No weak links: A W&M-based collaborative advances understanding of neural control of respiration
Patch-Seq is shorthand for “patch-clamp, followed by next-generation sequencing.” It’s a collaborative procedure that’s only been performed in a few labs. Christopher Del Negro says there’s a good reason for that.
“You have to have a team that is extremely highly skilled in absolutely every aspect because if you have one weak link — one weak link! — it all fails,” he said.
Del Negro is a professor in William & Mary’s Department of Applied Science. He is part of an interdisciplinary team that successfully used Patch-Seq to tease apart the physiological activity of two sets of neurons that are important in the generation of respiration.
The successful Patch-Seq work has produced a more detailed understanding of the neural control of respiration. The discovery has important and relevant implications for sleep apnea and other respiratory ailments with neural origins.
The team was led by Prajkta Kallurkar, a graduate student in Del Negro’s lab who received her Ph.D. in 2021. Other members of the team from the Department of Applied Science are Tina Picardo, a senior scientist and Greg Conradi Smith, a professor. Margaret Saha, Chancellor Professor of Biology, and Yae Sugimura, of the Jikei University School of Medicine in Tokyo, round out the team.
Kallurkar explained that the Del Negro lab investigates the neural basis of breathing. Respiration is controlled by a surprisingly small area of the brainstem known as the pre-Bötzinger complex. Small as it is, the pre-Bötzinger complex is packed full of neurons, and the lab wanted to identify the classes of neurons that were responsible for generating rhythm and motor pattern of respiration.
She said that their investigation requires a coordinated series of operations, rather than a single technique. Patch-Seq begins with recording the electrophysiological properties of the neurons within the target area of the pre-Bötzinger complex responsible for inspiration, the intake of breath. Sugimura flew over from Japan and trained Kallurkar on how to successfully extract a neuron and she helped the team set-up the recording and extraction process of Patch-Seq in less than a month.
“Once we have recorded the electrophysiological properties, we then extract all the cellular contents, and then we perform downstream RNA sequencing,” Kallurkar said.
The measurement of electrophysiological properties, a technique known as “whole-cell patch recording,” and cellular extraction were performed using robotic-guided pipettes, necessary when working with targets as small as a single neuron. The team sent their samples out for sequencing, but first the cellular contents — the RNA, painstakingly pipetted from the neurons — is reverse transcribed into complementary DNA by Picardo.
“We are actually targeting the RNA of the sample,” she said. “So, I employ some molecular-biological techniques to generate complementary DNA, and then make a library out of it.
The process sounds straightforward, even simple, but Kallurkar and Saha point out the challenges Picardo faced in her part of the project, a procedure known as reverse transcription and library construction. Many of the challenges stem from the exceptionally small amount of starting material.
“What Tina did is to amplify less than a picogram of starting mRNA into a high-quality cDNA library,” Kallurkar said.
Picardo acknowledged that a high sample failure rate comes with such a painstaking procedure involving such a small quantity of material: 18 successful amplifications out of around 150 samples.
“This is incredibly difficult,” Saha said. “Tina is dealing with so little material — and the RNA degrades when you blink! And she does all of the quality control on this herself before it ever goes for sequencing. So this is exceptionally difficult, and an incredible achievement.”
The next in the series of Patch-Seq steps is sequencing. Picardo’s processed samples were sent to an outside lab that “reads” the nucleotides in each cell’s cDNA library and then assembles those reads into text files, which are sent back to the W&M team.
The data-science duo of Saha and Conradi Smith helped guide the team throughout the study. They played a more prominent role in working with the data that came back from the sequencing lab, unraveling which genes are represented in the sequences and which ones are differentially expressed.
Kallurkar said that the data-science team zeroed in on the relevant genes in the two target pre-Bötzinger neuron classes, labeled Type-1 and Type-2. As their animal model is the mouse, the data science work begins with a mouse reference genome.
The data analysts use mapping software to compare a particular nucleotide sequence from their sample maps to a known mouse gene. This is a “big data” problem, tackled using genomic data science and the W&M supercomputer cluster. Kallurkar says the mouse reference genome contains about 55,000 genes, some 33,000 of which are expressed in the group’s samples.
“For perspective, we have more than 33,000 genes in an individual single cell, and we have 17 cells,” Kallurkar explained. “So our matrix is 33,000 by 17.”
She said that within that matrix, the researchers are looking for genes that belong to two sets of neurons that have different electrical properties. The samples contained one batch of seven neurons and a second of nine neurons.
“So, now, we want to know is: Out of this 33,000, how many are differentially expressed between these two sets of neurons?” Kallurkar said. She explained that if a particular gene is highly expressed in one type of cell, and it is not expressed in the second, that shows the gene is important for this one particular type of cell, “and we should go after that particular gene in order to target that population”
“The data science side of it,” Conradi Smith commented, “is a heavy lift. Neural transcriptome studies are big data par excellence. Prajkta’s success required proficiency with scripting in Python, familiarity with several sophisticated bioinformatics software packages, and deep understanding of statistical theory.”
The Patch-Seq technique was developed around five years ago, and Del Negro said only 25 studies have used the technique.
“And of those 25 studies, only two other studies have done it from a population that has a known behavioral function,” he added. “And those were pancreatic islet cells and cells of the retina. Everybody knows this population of neurons generates the breathing rhythm and expresses the initial stages of the breathing motor pattern. I think that's another important aspect of this work.”
The collaborators have a paper in preprint on the Cold Spring Harbor Lab’s bioRxiv server. Del Negro says the work is so novel and so cross-disciplinary that it’s a challenge to find the right peer-reviewed journal for it.
“Patch-Seq is still a somewhat new technique,” he said. “And we’re shopping it around, trying to find the right venue. I think we just haven’t found the right audience yet.”
Del Negro adds that the focus of the study — Type-1 and Type-2 neurons— are subsets of a larger class known as Dbx1 — was discovered in 2010 by Picardo, and was the topic of her doctoral dissertation.
“This is a follow up of a decade long effort to understand this singular population with a known behavioral function,” Del Negro said. “And it's a thread moving through decades.”