luminal BC or prostate) reaching even 5-20 years

Answer 7

- Aguirre-Ghiso (Albert Einstein, NY) – p38/ERK ratio, signaling, focuses on DTCs / niche - Ghajar (Fred Hutch, Seattle) – DTCs and the role of the niche - Agudo (Dana Farber, Boston) -> immune evasion, quiescence in TNBC (“quiescent CSCs”) - Bentires-Alj (Basel) -> Tumor heterogeneity & DTCs - Vallot (France) – TNBC using single-cell but ChIP-seq but dormancy not main focus of the lab - Luca Magnani (UK) – hormonal BCs, epigenetics

Answer 8

No, I have access to all the equipment, either in our building or through collaborations at Bordet and the KUL

Answer 9

The novelty of my approach is its holistic approach, integrating preclinical and clinical samples with multi-omic technologies. While single-cell sequencing has been used sporadically in dormancy before, we go much further by looking at multiple layers of regulation simultaneously. This approach will allow us to map the entire regulatory network with a precision never reached before.

Answer 10

If the cells are sensitive at the start, then targeting the dormant cells eliminates the 'surviving reservoir' early on, which in turn prevents the cells from acquiring additional mutations that lead to resistance. (It does not however address potential pre-existing resistance.)

Answer 11

Currently, no, our primary activity does not rely on AI and we already have analysis pipeline in place for the various omics approaches. However, we are 'keeping the door open' for Machine Learning specifically as a tool for multi-omics integration (e.g. MOFA+ = Multi-Omics Factor Analysis).

Answer 12

DCIS = Ductal Carcinoma In Situ. When abnormal cells are confined strictly inside the milk ducts and have not invaded the surrounding breast tissue. IDC = Ivasive Ductal Carcinoma. When The cells have broken through the duct walls and invaded the surrounding stroma. (= most common type of BC). IDC is the main clinical concern because they are much more at risk for relapse and dissemination.

Answer 13

CSC are a subpopulation of cancer cells with increased renewal capacity and the ability to recapitulate the heterogeneity found in primary tumors.

Answer 14

First, transcriptomics only shows us the end result; we need the additional layers to identify the upstream drivers. For instance, transcription factor activity or regulation of intergenic elements are much better captured at the genomic level. Second, these epigenetic layers represent druggable vulnerabilities in themselves. For example, detecting specific methylation patterns could point us directly toward using agents like DAC, allowing us to target the dormant state with precision.

Answer 15

First, we know that several genes that have been involved in drug-tolerance and dormancy are regulated by promoter methylation in BC (e.g. TSC1/2, MYC). Second, demethylating agents such as 5-Aza and DAC have been shown to influence the sensitivity to some treatment (e.g. radiotherapy).

Answer 16

We know that there is extensive transcriptomic rewiring at play. Notably there is often a global hypotranscription on one hand, and on the other hand specific transcription factors like NR2F1 and b-cat/TCF. And ultimately, chromatin accessibility is the structural foundation that makes this entire transcriptomic rewiring possible.

Answer 17

* RNA: 1M reads/cell, 50bp single-end, expected ~10K genes * DNA: 10M reads/cell, 150bp paired-end. * ST : 50 K paired reads/ spot => if 2000 spots ~ 100M paired reads

Answer 18

Yes and no. We expect a core survival program, a “hub”, to be conserved across different drugs. However, different subtypes have fundamentally different genetic and epigenetic landscapes, so the upstream regulation could be context-dependent.

Answer 19

NMRI mice lack functional T cells (athymic due to a mutation in the Foxn1 gene), the other component are present but unable to mount a coordinated and efficient response

Answer 20

Yes and no. We expect a core survival program, a “hub”, to be conserved across different drugs. However, we also expect the upstream signals that trigger this hub to be context-dependent. Especially when we take into account the known heterogeneity of persisters.

Answer 21

The main reason is logistics and timing. NMT-seq requires fresh samples and immediate processing, which is much easier to control in a pre-clinical setting than with patients. On top of that, setting up the ethical approvals and recruitment for new patient samples could take years to fully implement, whereas using existing public data is a far more efficient way to validate our findings right now.

Answer 22

The LiveDrop platform allows for picoinjection and real-time monitoring of droplet formation, which is crucial for setting up a hightroughput screening platform.

Answer 23

I do not assume all lymph node cells are dormant. However previous work has shown that a change in the environmental niche typically causes a period of dormancy in disseminated cells, before they can adapt to the new niche. And secondly, sentinel lymph nodes are often collected at the time of surgery, so in patients who have often already received neoadjuvant therapies, which in itself can induce dormancy. Furthermore, since we get information on gene expression from ST, we are able to discriminate dormant clusters vs small but proliferating metastases.

Answer 24

Sentinel Lymph Node Biopsy, from the biobank at Bordet, collected at surgery (often following neoadjuvant therapy).

Answer 25

First, within each NMT model, we define clusters based on transcriptomics. Then, within each cluster, we infer regulatory networks to identify state-specific programs. When comparing different models — or mouse and human datasets — we do not compare gene by gene. We assess whether similar dormant clusters emerge across systems. This allows us to identify consensus versus context-specific programs. At the end, the output is a set of clusters (shared or unique) and their regulatory networks, which guide the selection of targetable vulnerabilities.

Answer 26

It’s true that by excluding actively cycling cells, we may miss a fraction of slowly proliferating dormant-like cells. This is a trade-off. I chose to prioritize specificity here. Although it’s important to note that a significant proportion of slowly proliferating populations are expected to be in G1 so we are not completely excluding those populations either.

Answer 27

A combination of 3 criteria: * High dormancy score (using a signature I established previously) * Hypotranscription (using n_count as proxy) * Non-cycling (based on RNA-seq data)

Answer 28

The molecular effect is global, since these drugs target chromatin regulators or enzymes. But the functional impact can be selective, because cancer cells are particularly dependent on certain epigenetically maintained programs, e.g. DAC in AML.

Answer 29

Yes, we can be sure of identifying a few tumor cells in the lymph nodes presenting metastases. First because we will use matching H&E staining to visually identify tumor cells. Secondly, the mammary tumor cells will have a transcriptome very different from to the lymphoid and stromal cells present in the lymph node; and ST is efficient at identifying different cell types.

Answer 30

The lymph node provides a unique window into the early events of dissemination. While we could eventually expand this to DTCs, at the moment expanding to other organs would be a bit too ambitious, it is already a complex project. It is a logical next step, so hopefully in the future we will tackle DTCs.

Answer 31

1- preprocessing (check reads -> trimming -> alignment -> count matrix) 2- Create Seurat object 3- QC ~ nFeature_RNA/nCount_RNA & percent.mt 4. Normalization 5. Feature selection (focus on genes with highest variation) 6. Data scaling 7. Linear dimension reduction by PCA 8. Non-linear dimension reduction (tSNE or UMAP) 9. Cluster the cells

Answer 32

1- preprocessing (check reads -> trimming -> alignment -> deduplication -> sorting & merging of paired reads) 2- methylation extraction (location + status) 3.- data can be aggregated based on RNA for NMT

Answer 33

2 types of information: - Imaging: analysed by QPath - Sequencing analysed by SpaceRanger for preprocessing -> STutilility: selection of regions of interest and analysis with Seurat like a scRNAseq -> Deconvolution with CARD or Stereoscope if needed.

Answer 34

Yes, for instance, there is a study that in which the use of DAC sensitizes BC cells to paclitaxel. (Which is thought to be linked to the reactivation of TSGs).

Answer 35

We decided to focus on epirubicin, paclitaxel and 5FU, 1- because they are major drugs used in clinics and 2- because we have preliminary data in vitro that indicates that they can induce dormancy in several BC cell lines.

Answer 36

Epirubicin: Anthracycline (DNA intercalation/Topoisomerase II inhibitor). Paclitaxel: Taxane (Microtubule stabilizer). 5-FU: Antimetabolite (DNA/RNA synthesis inhibitor).

Answer 37

Yes, I will share generated data through repositories (e.g. Geo, GitHub).

Answer 38

Classical regimen combines immunotherapy (Pembrolizumab) with a chemotherapy backbone of Taxanes, Platinum agents, and Anthracyclines before surgery, possibly with capecitabine after if residual disease found. Although there are new classes of drugs that are being introduced, for instance ones where a drug is coupled with Ab targeting the Trop2 receptor (commonly expressed in cancer cells).

Answer 39

Therapy typically begins with surgery. Radiotherapy follows to ensure local control, and the process concludes with long-term endocrine therapy, often combined with CDK4/6 inhibitors, to maintain systemic control. For high-risk cancers, by chemotherapy can be used as adjuvant therapy as well.

Answer 40

* Loss or mutation of the target: ER downregulation, HER2 activation of downstream or alternative signaling * Genetic mutations, such as in the PIK3CA gene * Epithelial-mesenchymal transition (EMT) : TGFb, MEK, etc. * Tumor microenvironment interactions: ECM and TILs * Drug efflux pumps, like P-glycoprotein * Enhanced DNA repair or resistance to apoptosis => Dormancy is primarily linked to EMT and decreased susceptibility to apoptosis, as these programs allow cells to survive the shock of therapy. In contrast, genetic mutations or target loss are active mechanisms for continued growth, whereas dormancy is a reversible, non-genetic state of hiding.

Answer 41

It’s a highthrouput method that is compatible with both fresh-frozen and FFPE (thanks to the cyassist) and compatible with scRNA-seq analysis. (If they push why not HD ST: at the time the project was launched the technology ws not available yet)

Answer 42

Spheroids strike a balance between experimental scalability and physiological relevance. They better recapitulate drug penetration and cellular organization than a 2D model, which strongly influences treatment response. At the same time, they remain compatible with high-throughput screening, unlike in vivo models.

Answer 43

Our primary focus is TNBC, because it remains a challenging subtype with high rates of relapse. However, the nature of our workflow makes it easy to integrate other subtypes in a second step, especially in Aims 2 and 3 as we have access to sequencing data from all subtypes in our cohorts. This allows us to test if the mechanisms we identify in TNBC are also present in other breast cancer contexts.

Answer 44

PAM50 and histological classification, while old, remain a standard in the clinic, so they are still relevant. That being said, one of the strengths of the single-cell approach is that we can also classify individual cells (as tumors are known to be frequently heterogeneous in that regard), so it would be relatively easy to test if a certain subgroup of cells within a tumor correlates better with dormancy.

Answer 45

Dormancy is a hot topic but also a relatively new field, and addressing it is technically ambitious. While some single-cell studies have been performed, it hasn't been done at this scale before because it requires a specific convergence: the right expertise in single-cell multi-omics, access to large-scale patient cohorts, and the technology to bridge those findings into high-throughput 3D screening. We are the first in a position to do this because we have all three pillars in one place.

Answer 46

Most dormant cells are in G0 or early G1, because they are not progressing through S phase or mitosis. However, some dormant-like cells may cycle very slowly rather than being fully quiescent.

Answer 47

Mutations can still arise from DNA damage induced by therapy or oxidative stress, and from imperfect DNA repair. In addition, dormancy is often a dynamic state — some cells may transiently re-enter the cell cycle, providing opportunities for mutation accumulation. So low proliferation reduces mutation rate, but it does not eliminate genomic evolution.

Answer 48

Any field where cellular dormancy is present. This includes bacterial persistence in chronic infections, latent viral reservoirs like HIV, and stem cell niche maintenance in regenerative medicine, or even embryo survival in assisted reproduction.

Answer 49

First there is no reason because it’s a technology in place already at the LISCO institute. But also, we could use individual sequencing for each 'ome' and integrate them computationally. It's not quite as reliable as the NMT and it adds a layer of complexity to the bioinformatic analysis, but it serves as a robust backup with easier sequencing protocols.

Answer 50

It is difficult to give an exact number, but based on previous studies and my own preliminary data, we expect that 10-20 samples per subtype should provide sufficient power for scRNA-seq. For Spatial Transcriptomics (ST), the requirements are more complex due to the spatial architecture, but with a cohort of nearly 100 TNBC samples, we are confident we have more than enough power to characterize the subtype effectively.

Answer 51

First, on a technical level, we can validate our findings through traditional techniques like IHC or FACS on identified targets. But more importantly, we have a functional validation in Aim 3: we can use pharmacological inhibition to test if our candidates are truly required for dormant cell survival.

Answer 52

Two main approaches. First, we can eradicate the dormant pool (either by directly killing these cells or by preventing them from entering dormancy so they remain sensitive to chemotherapy). In doing so we eliminate the source of future relapse. Second, for advanced cases, for instance with multiple metastases where curing is no longer an option, we can instead maintain dormancy on purpose. This is more of a palliative approach, as keeping the metastases asleep as long as possible would ensure a better quality of life for the patient for the remaining time.

Answer 53

That is a potential risk unfortunately. However, our goal is not to target quiescence itself, but cancer-specific survival programs that sustain dormant cells under therapeutic stress. To distinguish these from physiological quiescence, we will compare dormant cancer cells with normal quiescent populations in human samples. Candidate targets would be selected based on cancer-specific dependencies rather than shared cell cycle suppression programs.

Answer 54

To integrate these diverse technologies and species, we use the RNA layer as a common denominator. For the mouse-to-human comparison, we can easily map the genes to their orthologs first. Then we can use tools like Harmony to align them into a shared space. (Harmony projects cells into a shared low-dimensional space and uses a clustering algorithm to identify matching cell populations across different samples. It then applies a correction to pull these clusters together in an iterative manner until the technical batch effects are minimized, without losing the original biological variation.)

Answer 55

If the PDX cells are too sensitive, we can adapt the treatment by lowering the dose to ensure a surviving dormant population remains. Conversely, if they are already resistant, we can pivot to the other chemotherapies in our pipeline; this is precisely why we plan to test several agents, allowing us to select the most appropriate drug-model combination to induce the dormancy mechanism.

Answer 56

For ST we are planning to use CARD (as it incorporates the spatial information). For bulk RNA-seq, CIBERSORT.

Answer 57

Mainly because there is a significant technical hurdle in capturing both RNA and ChIP signals from the same physical cells, especially from in vivo samples. On the other hand, NMT-seq is an established and robust method for capturing multiple layers, so much more efficient at building these multi-layer networks. As a caveat, we can to some extent take advantage of external data, such as the H3K27me3 study by Marsolier et al. However, this will require in silico integration, which naturally won't be as precise as the true single-cell matching we obtain from our NMT-seq pipeline.

Answer 58

No, only part of aim 1 will be performed with the support of the Leuven teams. The majority of the project will be conducted in Brussels, including all the in vitro work, the work related to human biopsies, and all the bioinformatic analyses.

Answer 59

At the moment we are not planning single cell proteomics, too out-of-the-scope. But we can easily at least validate some targets by IHC or IF. We even have access through our collaboration with Thierry Arnould at UNamur to a mass spectrometry platform, so we could perform a proteomics analysis if wanted.

Answer 60

It's a model developed by Alexandra Van Keymeulen, with a mutation in the Pik3ca gene that is TAM-inducible under a K8-CreER, with p53 deletion. (Injection at 4-5 weeks old => latency of about 5 months)

Answer 61

First, relapse does not necessarily require a pre-existing CSC pool. Tumor-initiating capacity is a functional property that can be reacquired through plasticity. If a sufficient number of persister cells survive therapy, some may regain stem-like features and drive relapse. Second, dormancy is a cell state, not a cell identity. CSCs themselves can enter dormancy, and dormancy may actually protect them during treatment. Therefore, targeting dormancy addresses both non-CSC persisters and potentially quiescent CSCs.

Answer 62

I don’t assume a single master regulator, but rather a network of interacting pathways. The goal is to identify bottlenecks — meaning pathways that dormant cells cannot compensate for. Importantly, under chemotherapy pressure, dormant cells are already constrained, which may reduce redundancy and create specific survival dependencies. These are the candidate bottleneck that we want to test experimentally.

Answer 63

The distinction is the transition from a acute reactive state to a more long-term survival program. An acute stress response is a short-term, high-energy 'panic' mode. The cell is metabolically hyperactive, burning ATP to fix immediate damage (like protein misfolding or DNA breaks), and it frequently leads to apoptosis if the repair fails. Dormancy is the opposite: The cell shuts down non-essential activity and enters a low-power state to wait out the threat. And if anything the cells are less susceptible to apoptosis in that state.

Answer 64

-a drop in biosynthetic activity (production of RNA/DNA/protein) -a drastic reduction in ATP consumption -sometimes: shift toward Oxidative Phosphorylation (OXPHOS) or Fatty Acid Oxidation (FAO) -inhibition of mTOR which can lead to the activation of autophagy

Answer 65

-lack of druggability -essential in healthy tissues => would cause side effects -lack of conservation (a very specifc one would not be my first choice at least)

Answer 66

That is a potential risk unfortunately. However, our goal is not to target quiescence itself, but cancer-specific survival programs that sustain dormant cells under therapeutic stress. To distinguish these from physiological quiescence, we will compare dormant cancer cells with normal quiescent populations in human samples. Candidate targets would be selected based on cancer-specific dependencies rather than shared cell cycle suppression programs.

Answer 67

Currently, clinicians track breast cancer patients by waiting for a mass to appear on an MRI or a lump to be felt during an exam. But these methods only detect the cells after they have already re-awakened and regrown a mass. Occasionally dormant cells have been found in biopsies (BM or ctDNA).

CNP practice Flashcards

(91 cards)