Advancing Novel Ideas through Harvard Catalyst's Reactor

Two investigators advance to Reactor's second stage, "Incubator"

Why participate?

  • Researchers have the opportunity to develop innovative technologies and methodologies as part of a community that fosters collaboration across the hospitals and Harvard schools. As investigators work on their projects, they receive assistance throughout Reactor's Innovator and Incubator stages.

Who benefits?

  • Harvard-affiliated researchers who are interested in collaborating with interdisciplinary and cross-institutional investigators.

Harvard Catalyst's Reactor Program offers opportunities designed to assist researchers in crystallizing and expanding their ideas, and ultimately translating them into solutions that will impact human health. There may not be a sure-fire recipe for successful translation of a research idea into a solution that benefits patients, but there are key ingredients: access to new technologies, financial support, scientific guidance, and project management to support project ideas. Investigators begin with the first stage, Innovator, which focuses on pilot funding, and several may move on to Incubator, which stresses team building and collaboration.

"We're trying to create a supportive environment that enables more successful translational outcomes," says Gary Gray, PhD, director of technology and innovation for Harvard Catalyst's Reactor Program. To establish and grow clinical and translational research communities, Reactor launches several initiatives and grant funding opportunities each year. In a recent initiative, Reactor funded 11 Innovator projects [PDF] , in collaboration with the Harvard Center for Biological Imaging (HCBI). The HCBI—under the direction of Jeff Lichtman, MD, PhD, and Douglas Richardson, PhD—is a renowned research facility that's home to the most advanced microscopes and microscopy expertise in the world. For this opportunity, the goal was to have research pathologists attain access to advanced microscopes to generate new ideas that would impact patient care.

After one year, Reactor's internal and external review committees selected two teams to advance to Reactor's Incubator phase, with each receiving an additional $150,000 per year for up to two years. "The teams made significant progress during the year, and their projects embody multi-disciplinary, innovative research that addresses clear biomedical needs," Gray says. Both projects are novel methods that support cancer diagnoses and risk assessments. "We anticipate the additional funding will lead to important clinical proofs of concept," Gray adds.

New Technique for Tumor Genotyping

Nucleus of cultured circulating tumor cell hybridized with a panel of ten different genes labeled with specific color combinations ('bar-coded'). This image is part of the development of a clinical grade high-throughput DNA-FISH platform that will enable the automated copy number analysis of large panels of genes in formalin-fixed paraffin embedded tumor biopsy samples and in isolated circulating tumor cells.

John Iafrate, MD, PhD, a pathologist and medical director at the Center for Integrated Diagnostics at Massachusetts General Hospital (MGH), has already successfully brought new genetic technologies to cancer diagnosis. His contributions to genetic fingerprinting of common tumor mutations over the past eight years have guided targeted therapies and become part of routine care for most advanced cancer patients at MGH and beyond.

The primary focus of Iafrate's research program is to create new genotyping assays that will better screen for copy number variations in genes known to be major drivers of cancers. There are important drugs based on copy number assessment such as for breast cancer driven by HER2 gene amplification.

Iafrate's MGH cancer genotyping platform, called SNAPSHOT, uses next generation sequencing (NGS) to rapidly identify known genetic mutations on tumor samples. "But currently there really is no optimal algorithm to analyze copy number for NGS platforms," says Iafrate, who is also a professor of pathology at Harvard Medical School (HMS).

"Harvard Catalyst does what its name says," Iafrate points out.
"It catalyzed our ability to get proof of principle and move things along."

As a complement to his lab's NGS efforts, his project team is using fluorescence in situ hybridization (FISH) techniques to determine gene copy number. FISH allows someone to look under the microscope at each tumor cell, one cell at a time, to see how many copies of a gene there are in any particular cell. Current FISH techniques are limited by standard microscopy to analyzing one to three genes at a time. However, there is such complexity within one tumor, Iafrate says, that there can be many closely related genes amplified in adjacent tumor cells. "The ability to understand heterogeneity, especially copy number heterogeneity, is limited since we have no technology that would allow us to analyze more than three genes at a time," he explains. "We are thus limited in our understanding of tumor biology, and challenged with developing rational drug approaches."

Deciphering Tumor Complexity

John Iafrate, MD, PhD

For his Incubator project, Iafrate and his co-investigator, Maristela Onozato, MD, PhD, are using an advanced fluorescent microscope at the HCBI to engineer a multiplex FISH assay that can rapidly and accurately visualize 30 to 50 genes at a time. Onozato is a research fellow in pathology at MGH. After the year of Reactor support as part of Innovator, Iafrate and Onozato had already developed the capability to examine up to 20 genes at a time.

Iafrate credits the Reactor program with keeping them on track and facilitating the expertise needed in bioinformatics, complex image analysis, and microscopy. Hunter Elliott, PhD, director of the Image and Data Analysis Core (IDAC) at HMS, became such a critical part of the project that he is now a co-investigator. "Harvard Catalyst does what its name says," Iafrate points out. "It catalyzed our ability to get proof of principle and move things along." He is excited about their progress thus far. "I'm hopeful that we'll be able to deliver something that works for patients and is affordable," he says. He also expects this assay will enhance understanding of tumor heterogeneity, one of the big quests in cancer research today.

Early Breast Disease Now in Better Focus

The accurate evaluation and classification of breast lesions represents a huge challenge for pathology because these assessments, based on conventional microscopy and qualitative criteria, haven't changed in decades. Andrew H. Beck, MD, PhD, a computational pathologist and director of bioinformatics at the Cancer Research Institute of Beth Israel Deaconess Medical Center (BIDMC), and his team have set out to see if the latest advances in microscopy, machine learning models, and computation can lead to better classification schemes and predict who's truly at risk of cancer.

3D morphological features of expanded breast tissue nuclei visualized with Lightsheet microscopy (HCBI); obtained by Octavian Bucur, Yongxin Zhao, and Humayun Irshad.

Tissue biopsies are necessary to look at tissue and cells under a microscope to confirm a diagnosis of breast cancer. "Improvements in non-invasive imaging have resulted in an increase in the number of biopsies of borderline lesions, where it's not clear from the clinical and radiology characteristics if the lesion is due to a benign process, a malignant process, or some kind of inflammatory or fibrotic process," Beck says. In these cases, accurate pathological evaluation is especially important to guide clinical care.

But recent studies have shown significant discordance in lesion interpretations among pathologists, especially in borderline cases. This has enormous ramifications. "In these challenging cases, what the pathologist says is perhaps the single most important piece of information that will direct all further management of the patient," Beck explains.

If the pathologist reports a benign proliferation with no atypia then the doctor will tell the patient she is fine. But if the report comes back that it's an intraductal proliferation with atypical or malignant-appearing cells, then that lesion will be treated more aggressively.

Beck, who is an assistant professor of pathology at HMS and an associate member of the Broad Institute, is the principal investigator of the team, which also includes Stuart Schnitt, MD, BIDMC, Rulla Tamimi, ScD, Brigham and Women's Hospital, Benjamin Glass, BS, Andreea Lucia Stancu, MD, and research fellows Octavian Bucur, PhD, and Humayun Irshad, PhD.

"The Reactor Incubator grant really pushed me to try new, advanced microscopy techniques."

In the first year of Reactor support, Beck's team developed a protocol using light sheet microscopy, which produces impressive resolutions in 3D, but had never previously been developed for use on breast pathology specimens. They used the 3D light sheet microscopy and image processing methods to measure new features of breast pathology specimens, which are not currently reported in clinical care. Machine learning and statistics teased out which findings are most clinically useful and they built models to generate the diagnoses.

The platform they created worked, "but we wanted to make it even better," Beck says. "The Reactor Incubator grant really pushed me to try new, advanced microscopy techniques."

Applying Expansion Microscopy to Tumor Cells

With additional program support, Beck's team is now building a platform using expansion microscopy, an innovative way of preparing the tissue sample that makes it expand five-fold without distortion. With it, infinitesimally smaller features are visible. The team is collaborating with Edward S. Boyden, PhD, at MIT, whose research group created the expansion method for studying neurons. Beck's group at HMS and BIDMC in collaboration with Boyden's team at MIT are the first to apply the method to tumor cells.

They are also collaborating with the Nurses Health Study, using their archival tissue specimens. The study has extensive clinical documentation, including pathologist reports related to each specimen. "We have data that suggest that we can train the computer more effectively on these expanded samples than the conventional samples," Beck reports.

They are also building a model to predict risk. They will analyze samples for distinguishing features from participants in the Nurses Health Study who never went on to develop cancer (those at low risk), and compare them to samples from those who were diagnosed with cancer in the following 10 years (those at higher risk).

The end goal will be a commercially available product for breast biopsies. Applying a similar approach to other diseases where better biopsy evaluation may improve treatment decisions, such as with prostate biopsies, will likely be next. "We think the technology and methods we are developing should be useful across a range of solid malignancies," Beck says.

As Reactor continues to offer new funding opportunities through Innovator, the goal is to engage these funded investigators in moving their work to further stages of development, identifying the most promising ideas and offering them the tools, resources, and support that Harvard Catalyst provides.

You can access Reactor from the Programs menu on the Harvard Catalyst website.

Try It