Genetic Engineering Breaks Ground In War Against Cancer

January 10, 2023

By Deborah Borfitz 

January 10, 2023 | Multiyear efforts using genetic engineering to study cancer development and interactions with components of the anti-tumor immune response were the subject of the Ernst W. Bertner Memorial Award keynote address delivered at the recent Leading Edge of Cancer Research Symposium of the University of Texas MD Anderson Cancer Center by Tyler Jacks, Ph.D., professor of biology at Massachusetts Institute of Technology. Jacks is also president of Break Through Cancer, a foundation established less than two years ago as a partnership between five leading cancer centers around the country. 

The focus of his talk was genetically engineered mouse models established to understand how cancer develops over time and to inform new approaches for both earlier detection and better treatment. Examples of these genetic engineering tools include the so-called “OncoMouse,” first used in the mid-1980s by researchers from Harvard Medical School to study breast cancer. 

Jacks explained how gene targeting technology can be used with embryonic stem cells to modulate genes of interest. Experiments starting in the 1990s targeted the p53 tumor-suppressor gene. Both “constitutively active” gain-of-function oncogene models and constitutional (germline) loss-of-function mutations in tumor suppressor genes have been created. 

Regulated expression systems have also been used to switch an oncogene on and off to establish the consequences of inactivating its function in a fully established tumor, he says. Conditional approaches have likewise been used for loss-of-function mutations. The first such example was when Cre-lox technology was used to delete the Apc (adenomatous polyposis coli) tumor suppressor gene in adult colonic tissue, leading to the establishment of adenomas and adenocarcinomas.  

One of the workhorse models in Jacks’ lab is mouse lung adenocarcinoma driven by conditional alleles for K-ras oncogene expression and p53 deletion—the KP model. With this model, viral delivery of Cre via an engineered adenovirus or lentivirus simultaneously accomplishes both of these steps, which is sufficient to drive the initiation of lung tumors that eventually progress to metastatic disease. 

The beauty of the model, and others like it, is that it allows researchers to watch the disease process unfold over time—from the normal counterpart cells (alveolar type 2, or AT2) to the early stages of hyperplasia and then the transition to true cancers such as adenocarcinomas that can become invasive and eventually spread, Jacks says. One can study the consequences of mutations specific to adenocarcinomas atop other foundational mutations shared by all or almost all tumor cells.  

‘Any mutation... Anywhere’ 

Experiments of this sort became much easier with the development of CRISPR-based methods for editing genes of organisms, including mice and humans, he adds. Jacks and his colleagues have applied these methods to accelerate the consequences of loss-of-function mutations in potential cancer genes. 

In early CRISPR-based trials, for example, they introduced viral vectors carrying Cre to manipulate K-ras and p53, as well as to guide Cas9 to specific genetic locations where it would introduce mutations so that their impact could be assessed, says Jacks. The most recent version of CRISPR-based gene editing being used in his laboratory involves Cas9 fusions with more precise editing, including prime editing, which allows point mutations to be introduced without generating double strand breaks. An example of this is knock-in alleles that precisely modify specific nucleotides in the vicinity of the Cas9 fusion protein after it is ushered to a specific site in the genome by a prime editing guide RNA (pegRNA).  

Jacks points to a study from the Liu laboratory (Nature Biotechnology, DOI: 10.1038/s41587-021-01039-7) where Cas9 nickase was fused to a portion of reverse transcriptase from a murine leukemia virus. Because nickase creates a single strand nick, the method is much less prone to non-homologous end joining with associated deletion and insertion mutations, he notes. In prime editing, the pegRNA has an extension that includes homology to the strand downstream of the nick and contains the programmed mutation—in reverse orientation, which is why the fused reverse transcriptase is necessary to generate the complement sequence for incorporation.  

A mouse model was subsequently built in Jacks’ lab that had the prime editor knocked into the genome at the Rosa26 insertion site (located on chromosome 6) but engineered not to be expressed except in the presence of Cre, he says. Researchers also knocked in a neon green recorder, so they’d know the cells in which the prime editor had been activated.  

The prime editor allele created in Jacks’ lab can be used in pancreatic organoids or fibroblasts in culture to install K-ras G12D or K-ras G12C mutations with “quite reasonable efficiencies,” he continues. “These are cell-based experiments, but we can also do it in the whole animal.” 

Jacks next discussed this scenario for an animal that is wild-type for K-ras and has two floxed alleles (markers that enable Cre recombinase to excise the intervening portion) of p53 and the conditional prime editor allele. “If we now introduce a vector that has Cre recombinase, plus pegRNA that either directs the mutation of K-ras G12 to D or C, one can develop tumors in the lung... that carry [those] precise mutations and as best we can tell no other mutations.” 

With this new tool, it should be possible “to make any mutation in any gene of any sort anywhere,” continues Jacks. He specifically mentions a series of oncomutations, tumor suppressor mutations, subtle mutations in regulatory sequences, and the creation of antigens in genes of interest. “I think this technology will supplant a lot of the things that we currently do to... accelerate and make more powerful our assessment of cancer-associated alterations.” 

Lineage Tracing 

In a 2022 article published in Cell (DOI: 10.1016/j.cell.2022.04.015), Jacks and his colleagues describe CRISPR-based lineage tracing in the KP model of lung adenocarcinoma. Here, researchers succeeded in  

randomly integrating tracer cassettes throughout the genome, each expressing 3 sgRNAs targeting 3 sites within the same cassette. Now, the introduction of lentiviral Cre initiated the KP tumor development program as usual, in addition to mutations within these cassettes, which continue to change over time. This serves to “mark those cells with a permanent mark, a scar if you will, and these different scars can then be evaluated to determine... [their] clonal relationship,” Jacks said. 

At the end, he continues, the tumor was harvested, and a phylogenetic analysis was conducted to affiliate various cells to each other based on their mutational patterns. This showed different modes of clonal expansion within the tumors and allowed the linking of clones within the primary tumor to metastatic clones elsewhere in the animal. 

When tumors were isolated at the conclusion of the experiment, gene expression analysis was performed at the single-cell level. It was therefore possible to associate the gene expression state of the cells with their clonal history and to look at the plasticity of cells during their development. This method allowed for a detailed picture of the trajectory from normal cells, through transient increased plasticity, to the establishment of rare subclones with transcriptional programs that allow them to rapidly expand with metastatic capability. 

Epigenetic Controls 

In terms of tumor evolution and progression, Jacks and his colleagues have long wondered what in the genome or epigenome drives the transitions they could readily see were occurring with cancer. They initially believed the changes were due to acquired mutations, which made sense based on precedence in human cancers. 

But after the genomes of many of these KP lung tumors were carefully analyzed in the lab, they were found to have very few acquired mutations, Jacks reports. “These tumors do not acquire secondary or tertiary mutations at high frequency, so we believe these transitions are not... genetically controlled [but] perhaps they are epigenetically controlled.” 

Indeed, a fair bit of evidence has accumulated over the years suggesting this is the case, he continues, including changes in transcription factors that are known to control cell states in the lung. “A good example of that is NXK2-1, which is the master regulator of lung development and is highly expressed in... alveolar type 2 cells,” the precursor cells of lung adenocarcinoma. Early, pre-cancerous lesions in the KP model resemble alveolar type 2 (AT2), including having high levels of NKX2-1 expression. 

Importantly, as tumors progress, levels of NXK2-1 drop over time due to epigenetic silencing, and the cells begin to change their character, or gene expression state, Jacks says. In late-stage disease, for example, the transcriptional regulator HMGA2, which is usually restricted to embryonic development, is upregulated, and in intermediate-stage tumors, HNF4 alpha activity becomes detectable. 

“That is interesting because [HNF4 alpha] is most associated with gut development and function... [and], in embryogenesis, the lung arises from the gut.” This may be an example of the tumor “undergoing a sort of reverse development.” 

Clustering Dots 

Researchers in Jacks’ lab decided to start exploring tumor evolution in the KP model at the single-cell level using a Smart-Seq2 protocol. To do this, he says, they created models where the cancer cells could be marked with a red fluorescent protein to distinguish them from non-cancerous cells in the tumor or tissue. 

“We could compare the gene expression profile of the counterpart cell, the AT2 cell, as well as cells from different stages of tumor development—hyperplasia or adenomas, early and late, or adenocarcinomas and metastasis,” says Jacks. With green representing the early normal counterpart cells and red the most distant phenotype seen in these lung cancers, individual cells were positioned on a plot as dots with those closer together being more similar in transcriptional space. 

It was readily apparent that AT2 cells tightly cluster together early in cancer development and that over time the tumor cells become “less and less like normal counterpart cells and also become quite heterogenous with distinct clusters,” he says. “[Tumor] evolution makes sense as we begin to look at the specific nature of alterations or gene expression patterns in these cells at different time points.”  

The research team has found, for example, that surfactant protein C—an important target gene of NXK2-1—is highly expressed in the normal counterpart cells (AT2 cells) in early-stage disease but decreases as the tumor evolves. HNF4 alpha, on the other hand, is not expressed in early stages of the disease but becomes more highly expressed in intermediate stages. 

HMGA2 is also not expressed in the early stages of the disease but gets turned on later in a subset of cancer cells, he adds. In very late disease stages, probably pre-metastatic tumors, there is evidence of epithelial-mesenchymal transition (EMT) that is marked by ZEB2 (zinc-finger E-box-binding homeobox 2). 

To get at the epigenetic regulators that dictate these transitions, continues Jacks, researchers turned to ATAC sequencing allowing them to look at the chromatin state of cells in evolving tumors. ATAC-seq can be run on bulk samples but more recently was converted to a single-cell test method that has been used to compare the chromatin profile of normal cells to developing lung tumor cells (Cancer Cell, DOI: 10.1016/j.ccell.2020.06.006). 

Here, the dot plot provided visualization of cell similarity to its neighbors in the chromatin profile space, Jacks points out. Cells derived from metastases look most different from the normal counterpart cells, but interestingly, some of the cells derived from primary tumors share the same epigenetic profile as cells derived from metastases. “We believe and have evidence that these are pre-metastatic cells, [that they] are buried in the primary tumor and seed metastases at a later point.”   

Jacks adds, “It is likely there are multiple paths cells take—it might be a northern path or a southern path, and there might be offshoots of this path that lead to dead ends, but broadly speaking, we think that tumor development moves from left [cells of origin] to right [cells of the metastases].” And, using bioinformatics analysis, this provides new insights into what controls these transitions.  

Cancer Modules 

Some cancer cells share the signature or “module” of AT2 cells, which are physically adjacent to cells that look more like alveolar type 1 (AT1) cells, says Jacks. Within the lung lineage, AT2 cells give rise to AT1 cells in both development and during normal homeostasis in wound healing, and this same type 2-to-type 1 transition has been confirmed in early-stage tumors, he notes. 

Using immunofluorescence, early-stage tumors have also been shown to have high levels of NKX2-1. While most of these cells are more “type 2-like,” some begin to upregulate the HOPX gene—a marker of AT1 cells. “These are cells on their way to acquiring different phenotypes associated with tumor progression... [and] we can play this out with other molecules,” Jacks says.  

In the middle zone between the type 1 and type 2 modules is another one that looks more gastric, he continues. The early module includes both AT2 and AT1 cells and is highly associated with the NKX2-1 transcription factor. As disease progresses, several gene promoters lose accessibility to NKX2-1 binding sites but gain accessibility to those for HNF4 alpha. 

In module 9—the undefined transition between NKX2-1 loss and EMT induction—the RUNX2 protein stands out, says Jacks. It is a transcription factor not normally expressed in the lung that becomes expressible as the gene turns on in late-stage disease. It is believed to be a driver of the premetastatic state that ultimately seeds metastases, as evidenced by the loss of metastatic potential when RUNX2 was knocked out of tumor cell lines that would otherwise be highly metastatic.   


Cancers are complex and involve multiple cell types beyond the cancerous ones that variably promote and impede tumor development, says Jacks. Immune cells—CD8 T lymphocytes in particular—are worthy of scrutiny, but the KP model has no clear neoantigens, or novel peptides on the surface of cells, meaning there is not much for the immune system to see. To overcome this problem, the vector used to initiate tumors was modified to include not just Cre but also luciferase and the synthetic antigens ovalbumin SIINFEKL and SIYRYYGL (SIY). 

Many antigens have been investigated in Jacks’ lab, and they all “kind of look the same, with some important differences,” he says. “Obviously, strength of binding to MHCI matters, and there are some interesting differences depending on the nature of the antigen, but the findings [from studies involving SIINFEKL]... are by no means dependent on this.” The nature of the antigens for CD8 cells and CD4 cells “likely make a big difference in priming a proper immune response,” and this needs to be considered when designing future vaccines.  

In early-stage disease in antigen-expressing tumors, “very strong immune infiltration” can be seen, says Jacks. But at later stages, such tumors have no trouble growing and killing animals. Rather than antigen loss or immune system exhaustion, the reason appears to be immunosuppression. Studies have found contraction of the anti-tumor immune response, as evidenced by the exclusion of ovalbumin SIINFEKL-reactive CD8 T cells from the center to the margins of a tumor and increased levels of regulatory T cells in the tumor microenvironment. 

Researchers in Jacks’ lab have used single-cell RNA sequencing as well as flow cytometry to study changes in the tumor microenvironment that lead to immunosuppression. This has resulted in publication of a series of papers over the last few years exploring different aspects of CD4 and CD8 function. For example, a subtype of CD8 T cells was found to be particularly important in mounting an immune response from the lymph nodes, and their control might be unleashed through various pharmacological methods. 

New Clinical Trial 

The same strategy used for lung cancer research can be applied to other organs, says Jacks, adding that organoid technology has been used by researchers to create immunogenic models of colorectal and pancreatic cancer—the latter of which has led to a recently initiated clinical trial. The organoids had few somatic mutations and were thus stealth to the immune system, so model antigens and fluorescent markers were required. 

When the organoids were transplanted into the pancreas of immunodeficient mice, they grew into “plump, glowing tumors” that killed the mice after a few months, he says. But when put into immunocompetent mice, the animals rejected the tumor about 50% of the time. “The CD8 antigen produced from the cancer cells is recognized as foreign, priming an immune response, and the CD8 T cells eliminate those early cancer cells before they take root.” A “battle of the wills” ensues a small percentage of the time, leaving the developing cancer in a state of equilibrium. In about 40% of cases, antigen-expressing “progressor” tumors arose, which were ultimately lethal. 

The observed failed anti-tumor immune response, he would argue, may be reflective of human pancreatic cancer in patients who likewise express cancer-specific antigens and yet have fully established disease. “Pancreas cancer is one of those cancers that does not respond well to immunotherapy, so we wondered whether this would be a useful model to begin to study the details of immunosuppression.” 

A subsequent study narrowed the focus of researchers to TIGIT, one of the immunosuppressive regulators expressed by CD8 T cells in the microenvironment of the progressor tumors that is in the same class as PD-1, TIM-3, and CTLA-4, says Jacks. In the case of TIGIT, its ligand, CD155, was found to be expressed in the tumors in the model as well as in human pancreas cancer samples. The research team went on to show that tumor-infiltrating lymphocytes of human pancreas cancer patients following a resection showed TIGIT staining at impressive levels (Cancer Cell, DOI: 10.1016/j.ccell.2021.07.007).  

In a “very large” preclinical study evaluating TIGIT-based combination immunotherapy in the model system, ultrasound imaging was used to identify mice with progressor tumors, reports Jacks. “We then enrolled animals with tumors of similar sizes into one of six groups getting no treatment, single-agent anti-PD-1, or anti-TIGIT, or combinations of those with agonistic CD40.”  

For animals in the single-agent and double-agent groups, the drugs had “very little effect,” he says. In contrast, the triple combination produced a substantial response with about 30% of the mice being cured of their disease and about half showing objectively clear, if sometimes partial, responses. 

The phase 1b/2 clinical trial now underway is enrolling pancreatic cancer patients who have responded to the FOLFIRINOX chemotherapy regimen and are in the maintenance phase, says Jacks. The objective is to see if the addition of immunotherapy improves their progression-free survival. 

Tumor Immunopeptidome 

In other recent work, Jacks and his colleagues have shifted their attention to the major histocompatibility class 1 (MHC1) antigens and their associated T-cell receptor as an approach to investigating the tumor immunopeptidome. These antigens can be immunogenic and tumor-specific. 

Computational methods have been deployed over the years to predict a good peptide for MHC binding based on mutation in both mouse models and in human cancers, he says. In vitro studies—to keep the focus on just the class 1 antigens expressed in cancer cells—can then be performed to isolate those peptides and conduct a proteomics analysis (i.e., immunopeptidomics). 

Jacks and his colleagues used a series of Cre-dependent mouse models to build an allele of MHC1 containing a strep tag that is expressed only in cancer cells and can be pulled down with an affinity reagent, Jacks explains.  

In this way, the group created models in both the lung and in the pancreas where only the cancer cells express the tagged MHCI, says Jacks. They could then homogenize a tumor, pull down the tag with an affinity reagent, and analyze the peptides associated with the isolated MHCI molecules to see which peptides the cancer cells are presenting. 

In the case of early-stage pancreas cancer, they were able to use the system to pull out cancer-specific peptides that were not detectable by other methods (Nature, DOI: 10.1038/s41586-022-04839-2). “We were able to find things in the haystack that were otherwise obscured,” says Jacks.  

Some of the detected peptides are “not present in the normal lung and therefore represent potential, tumor-specific tumor-associated antigens that could be useful for promoting immune responses,” he points out. In fact, many of them have no known expression anywhere else in the animal—so-called “putative tumor-specific antigens.”  

At least some of the peptides are demonstrably immunogenic, says Jacks, based on a vaccine testing approach. Further, research has already shown that when at least one of the antigens is given to mice with tumors, the animals mount an effective anti-tumor, antigen-specific T cell response. 

This creates the possibility for immune-based interception of early-stage disease “by priming the immune system to anticipate what will come in a developing tumor, and maybe [we can] even use that information to do immune-based early detection,” concludes Jacks. 

“Beginning with the OncoMouse from the mid-1980s to today, the use of more and more powerful tools in genetic engineering has allowed the development of a suite of mouse models of many forms of human cancer,” he says. “The careful study of tumor evolution and tumor-immune interactions in these models—using an almost dizzying array of analytical methods—has contributed significantly to our understanding of how cancers arise [and] progress and how they are seen by and evade the immune system. The lessons learned from these models are shaping how we think about human cancer, how it can be more effectively treated, and, one day, prevented.”