Useful News for You

A group of scientists, including several at Harvard, have dived deeper into the mammalian brain than ever before by categorizing and mapping at the molecular level all of its thousands of different cell types.

The researchers reported their work in Nature, through a series of 10 papers — six with Harvard affiliations. It’s part of the National Institutes of Health’s Brain Research Through Advancing Innovative Neurotechnologies initiative, which so far has focused on mice; future phases will shift to humans and other primates.

Mammal brains house billions of cells, each defined by the genes they express. This complexity is why true understanding of many brain functions, including molecular mechanisms that underlie neurological diseases, remains so elusive.

To create the first molecularly defined cell atlas of the whole mouse brain, a team led by Harvard’s Xiaowei Zhuang identified and spatially mapped thousands of unique cell types, most of which had never previously been characterized.

“We identified 5,000 transcriptionally distinct cell populations,” said Zhuang, the David B. Arnold Professor of Science and a Howard Hughes medical investigator. “Suffice it to say that the level of diversity we identified is really extraordinary.”

The brain-wide atlas of cell types cataloging cells, their distribution, and interactions could serve as a starting point for scientists studying certain brain functions or diseases. Someday the basic outlines of the atlas could be applied to the human brain, 1,000 times larger than the mouse brain.

“It gives me real excitement to see things that were not visible before. I am also thrilled when our technology is used by so many labs,” said Zhuang, referring to Multiplexed Error-Robust Fluorescence in situ Hybridization (MERFISH), a genome-scale imaging technology developed in her lab.

In collaboration with scientists at the Allen Institute for Brain Science, Zhuang and her team used MERFISH together with single-cell RNA sequencing data to not only identify each cell type, but to image them in situ. Their work provides new information about the molecular signatures of these cell types, as well as where they are located in the brain. The result is a stunningly detailed picture of the mouse brain’s full complement of cells, their gene-expression activity, and their spatial relationships.

In their Nature paper, the researchers used MERFISH to determine gene-expression profiles of approximately 10 million cells by imaging a panel of 1,100 genes, selected using single-cell RNA sequencing data provided by Allen Institute collaborators.

Retina findings could boost glaucoma research

In a separate paper in the Nature series, Joshua Sanes, the Jeff C. Tarr Professor of Molecular and Cellular Biology, co-led a team that captured new insights into the evolutionary history of the vertebrate retina.

A part of the brain encased in the eye, the retina boasts complex neural circuits that receive visual information, which they then transmit to the rest of the brain for further processing. The retina is functionally very different from species to species — for example, human hunter-gatherers evolved sharp daytime vision, whereas mice possess better night vision than humans do; some animals see in color, while others see predominantly in black and white.

But at molecular levels, how different are retinas, really? Sanes, in collaboration with researchers at the University of California, Berkeley, and the Broad Institute, performed a new comparative analysis of retinal cell types across 17 species, including humans, fish, mice, and opossums. Using single-cell RNA sequencing, which allowed them to differentiate types of retinal cells by their genetic expression profiles, the researchers’ findings upended some long-held views about how certain species’ visual systems evolved.

One striking discovery involved so-called “midget retinal ganglion cells,” which, in humans, carry 90 percent of the information from the eye to the brain. These cells give humans their fine-detail vision, and changes to them are associated with eye diseases such as glaucoma. No related cells had ever been found in mice, so they had been assumed to be unique to primates.

In their analysis, Sanes and team identified for the first time clear relatives of midget retinal ganglion cells in many other species, including mice, albeit in much smaller proportions. Since mice are a common model animal to study glaucoma, being able to pinpoint these cells is a potentially crucial insight.

“I think we can make a very compelling case that if you want to study these important human retinal ganglion cells in a mouse, these are the cells you want to be studying,” Sanes said.

Other Harvard-affiliated researchers, at Harvard Medical School, Boston Children’s Hospital, and the Broad, also contributed findings to the NIH’s cell census network, including a molecular cytoarchitecture of the adult mouse brain, and a transcriptomic taxonomy of mouse brain-wide spinal projecting neurons.

Harvard researchers have realized a key milestone in the quest for stable, scalable quantum computing, an ultra-high-speed technology that will enable game-changing advances in a variety of fields, including medicine, science, and finance.

The team, led by Mikhail Lukin, the Joshua and Beth Friedman University Professor in physics and co-director of the Harvard Quantum Initiative, has created the first programmable, logical quantum processor, capable of encoding up to 48 logical qubits, and executing hundreds of logical gate operations, a vast improvement over prior efforts.

Published in Nature, the work was performed in collaboration with Markus Greiner, the George Vasmer Leverett Professor of Physics; colleagues from MIT; and QuEra Computing, a Boston company founded on technology from Harvard labs.

The system is the first demonstration of large-scale algorithm execution on an error-corrected quantum computer, heralding the advent of early fault-tolerant, or reliably uninterrupted, quantum computation.

Lukin described the achievement as a possible inflection point akin to the early days in the field of artificial intelligence: the ideas of quantum error correction and fault tolerance, long theorized, are starting to bear fruit.

“I think this is one of the moments in which it is clear that something very special is coming,” Lukin said. “Although there are still challenges ahead, we expect that this new advance will greatly accelerate the progress toward large-scale, useful quantum computers.”

Denise Caldwell of the National Science Foundation agrees.

“This breakthrough is a tour de force of quantum engineering and design,” said Caldwell, acting assistant director of the Mathematical and Physical Sciences Directorate, which supported the research through NSF’s Physics Frontiers Centers and Quantum Leap Challenge Institutes programs. “The team has not only accelerated the development of quantum information processing by using neutral atoms, but opened a new door to explorations of large-scale logical qubit devices, which could enable transformative benefits for science and society as a whole.”

It’s been a long, complex path.

In quantum computing, a quantum bit or “qubit” is one unit of information, just like a binary bit in classical computing. For more than two decades, physicists and engineers have shown the world that quantum computing is, in principle, possible by manipulating quantum particles — be they atoms, ions, or photons — to create physical qubits.

But successfully exploiting the weirdness of quantum mechanics for computation is more complicated than simply amassing a large-enough number of qubits, which are inherently unstable and prone to collapse out of their quantum states.

The real coins of the realm are so-called logical qubits: bundles of redundant, error-corrected physical qubits, which can store information for use in a quantum algorithm. Creating logical qubits as controllable units — like classical bits — has been a fundamental obstacle for the field, and it’s generally accepted that until quantum computers can run reliably on logical qubits, the technology can’t really take off.

To date, the best computing systems have demonstrated one or two logical qubits, and one quantum gate operation — akin to just one unit of code — between them.

The Harvard team’s breakthrough builds on several years of work on a quantum computing architecture known as a neutral atom array, pioneered in Lukin’s lab. It is now being commercialized by QuEra, which recently entered into a licensing agreement with Harvard’s Office of Technology Development for a patent portfolio based on innovations developed by Lukin’s group.

The key component of the system is a block of ultra-cold, suspended rubidium atoms, in which the atoms — the system’s physical qubits — can move about and be connected into pairs — or “entangled” — mid-computation.

Entangled pairs of atoms form gates, which are units of computing power. Previously, the team had demonstrated low error rates in their entangling operations, proving the reliability of their neutral atom array system.

With their logical quantum processor, the researchers now demonstrate parallel, multiplexed control of an entire patch of logical qubits, using lasers. This result is more efficient and scalable than having to control individual physical qubits.

“We are trying to mark a transition in the field, toward starting to test algorithms with error-corrected qubits instead of physical ones, and enabling a path toward larger devices,” said paper first author Dolev Bluvstein, a Griffin School of Arts and Sciences Ph.D. student in Lukin’s lab.

The team will continue to work toward demonstrating more types of operations on their 48 logical qubits and to configure their system to run continuously, as opposed to manual cycling as it does now.

The work was supported by the Defense Advanced Research Projects Agency through the Optimization with Noisy Intermediate-Scale Quantum devices program; the Center for Ultracold Atoms, a National Science Foundation Physics Frontiers Center; the Army Research Office; and QuEra Computing.