Sick Papes Interview with Luke Gilbert!
We recently received the following letter:
Dear Sick Papes,
I am a third grader who lives under a rock. Furthermore, I cannot read. Thus, I do not know the difference between CRISPR, CRISPRa and CRISPRi. Can you please help?
You are in luck, reader! Sick Papes recently sat down with one of the world’s most qualified people to answer this all-too-common question. Dr. Luke Gilbert has spent the past few years as a post-doc in Jonathan Weissman’s lab at UCSF, a notorious “scientific paradise,” during which time he has been going, quite literally, buck wild. Following a now-classic pape led by Stanley Qi, showing that catalytically-dead Cas9 can be used to block gene transcription, Luke has published a truly magnificent run of papers demonstrating a wide variety of ways that CRISPR can be used besides just cutting DNA. If I had to describe the addictive and jaw-dropping experience of reading these papers in real time as they came out from 2013-2016, I would say it’s something like listening to Serial, watching Making a Murderer, and reading The Mastermind, all at the same damn time. Please enjoy this educational interview!
(Full disclosure: Luke and his co-authors are on some patents for CRISPR-related techniques).
Figure 1. Effective and highly specific knockdown of GFP using CRISPRi (see below for details). From: Gilbert et al. 2013, Cell 154).
Sick Papes: Hi Luke! What is CRISPRa and CRISPRi?
Dr. Luke Gilbert: You can think of Cas9 nuclease as a pair of programmable molecular scissors. Our work started when we decided to break the scissors, creating a nuclease inactive form of Cas9, which we call “dead Cas9” or dCas9, and which be used as a programmable DNA binding protein in both prokaryotes and eukaryotes. Why would anyone want a dead Cas9? Because we can use it to recruit any protein to any region of a genome in any organism to do things like control transcription, edit the epigenome, or image the genome in living cells [e.g. by bringing a fluorophore to a specific genomic location].
I have spent most of my time as a postdoc in Jonathan Weissman’s lab showing that we can turn any gene on and off at will by using dCas9 to recruit different protein domains that activate (CRISPRa) or inactivate (CRISPRi) transcription. Another way to describe this work is that we have created a set of programmable synthetic transcription factors.
Synthetic transcription factors are not new - if you want to hear me drone on and on and on (seriously I talked for 2 hours) about the history of synthetic transcription factors you can click here. Or if you want to hear Jonathan explain what we did in a way more coherent way click here. [editor’s note: these talks are great, and give a deep history on research into transcription.]
SP: One of the sweet things about CRISPR is that has opened up entirely new ways to perform genetic screens, where you interrogate the function of every gene in the genome. Can you explain how this field possibly developed so fast, and what benefits CRISPR offers over existing strategies?
LG: Genetic screens have a long, illustrious history in biology. Traditionally, people have done screens in model organisms using mutagenesis or transposons or have created knockout collections. In human cells, genetic screens really took off in the early to mid-2000s when RNAi screens were invented. So CRISPR screening platforms really owe a ton to the ~10 years of work using RNAi for genetic screens in human cells. This is reflected by how quickly people went from CRISPRing one gene to CRISPRing the whole genome - the first CRISPR systems were used in human cells in 2013, and within a year several groups had already created genome scale libraries. This was possible because biologists already knew how to construct a pooled library, do a screen (don’t forget the negative controls!), process and sequence screen samples, and analyze the data. I do not want to get into the relative merits of CRISPR, RNAi and CRISPRi as this conversation would literally go on for like two days [editor’s note: Dr. Gilbert politely declined our follow-up request to have a two-day conversation about this at a crystal-strewn clearing betwixt a waterfall and a lunar eclipse]. I think CRISPR and CRISPRi are new methods – in five years we’ll see what people like best.
Figure 2. Using a tiling array to determine the most effective window for sgRNA placement for CRISPRi. TSS = transcription start site. From Gilbert et al 2014, Cell 159
SP: In cell culture, genetic screens can be done in two ways: “arrayed,” where you perturb each gene in it’s own well, or “pooled,” where you just have a single large flask of cells, each with a single gene perturbed, and you measure the outcome of some biological challenge via sequencing. Can you tell us more about just how “cheap” and “easy” these really are?
LG: Arrayed screens are great, but they are expensive because they require robots and more reagents [editor’s note: for example, to study 10,000 genes in an arrayed screen would require 27 384-well plates for one replicate]. Thus, genome scale pooled screens are awesome because they are cheap and relatively easy. All the cells are grown together, and we use high throughput sequencing to measure how every gene included in a library controls response to a perturbation by measuring the relative abundance of each sgRNA in the library at the start and end of an experiment.
In our lab one person can do 4-8 screens at the same time because we use non-adherant leukemia cell lines that are relatively easy to passage. A pooled screen from start to finish takes about a month including sample prep post screen and sequencing and each sample costs roughly $5,000 including serum costs, sample prep and sequencing. Most people either do duplicate or triplicate screens to estimate your technical noise in a screen.
There are two main types of screens you can do in human cells, typically referred to as “positive selection screens” and “negative selection screens.” Positive selection screens [where you identify those mutants that survive some life-threatening insult] reveal genes that control resistance to a process, while negative selection screens [where you see which mutants disappear following a challenge] reveal genes required for a process. However, I don’ really think about screens as positive or negative because we do our screens in a way that reveals genes that confer both resistance and sensitivity to a process. To do this we choose to apply a measured or modest amount of selective pressure that enables the detection of genes that confer resistance and sensitivity to a process or drug.
SP: I’ve heard people draw a distinction between “technique building” and “biology” as if the two were somehow separable. I personally think this is a moronic comparison because new techniques come from biology, and techniques are what we use to do biology. What do you think?
LG: I do think there is a distinction between methods development and biology but agree that they are inseparable. I think as long as you are doing innovative science for the right long term reasons the rest takes care of it self. If you do methods development you often end up doing types of biology that you did not expect to when you started [editor’s note: and vice versa, e.g. CRISPR itself]. For example, I never thought I would be working on transcription or synthetic biology when I joined Jonathan’s lab. Or another example, many labs have now studied the biology of the Cas9 protein in vitro and in vivo: structure of the Cas9/sgRNA/DNA complex, mechanism of action, 2D vs 3D search mechanisms, fidelity of target DNA recognition. This could easily be described as either methods development or biology.
I feel like we have just started to use CRISPR for biology. Much of the published CRISPR work has been proof-of-concept or methods development. I think we are just now at a point where doing an experiment using CRISPR reagents is not a paper in and of itself but rather just another way to ask new questions about biology.
SP: What are the big questions that you’re interested in right now, that you think these tiling screens are allowing you to study in new ways?
LG: Tiling screens are very cool. In our published work, we used CRISPRi/a tiling screens to define rules for how to use our artificial transcription factors to control transcription most effectively and to derive rules about CRISPRi/a specificity using mismatch sgRNA libraries (Figure 2).
I think tiling screens using either Cas9 or dCas9 will be incredibly useful for mapping enhancer and promoter elements. A really cool example of this was published last year by Feng Zhang, Stuart Orkin and Daniel Bauer. I am also a huge fan of Chris Vakoc’s tiling screens. He showed you can functionally map the domain architecture of a protein using Cas9. I could see a two-tiered approach where you tile all coding exons of a gene to identify domains required for protein function and then use Jay Shendure’s saturation mutagenesis CRISPR approach to dive deep into the biology of a region of a protein you are interested in being very useful for many types of biology.
SP: You have a background in cancer biology, right? How did this lead you to CRISPRa/i?
LG: Hah – good question. I joined Jonathan’s lab to study a new type of anticancer drug that targets protein folding, but CRISPR seemed awesome and so I got side-tracked. I am a cancer biologist, though and have managed to talk my way into a faculty position at UCSF in the Helen Diller Cancer Center where I will be studying cancer in cancer cells and mouse models of cancer. [editor’s note: hell yeah congratulations!].
SP: Any particularly papes you think we should all be reading right now?
LG: I will make a shameless plug for my partner-in-crime Max Horlbeck’s paper in eLife that just came out showing Cas9 and dCas9 access to DNA is impeded by nucleosomes in vitro and in vivo (Figure 3). This work was a collaboration with Lea Witkowsky in Robert Tjian’s lab. I love this paper because they used an amazingly diverse set of methods from machine learning to hard core biochemistry to show Cas9 has a problem with nucleosomes. Makes sense to me as SpCas9 never saw a nucleosome until just a couple of years ago [editor’s note: whoa].
Figure 3. sgRNA effectiveness is anti-correlated with nucleosome position, suggesting that dCas9 cannot bind nucleosome-bound DNA. From Horlbeck et al 2016, eLife.
But beyond this I would say that I have always benefited enormously from reading as broadly and as much as I can. It does not matter what you read just read constantly.
(Top Image Credit: Gilbert et al 2013 Cell)