Treatment validation on mouse models

Treatment validation on mouse models

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I am struggling to set up a project proposal for validation of a known treatment for metastatic colorectal cancer in mouse models. I want to see how SNPs in patients contribute to their drug resistance and/or toxicity. For the validation of these SNPs as predictive biomarkers, I will use cell lines and mouse models. But, how am I supposed to "induce" these specific SNPs in the mice? CRISPR/Cas9 can be used for the alteration of bigger parts. Any idea is a useful idea!

Metastatic brain tumor models and drug brain penetration analysis

3.3.3 Advantage of GEMMs

GEMMs of brain tumors with specific genetic alterations observed in human brain tumors have been created and characterized. Since the created mutations in GEMMs are similar to those observed in human patients, the GEMMs more faithfully reflect important genomic characteristics of clinical brain tumors. In many cases, GEMMs had similar histopathological and biological features to human brain tumors. Therefore, these GEMMs can be used for investigation of brain tumor development, progression, genetic alterations, and therapeutic strategy. One of the most important advantages of GEMMs compared to xenograft models is that using GEMMs could address whether a specific molecular alteration is responsible for brain tumor development and progression. Moreover, GEMMs provide information about the sequential events or alterations after specific gene mutations. In addition, GEMMs also provide a tool for investigation of tumor and stroma interactions during metastatic brain tumor homing, invasion, and progression. GEMMs are good models for investigation of biological features of genetically altered cancers in immuno-competent animals. It can also be used in studies on drug penetration to the brain through the BBB.


Cancer is a heterogeneous disease with intra- and inter-tumor genomic diversity that determines cancer initiation, progression and treatment. The understandings of cancer biology and the development of therapeutics have been aided greatly by a variety of mouse tumor models, including cell line-derived xenografts (CDXs), patient derived-xenografts (PDXs), genetically engineered mouse models (GEMMs), cell line- or primary tumor-derived homografts in syngeneic mice and so on (reviewed by [1,2,3,4]). These models differ in their generation, host and tumor genomics and biology, availability, and research utilizations. For example, immunotherapies are tested in immunocompetent models such as GEMMs and syngeneic models.

Past decades witnessed the accelerated creation, distribution, profiling and characterization of mouse tumor models [5,6,7,8,9,10]. The abundant collections made it possible to conduct the so-called “mouse clinical trials (MCTs)”, in which a panel of mouse models, dozens to hundreds, are used to evaluate therapeutic efficacy, discover/validate biomarkers, study tumor biology and so on. MCTs demonstrated faithful clinical predictions in multiple studies [6, 11,12,13,14,15]. While most reported MCTs used PDXs, MCTs using other mouse models, such as syngeneic models, are now widely performed as well.

Because of their resemblance to clinical trials, MCTs are often analyzed by methods for clinical trials. For example, overall survival (OS) and progression-free survival (PFS) are estimated by tumor volume increase, Cox proportional hazards models are used for survival analysis, response categories are defined by tumor volume change and objective response rate (ORR) is calculated [6, 13, 16]. However, MCTs differ from clinical trials in many ways. (1) In an oncology clinical trial, a patient is enrolled in only one arm, while in a MCT, multiple mice bearing tumor from the same mouse model are made so that mice can be placed in all arms. Mice from the same mouse model capture intra-tumor heterogeneity for tumor growth and drug response, and mice from different mouse models capture inter-tumor heterogeneity. Measurement error can be quantified when multiple mice are used in each arm. Furthermore, since there are mice of same mouse models in both arms, they themselves can serve as control across arms for better measurement of drug efficacy. (2) tumor volumes are routinely measured every few days (3) mouse models are usually characterized with genomic/pharmacology/histopathology annotations (4) MCTs are done in labs that reduces/removes various noise and inconvenience encountered in clinical trials, such as dropouts, long trial time and concomitant medication.

In this study, we combine empirical data analysis, statistical modeling and computational simulations to address some key issues for MCTs, including the determination of animal numbers (number of mouse models and number of mice per mouse model), statistical power calculation, quantification of efficacy difference between mice/mouse models/drugs, survival analysis, biomarker discovery/validation with and beyond simple efficacy readouts, handling of mouse dropouts, missing data and difference in tumor growth rates, study of mechanisms of action (MoA) for drugs. We will also show MCTs can explain discrepant clinical trial results.

An Easy to Use, Inexpensive Model

Obviously the main feature of syngeneic models is that they are immunocompetent, featuring full murine immunity and comprehensive stroma. Another key factor in their increased use is their relative simplicity compared with other immunocompetent models (e.g. GEMM, humanized mice).

Syngeneics are the immunocompetent model that most closely resemble running a standard xenograft study. Syngeneic cell lines can be easily cultured and expanded in any lab. The models have 100% penetrance and subcutaneous injections can be carefully timed to synchronize tumor development and mimic xenograft study design. This results in short efficacy studies (potentially 2-4 weeks) performed with statistically meaningful numbers of mice per group, which are inexpensive and reasonably simple to run.

Cellular Housekeeping: How autophagy models in mice could lead to treatment in humans

In this modern era of medicine and research increased emphasis is being placed on how we age and how we develop age-related diseases. In this vein, the Levine lab at UT Southwestern has conducted studies on the process of autophagy. Autophagy is the highly regulated natural process through which cells degrade and recycle dysfunctional cellular components, critical in protection against disease and starvation. We know that our ability to perform autophagy declines as we age, and that reducing the ability of animals to perform autophagy significantly reduces their lifespan. Beyond that, however, the effect of autophagy on longevity is not well understood.

In the most common form of autophagy, the components marked for recycling are engulfed to form a structure called an autophagosome. Then the components are transferred to a very acidic lysosome and are eventually degraded through specialized enzymes. Researchers can exploit these processes to change how much autophagy an organism can perform.

In order to study autophagy, researchers have created a mouse model that has increased levels of autophagy. This is performed by mutating a component of what is called the beclin 1 -BCL2 regulatory complex. When BCL2 binds beclin 1, autophagy is turned off. The engineered mutation in beclin 1 prevents BCL2 from binding, and allows beclin 1 to continue to promote the formation of the autophagosome, which results in continuously higher levels of autophagy in the mice.

The results of this study demonstrate that the mice with increased levels of autophagy have a significantly increased lifespan. Studies showed that not only do these beclin 1 mutant mice live longer, but also healthier, having better kidney and heart function as well as less spontaneous tumor formation. Additionally, their premature lethality and infertility is rescued. These results suggest that promoting autophagy in this manner can promote mammalian healthspan and lifespan and should be further studied.

The researchers then wondered if known anti-aging compounds could be producing their effects through a pathway similar to their genetic mouse model. Klotho, a membrane protein, was one such compound they examined. It has previously been shown that animals genetically engineered to be deficient in klotho have reduced lifespan and that administering klotho could extend lifespan. Additionally, it was observed that administering klotho promoted more autophagy. The Levine lab took klotho-deficient mice and observed a noticeable increase in beclin 1 - BCL2 binding, leading to less autophagy. By taking these klotho deficient mice and mutating beclin 1 they were able to rescue the effects of klotho deficiency and return autophagy to normal. Furthermore, by administering klotho to human HeLa cells they were able to reduce beclin 1- BCL2 binding showing that this effect is not isolated to mice, but applicable to humans as well.


The most predominant and striking sign in an AD patient is the progressive decline in cognition, primarily due to loss of neurons and synapses in the hippocampal formation and related areas [18]. As such, a “must have” feature of a valid AD-transgenic model is the ability of the model to accurately reflect the behavioral changes observed in human AD patients. To accurately interpret behavioral results from transgenic mouse models of AD, it is important to intimately understand the behavioral tasks that are most often used to test cognitive changes in mice, as well as what each cognitive test is actually measuring. When examining cognition in animals, behavioral tasks are typically divided into either associative or operant learning tasks. Associative learning tasks use cues in the environment to condition a specific response in the animal. Operant learning tasks require the animal to make a particular response to a specific stimulus in order to receive an outcome. Cognitive tasks are further divided into groups by the type of memory being tested. The following are some of the most often used tasks to determine cognitive changes in mouse models, transgenic or otherwise.

1.2.1. Spatial Memory Tasks The Morris Water Maze

The Morris water maze (MWM) is a particularly sensitive task to examine age-related/AD-like deficits because it is highly specific for hippocampal function, one of the first and most affected brain regions in AD [18]. As a result, the MWM test is one of the most common behavioral tasks used to determine hippocampal spatial memory deficits [19]. The test consists of placing the rodent in a circular tank filled with cloudy water, which is used to motivate the animal to escape the water by swimming to a hidden platform located right below the water’s surface. Over several days the rodent learns to find the hidden platform by using spatial cues, such as posters or taped objects strategically placed on the walls outside of the water maze, in the testing room. Distance swam, latency to reach the platform, and swim speed, most often recorded on video, are common measures of this test. The capacity of the animal to retrieve and retain learned information or the flexibility to purge and relearn new strategies can be determined using a probe trial and reversal trial. In the probe trial the platform is taken out and the animals are allowed to swim in the pool. Time spent in the region that previously contained the platform, crossings over the platform area, and time to reach the platform location are measured. The reversal trial is identical to the training trials, but in this case, the platform is switched to the opposite region of the pool, testing the cognitive flexibility of the animal that is necessary to relearn a new location. A cued version of this task, rendering the platform visible, can also be used to measure nonspatial strategies as well as visual acuity [20]. Variations include the radial arm water maze (RAWM) or plus-shaped water maze [21].

One desirable aspect of this task is that the motivating stimulus, i.e., escaping the water, does not require the food or water deprivation that is common in other spatial memory tasks. However, it has certain limitations as well, one of which is the fact that the various components of memory, i.e., reference and working memories, cannot be tested simultaneously. Radial Arm Maze

One task that can accommodate simultaneous measurement of memory components and has also been widely used to study spatial memory performance in rodents is the radial arm maze (RAM). This maze consists of 8� equally spaced arms radiating from a central platform, which the rodent has to enter in order to attain a food or water reward placed in some of the arms. In this task, the animals guide themselves using spatial cues around the room, with the goal to enter each arm only once to receive the maximum amount of food or water rewards in the shortest period of time and with the least amount of effort. This maze requires the use of working memory to retain information that is important for a short time (within trial information), as well as the use of reference memory to retain the general rules of the task across days. Specifically, the animal must be able to remember which arms were baited as well as which it already entered (working memory), but it also must know to avoid non-baited arms across trials (reference memory), all of which takes place by being able to successfully encode spatial information. However, while this task permits the examination of both reference and working memory, major limitations are the use of food or water deprivation in this task, as well as the presence of odor confounds [22�]. Radial Arm Water Maze

A relatively new spatial memory task, the RAWM, has been designed to eliminate the limitations of the above-mentioned tasks by combining the positive aspects of the MWM and RAM. The difference between the MWM and RAWM is that performance in the RAWM entails finding a platform that is submerged in water located in one of several arms (6𠄸) in the water bath, compared to the classic MWM which only has an open swim field. This makes the task a bit more difficult, but forces the animal to use spatial cues and working memory (keeping track of the arms it has already visited) to remember where the platform is located. Several variations of this task, using different numbers of platforms and platform location organization, have been used to examine spatial memory differences after pharmacological treatment [25,26] and differences across species [27], gender [28], and, importantly, models of AD [24,29].

1.2.2. Contextual Memory Fear Conditioning

Freezing response, defined as a complete lack of movement, is the innate response of rodents to fear. In a fear conditioning paradigm, the animal is placed in a box containing a grid that delivers a mild aversive stimulus for two minutes. In the box, the animal is presented with a tone (usually 80 dB) (conditioned stimulus) that is paired with a mild shock (unconditioned stimulus) at the end of the trial with the result that the tone elicits the freezing response. Repeated exposures are sometimes necessary depending on the strain used or the interval time between the tone and the shock. Some researchers use trace fear conditioning, which increases the time gap between the tone and the shock in order to investigate prefrontal cortical activity. Here, the animal is taken out of the box and returned 24 hr later to evaluate its learned aversion for an environment associated with a mild aversive stimulus (context-dependent fear) by measuring freezing behavior in the absence of tone or aversive stimulus. Cue-dependent fear can be measured by placing the animal in a new box that is different in color, shape, etc., and presenting it with the tone as it explores the new environment freezing behavior associated with the tone is measured.

Fear conditioning is a widely used test to measure hippocampal-dependent associative learning. This test is thought to be sensitive to emotion-associated learning and therefore is a useful measure of amygdalar–hippocampal communication. Many of the transgenic mouse models of AD display impairments in fear and anxiety, which is primarily a function of the amygdala. The hippocampal function used in fear conditioning may be different from learning in a spatial task [30�]. Passive-Avoidance Learning

In the passive-avoidance learning task, the animal must learn to avoid a mild aversive stimulus, in this case darkness, by remaining in the well-lit side of a two-chamber apparatus and not entering the dark where it receives the aversive stimulus. Note that since rodents innately gravitate to darkness, the animal has to suppress this tendency through pairing the negative stimulus with the desired compartment. Animals that do not remember the aversive stimulus will cross over earlier than animals that remember. Dependent measures include the median step-through latency (latency to cross into the unsafe side) and the percentage of animals from each experimental group that cross the threshold within an allocated time [20,33,34].

1.2.3. Working Memory/Novelty/Activity Y-Maze

This test is based on the innate preference of mice to alternate arms when exploring a new environment. Various modifications are available with different levels of difficulty and different demands on specific types of cognition. One version that is particularly popular for the study of cognitive changes in AD transgenic models is the spontaneous alternation version of the Y-maze. In this instance, test animals are placed in a Y-shaped maze for 6𠄸 min and the number of arms entered, as well as the sequence of entries, is recorded and a score is calculated to determine alternation rate (degree of arm entries without repetitions). A high alternation rate is indicative of sustained cognition as the animals must remember which arm was entered last to not reenter it [35].

A short-term memory version can also be carried out in which one arm of the Y-maze is blocked and the subject is allowed to explore the two arms for 15� min. The animal is then removed from the maze for a few minutes or up to several hours, depending on the experimental manipulation, and then placed back into the maze, this time with all arms open, to explore for 5 min. Animals with preserved cognitive function will remember the previously blocked arm and will enter that one first on the second trial. This test can also be repeated a week after the last trial with a delay time of only 2 min between the trials in order to test long-term memory and the time it takes the animal to relearn the task. Typically measured parameters include the first arm entered, amount of time spent in each arm, and total number of arm entries [35]. T-Maze

T-maze tasks are incredibly well characterized and are widely used for cognitive behavioral testing in both mice and rats. Animals are started at the base of the T and allowed to choose one of the goal arms abutting the other end of the stem. If two trials are given in quick succession, on the second trial the rodent tends to choose the arm not visited before, reflecting memory of the first choice. This is called “spontaneous alternation.” This tendency can be reinforced by making the animal hungry and rewarding it with a preferred food if it alternates. Both spontaneous and rewarded alternations are very sensitive to dysfunction of the hippocampus, and hence are sensitive to AD-like symptoms, but other brain structures are also involved. Each trial should be completed in less than 2 min, but the total number of trials required will vary according to statistical and scientific requirements [36]. Object Recognition

The object recognition test is based on the natural tendency of rodents to investigate a novel object instead of a familiar one, as well as their innate tendency to restart exploring when they are presented with a novel environment. The choice to explore the novel object, as well as the reactivation of exploration after object displacement, reflects the use of learning and recognition memory processes. The available object-recognition tasks to test cognition in rodents use different numbers of available objects and environments in which the animals are tested, as well as types of configuration aimed to test spatial recognition and novelty, among other things. One particular object recognition task that is sensitive to age-related deficits is very suitable to test AD-related deficits [37�]. In this task, a rodent is placed in a circular open field filled with different objects (i.e., various plastic toys of different sizes and shapes) for 6 min. After a series of trials, during which the animal has habituated to the configuration and properties of the different objects, some of the objects are switched from one location to another to assess spatial recognition. Subsequently, some of the objects are replaced with new ones to evaluate novel object recognition. The time spent exploring the open field (movement/inactivity) as well as number of times and length of time inspecting each object over the different trials is calculated. Open Field

The open field locomotion test is used primarily to examine motor function by means of measuring spontaneous activity in an open field. The circular or square open fields vary in size depending on the experiment and are divided into distinct quadrants or sections. The animal is placed in the open field and the movements of the animal are either videotaped or monitored by automated computer programs. Rearing, line crosses, cleaning, general movement, number of lines crossed, preference for particular sections, and/or fecal movements can all be calculated to examine behavior and anxiety [40,41].


5.1. Personalized Medicine: Tumor Neoantigen–Based Therapies

The use of next-generation sequencing technologies to study the mutational landscape of tumors allows for the identification of tumor neoantigens. These neoantigens, arising for tumor-specific mutations, represent ideal targets for personalized cancer immunotherapies, as they should be absent from normal tissues and could be recognized as a non-self-antigen by the immune cells. Although HIS models today are not being used to study vaccination strategies using neoantigens, they can be useful for the in vivo validation of candidate in silico–predicted neoantigens in adoptive T cell transfer approaches. For example, human CD8 + T cells enriched for predicted breast cancer neoantigens were able to protect mice from tumor challenge with autologous PDXs (Zhang et al. 2017), and neoantigen-specific TCR-transduced T cells showed antitumor efficacy in NSG mice engrafted with an acute myeloid leukemia cell line expressing the neoantigen (Van Der Lee et al. 2019).

5.2. Personalized Medicine: Avatar Models

HIS mice engrafted with autologous human immune cells and PDXs constitute so-called ideal avatar models, inasmuch as they can be used to predict a patient's response to immunotherapies and can help adjust personalized treatment at relapse. Nevertheless, a relatively low number of patients with advanced cancer survive long enough to benefit from an avatar trial performed with his or her own PDX. Indeed, the setup of an avatar model requires three time-consuming steps: the PDX establishment (in general, the faster the PDX engrafts, the more aggressive the tumor is in the patient), its in vivo expansion to obtain enough tumor tissue, and the preclinical evaluation of standard versus novel therapeutic strategies. Some studies have already succeeded in generating PBL-HIS models using autologous PBMCs and PDXs. In such settings, anti-checkpoint mAbs delayed a gastric tumor growth (Sanmamed et al. 2015), and a bispecific Ab inhibited a HER2 + breast tumor with better efficiency when mice received autologous PBMCs than when they received nonmatched PBMCs (Rius Ruiz et al. 2018). Interestingly, some studies have shown a correlation between therapeutic outcomes in the avatar mice and those in patients. Along these lines, adoptive transfer of autologous TILs inhibited melanoma PDX growth in hIL2-NOG mice following an analogous responder/nonresponder pattern in the patient similarly treated with ACT (Jespersen et al. 2017). Additionally, the anti-PD1 response observed in an avatar model using the patient's own melanoma PDX and TILs supported therapeutic decision for this patient (Ny et al. 2020). The avatar approach using HIS models still needs to be validated, but it could eventually guide the future use of immunotherapies as precision medicine and help to identify predictive biomarkers of clinical responses.

We thank Hayley Donnella and Attila Gabor for useful discussion and feedback on the manuscript. We thank Diana Panayotova Dimitrova and Maria Feoktistova for constructive discussion about apoptosis pathway. FE thanks European Molecular Biology Laboratory Interdisciplinary Post-Docs (EMBL EIPOD) and Marie Curie Actions (COFUND) for funding, and JRC for Computational Biomedicine which was partially funded by Bayer AG.

JS-R and CAM conceived the project. JS-R and FE designed the project. FE performed the analysis under the supervision of JS-R. PJ, MJG, JW, and TC performed experimental validation and helped with results interpretation. FE and JS-R wrote the manuscript. PJ contributed to manuscript finalization. All authors approved the final manuscript.

Data Availability Statement

The original contributions presented in the study are publicly available. This data can be found here: The mouse RNA seq/microarray data have been deposited in the NCBI Bioproject database under the accession number PRJNA690520 which can be accessed using the following link: The human microarray data is available online via Mendeley Data repository with DOI link at The RNAseq data in mouse GC after ivermectin treatment is available from the authors upon request.

Single-Cell Method Could Eliminate Need for Mouse Models to Study Cancer Treatment Resistance

NEW YORK – To study how cancer cells develop resistance to treatment, researchers are largely limited to using patient-derived cell lines or mouse models. But such methods, while informative, don't always give the whole picture: For example, they can't show how the mutated cells that are able to evade treatment interact with the tumor microenvironment.

To study these cells and their evolution in their native context, researchers need a way to study them in patient samples. A new single-cell transcriptomics method developed by researchers at Weill Cornell Medicine, the New York Genome Center, and elsewhere may offer investigators this opportunity.

In a study recently published in Nature, the researchers described their development of Genotyping of Transcriptomes (GoT), a method to integrate genotyping with high-throughput droplet-based single-cell RNA sequencing.

The challenge they were addressing was how to connect genotypes to phenotypes in single cells, according to the paper's senior author Dan Landau, a cancer researcher at Weill Cornell Medicine and a core member of the New York Genome Center. Although there have been many studies looking at oncogenic mutations in single cells, annotating somatic mutations at high throughput has been a problem.

"Somatic mutations are present in the single cell transcriptome because coding mutations are often transcribed," Landau said. "We did some analysis in the paper showing that the large majority of driver mutations are transcribed, often either at the equivalent rate or at a superior rate compared to what we see in the exome."

Further, defining the transcriptomic identity of malignant cells is challenging when the cancer clones lack any cell surface markers that could distinguish them from one another. But while droplet-based sequencing — such as the 10x Genomics platform, which the team modified to create the GoT method — enables researchers to profile the transcriptomes of thousands of cells, current methods provide sequence information for only short fragments at the transcript end, limiting the ability of these techniques to jointly genotype somatic mutations.

"Single-cell methods [like the 10x Genomics platform] have used digital sequencing, which means that they only preserve a small tag at the end of the transcript to quantify the number of transcripts found in each cell," Landau said. "We reasoned that we would need to extend this to loci of interest. This is essentially the crux of the method."

The researchers modified the 10x Genomics platform to amplify the targeted transcript and locus of interest, then investigated amplicon reads for mutational status and linked the genotype to single-cell gene-expression profiles using shared cell barcodes. They used GoT to profile 38,290 CD34+ cells from patients with CALR-mutated myeloproliferative neoplasms to study how somatic mutations can corrupt the process of human hematopoiesis.

They mixed mouse cells harboring a mutant human CALR transgene with human cells containing a wild-type human CALR transgene and applied GoT in order to test its ability to co-map single-cell genotypes and transcriptomes in a mixed-species context. They found that a significant majority of the cells with transcripts aligned to the mouse genome showed mutant CALR whereas cells with transcripts aligned to the human genome showed wild-type CALR — in total, 96.7 percent of the cells matched their expected species.

"CALR is expressed at the level of something like 200 to 300 [transcripts per million] in bone marrow cells, and we genotyped 90 percent of the cell," Landau said. "I think what we're offering here is a method that includes the ability to genotype across a wide range of scenarios in terms of high efficiency across the expression range. Obviously the more the gene is expressed, the easier it is to capture it in the transcriptome [and] if the mutation is very close to the transcript end, it is easier to capture. If it's further away, you'd need the methods we've developed."

The researchers have also created an analytical platform called Iron Throne that is capable of connecting the single cell transcriptome to the genotype information, while filtering out multiple sources of background noise such as PCR recombination and PCR errors, to provide high-accuracy precision genotyping, Landau noted. They've made both tools available on GitHub.

The potential for being able to perform controlled gene expression analysis experiments is significant, according to Landau. Using this method, a researcher can compare wild-type and mutant cells within the same individual, with the wild-type cells serving as the ultimate controls.

"Everything is the same — the microenvironmental conditions, the technical confounders, the patient's breakfast. The only thing that's different is the genotyping," Landau said. "Now, we can read it out with this very high-precision tool."

In the context of myeloproliferative neoplasms, which the team studied for the Nature paper, the ability to distinguish cells from one another with high fidelity is especially important. "These patients have both normal and mutated bone marrow development coexisting within their marrow, and in this case the cells are phenotypically indistinguishable with current methods. If you look under the microscope, you'll see that there are too many cells, but you can't distinguish between the wild-type and the mutant — the flow markers are the same," Landau said. "Therefore, there was a significant challenge. We also show that in the RNA-seq data alone, again these two processes are intermingled."

Another context to which the researchers applied GoT was clonal growths in normal tissues. "It's been reported by multiple groups that each one of us essentially harbors hundreds of distinct clonal growths in multiple tissues," Landau said. "And this is in normal-appearing tissue — the tissue is morphologically normal, and the cells look the same. And yet somehow, these clones are growing, and they contain somatic mutations."

Studies have described mutations affecting NOTCH1 in the esophagus, NOTCH1 and p53 in the skin, and mutations in various genes affecting the blood, he added. In this context, studying the transcriptome is not likely to provide answers, so there's a need for a method that can annotate cells that look alike and yet may be differentially mutated.

"We need to be able to start reading out what allows these cells to grow better than their wild-type counterparts," Landau said.

Importantly, this method could change the way that treatment resistance is studied. The way researchers currently probe questions on resistance is by using mouse models or cells lines, but to understand how resistance forms in a native context, "if we want to know what's the relation to the immune microenvironment, we need a method that can actually capture this information in patients," Landau said.

By isolating cells that are both wild-type and mutated for a particular subclonal mutation that confers resistance, and studying them within patient samples, researchers can begin to look at the differences between them in a well-controlled way, without the problem of patient-specific confounders.

"Let's say that you have a genotype that you hypothesize is related to therapeutic resistance. Now you're faced with the question of finding the mechanism. Why are these cells more resistant or less resistant?" Landau said. "The cells are not distinguishable by any other means. They don't have different cell surface markers. They're morphologically the same, so you can't sort them."

Therefore, he added, "you need a method that can annotate individual cells by their genotype and, at the same time, capture at high throughput their single-cell transcriptomes. That allows you to develop a hypothesis directly in patients. You take these cells and you ask what's different about them when they are exposed to therapy in the patient's blood or in the patient's tissue."

At this point, he noted, the GoT method is better suited for basic research, though there wouldn't be a limit to the types of cancers it could be used to study. The method could even be used as a drug discovery tool, to help pharmaceutical companies determine treatment efficacy.

"I think that this could be something that would empower a combination therapy design because you can — and we are collecting this sort of data — have a sort of clinical trial where combination therapies are introduced gradually," Landau added. "And again, you have this unique ability to compare wild-type and mutated subclones."

The team is continuing to develop GoT in multiple ways, and is looking at the possibility of adding layers of information to the analysis, such as DNA methylation.

Authors' Contributions

Conception and design: E. Bruckheimer, V.E. Velculescu, D. Sidransky, M. Hidalgo

Development of methodology: K. Paz, P.P. López-Casas, F. López-Rios, F. Sarno, V.E. Velculescu, M. Hidalgo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Garralda, P.P. López-Casas, A. Katz, F. López-Rios, F. Sarno, D. Vasquez, E. Bruckheimer, A. Calles, D. Sidransky, M. Hidalgo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Garralda, K. Paz, P.P. López-Casas, S. Jones, F. López-Rios, F. Al-Shahrour, S.V. Angiuoli, L.A. Diaz, V.E. Velculescu, D. Sidransky, M. Hidalgo

Writing, review, and/or revision of the manuscript: E. Garralda, K. Paz, P.P. López-Casas, S. Jones, F. López-Rios, A. Calles, L.A. Diaz, V.E. Velculescu, M. Hidalgo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Garralda, L.M. Kann, F. Sarno, D. Vasquez, E. Bruckheimer, A. Calles

Study supervision: K. Paz, V.E. Velculescu, A. Valencia