Special Report on Cancer: Aiming for avatars
Oncologists seek clarity via patient-derived models
Randall C Willis
Patricia relaxes in a lounge chair, reading the latest thriller from her favorite author. Scrolling up on her tablet, she is distracted briefly by the IV tubing that dangles at her side, connecting her to the medication that’s hopefully destroying the tumor that has become a preoccupying force in her life.
More disconcerting than the tubing, however, is the cart of cages that sits next to her, each cage containing a mouse that is also burdened by her tumor. But whereas Patricia is receiving one drug, the various mice are receiving other therapies either singly or in combination, treatments that Patricia will hopefully never have to experience.
And beyond the cages is an incubator filled with 384-well plates also burdened with Patricia’s health challenge. Within each well are small clumps comprised of cells from her tumor, bathed in various cocktails of approved and experimental small molecules and biologics, each clump providing insights both positive and negative for future angles of attack against her disease.
A wishful fantasy, perhaps, but possibly not for long as researchers and clinicians work toward patient-derived in-vivo and in-vitro models to be used to characterize, monitor and therapeutically screen in real-time co-clinical trials.
Impatient with cell lines
Historically, and continuing today, in-vitro models of various cancers have relied on cell lines originally derived from patient tumors, and to a large extent they have served admirably, as initial screens for a variety of drugs that form the fundamental oncology armamentarium.
That said, as molecular characterization tools have evolved and as more candidate treatments have failed to translate from multiwell plate to hospitalized patient, there has been growing concern that the cell lines no longer represent the cancers they were established to mimic.
“Tumor cell lines are a nice model, but they represent something that a long time ago was isolated from a patient and has evolved for quite some time in vitro, and is a very artificial model,” says Patrick Guye, chief scientific officer at InSphero.
Just as there has been a slow evolution from 2D cell culture to 3D development, so too have researchers expanded their use of patient-derived tumor tissues, hoping that by moving the experiments closer to the source, they will improve the translation of efficacy, toxicity and pharmacokinetics between bench and bedside.
These 3D cultured models generally fall into three categories—spheroids, organoids and microtissues—the conceptual boundaries of which are somewhat vague depending upon who you ask.
Spheroids are self-assembling aggregates of isolated cells, not always of a single type, that provide a 3D architecture much closer to the original tumor than cells grown as a monolayer. That said, they typically lack the cellular organization found in human tissues, which is where organoids come in.
“[Organoids] refers to a group of cells growing in a 3D structure that is generated directly from primary tissues, embryonic stem cells or pluripotent stem cells, with self-renewal and self-organization capacity, maintaining similar appearance and functionality as the original tissue,” explained the University of Pennsylvania’s Anil Rustgi and colleagues in a recent review. “Organoids can be maintained through indefinite passage and preserve genetic stability.”
According to Guye, microtissues take this concept one step further, removing the need for external matrices (such as Matrigel) and allowing the tissues to self-aggregate and -organize, providing a model that better reflects the original source material.
But how well do these 3D models recapitulate what is happening in the parent tumor?
Earlier this year, Michael Shen and colleagues at Memorial Sloan Kettering Cancer Center and Columbia University Medical Center generated and characterized patient-derived bladder cancer organoid lines to determine whether the cultured tissues retain the heterogeneity of the parental tumors and if they demonstrate clonal evolution during culturing and xenografting. Additionally, because many of the bladder cancer patients experienced multiple biopsies during treatment, the researchers were able to generate chronologically distinct organoid lines from the same patient.
Both immunohistochemistry (IHC) and targeted sequencing suggested the organoid lines remained very similar to the parental tumors and, using whole-exome sequencing (WES), the researchers noted tumor evolution in the organoid cultures that resembled those described for recurrent bladder cancer in vivo.
The researchers then screened the organoids with a variety of drugs relevant to bladder cancer, either as standard-of-care treatments or investigational agents being used in clinical trials, again finding results consistent with tumor type (i.e., muscle-invasive vs. non-muscle-invasive) and mutational profile.
“Positive drug responses identified by screening in organoid culture could be validated in organoid-derived xenografts and, in principle, could then be used to guide intravesical therapies,” the authors suggested. “Consequently, as a next step, it will be essential to pursue co-clinical studies to determine whether the response of patient-derived bladder tumor organoids to drug treatment in culture recapitulates patient response to the same treatments in vivo.”
Unlike their animal xenograft counterparts (more below), however, the in-vitro models allow researchers to visually monitor a variety of aspects of tumor biology beyond simply providing a high-throughput efficacy screen for candidate drugs, such as allowing visualization of immune infiltration and drug penetration.
“For this, 3D is perfect,” Guye enthuses. “We see this, for example, with minibodies, which are small fragments of antibodies that penetrate the tissue much more efficiently than full-sized antibodies. And then you have the penetration efficacy of modified cells like the CAR-T cells, which are vastly different depending on how you activate them.”
“It’s extremely impressive because you see 3D microtissues that are so compact, that have so much ECM [extracellular matrix] that it’s almost impossible to dissociate the tissue into single cells using proteases,” he continues. “But if you add some activated CAR-T cells, they express matrix metalloproteases that drill through this tissue in hours and completely dissolve it. So, you can really see the power of the immune system once you activate it.”
As well, Guye presses, the in-vitro platforms offer throughput that simply cannot be accomplished with in-vivo systems.
“If you think about an animal, you can have multiple cohorts but then you’re running into scalability issues,” he says. “If you have microtissues, if you have 96 or 384 in one plate, you can test for different induction conditions—you really can titrate activating stimuli and see which one would be the correct one.”
Speaking specifically about pancreatic organoids, Rustgi and colleagues suggested: “Organoids represent a novel platform for the analysis of gene expression in epithelial cells without contamination by hematopoietic, mesenchymal or immune cells.”
“This model represents a novel approach to validate the genetic alterations that are required for cancer progression, and to identify and elucidate genes associated with early-stage to late stages of cancer progression, treatment response and outcomes.”
At the same time, however, the authors were quick to note a significant limitation of organoids being the lack of important components found in vivo, such as the influence of other tissues.
“To surmount this limitation, efforts are ongoing to coculture organoids with other cell types, in order to generate a more ‘physiological’ microenvironment, and to study potential cell-cell interactions,” they explained.
For its part, InSphero has taken a nod from the organ-on-a-chip movement, and at the SLAS2018 conference introduced Akura Flow, a microfluidic device designed to mimic the multiple organ systems found in vivo.
As Guye explains, the microphysiological device allows researchers to combine multiple microtissues in a single experiment; for example, combining liver microtissue with tumor microtissue to perform bioactivation studies.
“There are certain chemotherapeutic compounds which are by themselves not toxic, but get activated by the liver,” he offers. “And we have very nice data showing that if you add this chemotherapeutic to a tumor microtissue, nothing happens. But if you have a liver present, it gets metabolized into the active ingredient and then kills the tumor microtissue.”
Guye also points to examples where tissues like those in the liver produce cytokines or glucose to support the growth of tumor microtissues, or metastasis studies where they can monitor the migration of tumor cells from one location to another.
“The other application that is also interesting is toxicity studies,” he continues. “You can have, for example, a liver microtissue and a tumor microtissue, you add your therapeutic agent, whatever this is—small molecules, biologics, engineered cells or combinations thereof—and what you want to see is the killing of the tumor microtissues but as little impact as possible on the healthy microtissues, which can be there as a proxy for liver.”
3D cultures and organ-on-a-chip platforms can only get you so far in your efforts to mimic in-vivo tumor and tumor microenvironment (TME) biology. So, taking their cues from cell line xenografts, researchers are implanting patient-derived tumor materials into model organisms, including immunocompromised mice and rats, to form patient-derived xenografts (PDXs).
As suggested earlier, there is a big difference between cancer cell lines and patient-derived tumor tissue, and it took some time for this to be reflected in PDX approaches, which aimed to be closer to the patient experience.
“If you think about the first time that PDXs were established, a lot of the time, the convention was following what we used to do for xenograft models,” explains Martin O’Rourke, senior director of the Oncology In Vitro Biosciences Discovery operation at Charles River Laboratories. “The tumor itself would be homogenized and generated into cell lines.”
“Now, instead of having a suspension or a slurry that was injected subcutaneously, we’re implanting fragments of tumors,” he continues. “So basically, we’re retaining the stromal content of the tumor—a big step forward from maybe 10 years ago.”
PDX implantation efficiencies can vary quite extensively based on tumor type, according to Seiji Okada and colleagues at Kumamoto University.
“Digestive system tumors such as colon, gastric and esophageal cancers tend to have high rates of PDX success,” they noted in a recent review. “In general, clinically aggressive and metastatic cancers tend to have high PDX model engraftment rates compared with less aggressive and primary cancers.”
They also suggested that the site of implantation may contribute greatly to the success of any PDX (see also the sidebar “Is location everything?”).
“Subcutaneous transplantation is commonly used due to the simple procedures involved and easy measurement of tumor size,” the authors wrote. “Orthotropic implantation results in behavior more similar to that of patient tumors, however.”
The model host is also a significant factor as researchers look to build their PDX models, particularly in the era of immuno-oncology and the potential to humanize the immune systems of recipient mice.
“If you’re interested in using a specific immune cell, you should be really careful about, again, what model you are going to use,” says Azusa Tanaka, product manager of precision research models at Taconic. “Many of the immune cells require cytokines for its survival. So how are you going to introduce that in vitro?”
Thus, she continues, some researchers are looking for models where cytokines are expressed via transgenic modification or have been introduced via viral expression.
Tanaka offers an example of a human IL-15 producing mouse line (hIL-15 NOG) developed by Taconic’s Japanese partners at the Central Institute for Experimental Animals (CIEA), led by Takeshi Takahashi. Not only did the IL-15 facilitate proliferation and maintenance of introduced human natural killer (NK) cells, but the NK cells delayed tumor growth in vivo.
Similar efforts are also being performed in collaboration with Experimentelle Pharmakologie & Onkologie (EPO), which has developed several PDX models using immune-humanized mouse strains, including hematological cancers such as AML, ALL, NHL and Hodgkin’s lymphoma.
Humans and mice are not identical organisms, however—so how closely does the PDX tumor match its patient counterpart?
Last year, Johns Hopkins University’s David Sidransky and colleagues generated PDX models with tumor samples from 1,163 patients and fully characterized both systems to see how well the model recapitulated what was happening in the patients.
Examining early-passage mice (P2 or P3), the researchers noted that the PDX models largely retained the histopathological profiles and mutation and copy-number variation (CNV) frequencies of their sources. The researchers then treated the PDX models with the same drugs given to their respective patients and noted close correlation between success or failure in both groups.
They then took this analysis one step further.
“To test whether PDXs developed from early resections retain the ability to replicate treatment outcomes observed for recurrent disease, we screened such models against all therapies employed clinically from disease presentation to subsequent progression,” the authors explained.
Offering several specific examples, the researchers noted that PDX responses largely matched known patient outcomes.
“Collectively, these results suggest that PDXs established early in the disease course may retain the capacity to reproduce patient outcomes to therapies used months later, despite probable tumor evolution during treatment,” they recorded.
This work and other experiences like it suggest that there may be room to use PDX models as avatars of the patients, opening the door to prospective treatment screening rather than the retrospective analysis offered here.
“Advances in high-throughput genomic technologies allow characterization of the cancer genome in a time frame compatible with treatment decisions,” concluded the authors, who include several individuals from Champions Oncology. “The PDX platform can be used to test different empirical treatment strategies potentially targeting the genomic aberrations that drive tumor behavior, offering the unique opportunity to increase clinical benefit.”
“However,” they cautioned, “predicting treatment response to known oncogenic pathways is still not straightforward, and future studies are warranted to determine the feasibility of this approach.”
Champions is one of a handful of companies, according to O’Rourke, that is offering such fee-for-service PDX development and therapeutic screening to cancer patients and their oncologists.
A possible critical factor for the results described above, however, is the fact that the researchers analyzed early-passage mice, when most of the tissue involved in the PDX is still human.
“It’s very important that we totally understand that we have a tumor that has, as much as possible, a complete human component of stroma and tumor cells,” says O’Rourke, “because once you have mouse stromal matter in there, the response will definitely change.”
Monitoring that evolution is a big part of Charles River’s mandate, he says.
“We would take sections of the fragment through the various passages and look at the IHC profiles, so looking at how much human stromal matter and how much mouse is there,” he explains. “Usually, whenever we’re developing a bank, we would take those sections and compare the histology all the way through. Once we find a point where we get to see more than ten percent invasion of the mouse stroma, we usually stop that passage.”
To make this information available to its customers, the company recently launched an online portal called the Compendium. As O’Rourke explains, researchers can identify how many passages have been carried out for an individual tumor, and the percentage of human stromal component that is viable and visible at the passage of interest.
Hera Biolabs’ vice president of research and develoment, Tseten Yeshi, suggests that such efforts to align characterization data with PDX models are unfortunately not yet the norm.
He offers that it is not uncommon to examine the data set of a biobanked PDX line only to realize that whereas the characterization and the histology was done on P3 or P5, the actual sample you receive might be P20 or later.
“Having learned from cell line establishment how different they look over time, the field is a lot more cautious about getting samples and matching them up to the data sets provided,” he continues, forcing people to spend time and effort to molecularly characterize the tissue before they use it.
Such challenges have raised the call for changes to the process of creating PDX models, and Tanaka recalls attending the EurOPDX meeting about two years ago where the need to standardize PDX characterization was front and center. Now is the time, she says, for the field to mature and have agreed-upon standards.
Okada and colleagues echoed this sentiment in their review.
“PDXs have largely been developed within individual institutions,” they wrote. “There is a growing recognition of the need to develop large collaborative groups to create large stocks of PDXs or PDX banks.”
They offered the examples of the 16 European institutions that established EurOPDX, which has already accumulated more than 1,500 samples in its PDX bank, as well as the 450 models available from The Jackson Laboratory. The latter is a co-developer of PDX Finder, which itself contains information on more than 1,900 models.
“Since the character of tumors differs by region, race, etc., it is necessary to establish Japanese and Asian PDX banks in the near future to establish precision cancer medicine for Japanese and other Asians,” they added.
This last point is particularly important, according to Yeshi, as cancer is increasingly being viewed as a personalized disease. Even if you have the same type of cancer as another individual, you might respond very differently to treatment.
“It might be due to the genetic mutations,” he acknowledges, but it could also be other factors, such as the microbiome or epigenetic effects.
Tanaka echoes his thoughts regarding the microbiome.
“There was a publication in Science in January 2018 that showed if you transplant fecal microbiome from a patient who responded to anti-PD-1/PD-L1, you can recreate that response in a mouse model,” she recounts. “That again really shows that perhaps the diversity is not only in the human genetics, but also in the microbiome it carries, and how that is influencing our immune cells.”
“There is clear, established documentation of prevalence between different race types and responsiveness to treatment between different race types, given the same type of cancer,” Yeshi continues. “And the understanding behind that is your whole genome contributes to your disease progression and drug response.”
Thus, it is important to identify and characterize those underserved communities with their own murine avatars.
Rat here, rat now
Despite the growing ubiquity of mouse PDX models, however, some groups are starting to look back to a classic animal model system: the rat.
In part, suggests Yeshi, the switch from the stalwart rat to the mouse had a lot to do with the need for genetically engineered strains, which were very difficult to accomplish in rats. More recently, however, technical innovations are slowly tilting that balance.
“We were spun out of Transposagen, which was primarily a gene-editing tools, technologies and services company,” explains Jack Crawford, Hera’s vice president of business development. “Some of the great gene-editing tools we got to license into Hera straight away, one of those being the Cas-CLOVER CRISPR technology to do targeted genome editing and another being the Piggyback transposon technology.”
Those gene-editing tools, he presses, offered the company an approach to develop new genetically modified rat strains to facilitate, among other things, rat PDX models such as the OncoRat SRG.
Aside from being a better physiological match with humans than mice, Yeshi and colleagues say, the rat offers a few technical advantages in terms of PDX modeling.
Engraftment efficiencies, for one, are much higher in rats. Yeshi offers the example of non-small cell lung cancer (NSCLC) tumors, where the engraftment rate is only about 20 to 30 percent in mice but 85 to 90 percent in rats.
Although such rates might not be mission-critical for common tumor types, they are for rare tumor subtypes or tumors in minority patient populations, says Chris Brenzel, business development manager at Hera.
“You want some kind of prostate cancer from a Caucasian person in the United States, you can find that,” he offers as an example. “If you only find one person with a specific tumor type in a specific minority group, you’ve only got one shot at making a PDX.”
Examining just this question, the National Cancer Institute recently contracted Hera to generate and characterize PDX models of NSCLC tumors from patients in specific minority populations, says Tom Isett, company CEO. The analysis will be two-pronged, they say, with an initial analysis for known EGFR and KRAS mutation hotspots, followed by WES to fully characterize the mutational backgrounds.
Another advantage of rats over mice is faster tumor growth kinetics.
From the same amount of starting material, Yeshi says, a rat tumor may reach 30,000 mm3 versus only 2000 to 3000 mm3 in mice. This means that within a single generation, researchers have enough material to begin efficacy assays in the rat that would take three to four generations in the mouse.
“When you are doing your efficacy studies from that one transplant into the rat, genetic drift-wise and tumor heterogeneity-wise, you’re going to be a lot better off compared to the patient sample than if you had passaged for several generations in mice,” he explains.
And beginning studies on a much earlier passage also increases the likelihood that the stroma surrounding the implanted tumor remains that of the original patient and has not yet been replaced by the host stroma. This can be critically important in ensuring that tumor response to treatment more closely parallels what would happen in the human patient.
“In the drug development pipeline,” Yeshi continues, “the safety, PK [pharmacokinetic] and toxicity studies are all done in the rat right now, whereas efficacy gets done in the mouse. So you’re generating data from different sets of species.”
In contrast, he argues, if you can generate your efficacy data in the rat, as well, then your data set is going to be more cohesive.
And again, the small size of the mouse can be a problem when studies rely on invasive tests such as blood collection timepoints.
“In a study when you’re drawing blood for a clinical pathologist, it is a terminal blood draw,” Yeshi offers. “So, let’s say in a 14-day study, you draw blood from a set of animals on day 7 and another set at day 10 and day 14, these are all going to be different animals from which you’re drawing blood, and then compiling the data.”
The rat, however, offers ten times the volume of blood, and so blood samples are drawn from the same set of animals, potentially providing higher quality data.
“Not only that, but to extend it out further, we’re doing some studies here where the PK, toxicity and efficacy studies are all being done in the same cohort of animals,” he says. “So all of this data is not only from the same species, but from the same cohort of animals, which again means your data quality is going to be much higher and you’re going to have much higher confidence in what your data says.”
Room for all
Obviously, no single model can address all aspects of drug development in oncology. In many ways, to offset the heterogeneity and genetic drift inherent in tumors, companies and researchers are being forced to take a multipronged approach to characterization and screening.
Charles River Laboratories, for one, has initiated collaborations and partnerships with companies like Ocello and InSphero to reveal the synergies between the PDX and 3D culture approaches.
“I think this is going to be very beneficial for the PDX platform moving forward because it’s going to give us a real feel for how well T cells, once they’re activated, will penetrate into the tumor, how quickly they’ll release chemokines and cytokines, which will affect the tumor growth,” says O’Rourke. “And the only way that we can really see that directly in tumors is by using the organoid cultures.”
He envisions service offerings this year that will allow people to screen compounds against spheroids generated from PDX models.
“You may be able to get some type of tumor that is specific to your therapy or targeted toward your therapy, and through our PDX collection, you can select that line, we can generate the 3D spheroids using the information we’ve got from InSphero and Ocello and actually screen multiple compounds against that background,” O'Rourke offers.
“What we’re hoping that will do is fast track the selection of compounds to go into the appropriate in-vivo model and increase the likelihood that you’ll see a beneficial response in terms of efficacy whenever you then run the in-vivo study itself,” he adds.
Co-clinical trials will likely put this idea to the test.
Is location everything?
When thinking of xenograft models, one traditionally envisions a mouse with a bulbous growth on its back, the result of a tumor that has been implanted subcutaneously. To a large extent, the location is a matter of ease of implantation and ease of analysis.
Simply put, the researcher can monitor the size of the tumor in response to treatment and easily access it as he or she passages the tumor to subsequent generations of mice.
More recently, however, there has been growing interest in orthotopic implantation, engraftment of the tumor sample into its tissue of origin within the test subject (as opposed to heterotopic implantation).
To measure tumor size in orthotopic xenografts of cell lines, says Tseten Yeshi, vice president of research and development at Hera Biolabs, researchers typically modify the cell line with something like a luciferase reporter. This allows them to do in-vivo imaging to accurately measure treatment efficacy.
But with patient-derived lines, he continues, modifying the patient material to express a marker would largely defeat the whole purpose of working with patient samples.
“We have not done orthotopics with patient material for that reason, but there are new in-vivo imaging technologies emerging that might allow in-vivo measurements with unmarked cells or tumors,” he enthuses. “It would be exciting if we could utilize those and do some orthotopic work.”
For Azusa Tanaka, product manager for Precision Research Models at Taconic, another significant consideration for choosing between orthotopic and heterotopic models is the anatomical location of the original tumor.
She gives the example of a brain tumor, suggesting researchers need to consider how challenging such microsurgeries would be in terms of both technical expertise and time, particularly when the goal is large drug screens.
“I think that is one of the reasons why many people are doing subcutaneous,” she offers, “because it’s easy and quick, and [you] do all that screening, and perhaps maybe after you have identified candidates, you maybe want to go to more complex and time-consuming models such as orthotopic.”
Despite these challenges, however, the rationale behind orthotopic implantation is strong.
“The way that I look at it is if you implant a tumor into the tissue where it was originally excised from, you’re putting it back into its specific architecture,” says Martin O’Rourke, senior director of the Oncology In Vitro Biosciences Discovery operation at Charles River Laboratories. “So, if it’s a very muscle-rich tumor and you’re putting it back into a muscle bed, it should develop and take on the infiltrates in the way that those tumors do.”
The second reason he describes as more of a fundamental drug discovery question. He gives the example of trying to treat a glioblastoma implanted subcutaneously as a PDX.
The easily accessible subcutaneous tumor, he suggests, doesn’t really reflect the restrictions that surround glioblastomas in their natural environment where, among other things, therapeutic agents would need to cross the blood-brain barrier. Thus, it is less likely that PDX results would translate to human patients.
“The final thing I think that’s important is that by putting them into their natural location, you’re more likely to develop metastatic disease from those tissue banks,” O’Rourke continues. “If you do a prostate implant or breast cancer implant into their natural beds, you tend to see the tumor develops in a significantly different way and metastasizes to clinically relevant tissue types. It’ll go into lymph nodes, go into lung, and I think that’s something that is quite important.”
“If you wanted to not only look at the effect [of prospective therapy] on the primary tumor, which may be minimal, but also look at the reduction of metastatic development, I think that becomes more relevant if the tumor is implanted in its natural tissue of origin,” he concludes.
Recently, AntiCancer’s Robert Hoffman and colleagues examined the molecular impacts of orthotopic and heterotopic implantation in PDX models of pancreatic cancer: PDOX and PDHX, respectively.
Starting with tumors from 25 patients, the researchers managed to establish 13 simultaneous PDOX and PDHX models. They then examined immunohistochemistry, SNP and DNA methylation profiles and metabolite levels in four of these pairs.
“The majority of the molecular-genetic characteristics of the original patient tumors are maintained in the PDOX and PDHX models,” the authors concluded. “Many metabolites were maintained in the PDOX and PDHX models with respect to the original patient tumor, confirming the possibility of conducting accurate metabolic analysis of human tumors using PDX models.”
Despite the significant overlap in tumor tissues, and echoing O’Rourke’s comment above, the researchers acknowledged that PDOX models reflect the metastatic nature of the original patient, whereas PDHX tumors do not metastasize.
“A specific tumor microenvironment may therefore have a great influence on metastasis, rendering the subcutaneous site metastatic resistant,” they suggested. “The relationship of DNA hypermethylation and metastasis will be a subject of future experiments.”
Given these confounding results, location may not be everything, but there is little doubt in most researchers’ minds that it is something.