We are searching data for your request:
Upon completion, a link will appear to access the found materials.
How are the techniques: Knock-sideways, knockout & knock-down different?
Hello and welcome Adil Amchi.
In genetics and molecular biology there are differences between the terms knock-out, knock-down and (the lesser known term) knock-sideways.
Knock-sideways: Inactivates proteins. Could be done using small molecule inhibitors.
Knock-out: Gene removal, no gene expression. Could be done using CRISPR.
Knock-down: Gene expression is reduced. Could be done using interference RNA.
(Another reference for Knock-sideways:
"Importantly, the rapidity of the “knock-sideways” system allowed the researchers to observe a phenotype distinct from siRNA-mediated knockdown of the same protein… ")
Gene knockout: Loss of a gene can be compensated by another gene
New methods for modifying the genome are currently widely discussed: Using CRISPR/Cas for instance, scientists can remove parts of the genetic code of a gene, thereby knocking it out. Furthermore, there are ways to inhibit translation of a gene into a protein. Both methods have in common that they impede production of a protein and should therefore have comparable consequences for an organism. However, it has been shown that consequences can differ, after a gene is either knocked, out or only blocked. Scientist from the MPI for Heart and Lung Research in Bad Nauheim now found that additional genes compensate for a knocked out gene and either attenuate consequences or completely compensate deficits. The results suggest caution when interpreting data from molecular biological studies or developing gene therapies to treat various diseases.
To analyse function of an unkown gene, scientists often extinguish the gene and investigate the consequences of this treatment for the organism. To do so, they cut DNA-fragments from the gene using enzymes deleting the genetic information for a functioning protein. Such method is called "Gene knockout." In contrast, in a "gene knockdown" scientists block protein production using particular substances, e.g. microRNAs.
Recent studies, however, have shown that results may vary between knockout- and knockdown animals. Scientists from Didier Stainier's group at the Max Planck Institute for Heart and Lung Research have now identified the reason for this. The Bad Nauheim based researchers have investigated a gene called egfl7 in zebrafish. The gene is involved in the production of connective tissue in blood vessel walls, thereby stabilizing them. Doing so, egfl7 regulates blood vessel growth.
Developmental biologists, however, are not sure, what happens in a fish organism, after the egfl7 gene has been deleted. "If the gene has been blocked in a knockdown, blood vessels do not develop normally," explains Andrea Rossi, together with Zacharias Kontarakis first author of the study. In contrast, if the gene itself is deleted by a genetic manipulation, blood vessel growth is not affected.
In the beginning, Max Planck researchers excluded potential side effects of the knockdown substance being responsible for interference in vascular development. To this end, they injected the substance into fish larvae in which the egfl7 gene had already been deleted. However, the larvae almost developed normally.
"Since the substance did not cause disturbances in blood vessel growth, we thought of a different mechanism: The gene loss could be compensated by another gene taking over the function," Kontarakis says. "Therefore, we were looking for rescue genes, which might have been produced in animals without a functional egfl7 gene."
The researchers compared the mRNA molecules and proteins in fish with or without a functional egfl7 gene and detected several mRNAs and proteins being present in higher amounts in fish without egfl7. An example is emilin 3B. When "knockdown" animals are treated with emilin 3B after egfl7 has been blocked, blood vessels develop almost normally. "This tells us that emilin 3B can compensate for the loss of egfl7. In egfl7 knockout fish, emilin production is getting upregulated. This is not the case in knockdown fish," Stainier explains.
As the next step, the group plans to analyse how genes "know" that another gene has been deleted and then compensate for the loss. Several researchers worldwide are trying to delete disease genes for therapeutic reasons. Before we establish such therapies, we have to fully understand the consequences the loss or blockade of a gene might have. "In addition, our study illustrates the power of comparing knockouts and knockdowns to identify modifier genes, a goal that remains a major challenge in the field of human genetics" says Stainier.
RNAi knockdown versus DNA knock-out - (Jul/28/2005 )
I'm new in this field and am hoping someone can explain to me why I would want to use RNAi to create knockdown mice over knockout mice.
So far I can understand why short-term RNAi using siRNA is beneficial because the results are fast observed and temporary. But the process in creating a knockdown mice seems almost the same as when creating a knockout mice. so y bother?
my understanding of the issue is that some genes cannot be knocked-out. or if you do knock out the gene in your mice they become very sickly and will not breed, which makes the long process of creating a knockout not particularly helpful. i haven't used RNAi in mice, so until you commented i figured it would be a lot easier than designing and breeding knockout mice.
Wow, I didn't even know that it was possible to have knock-down mice. I wouldn't really do that anyway because if your silenced protein is crucial for the cells it is likely that you will counterselect them or you may select cells that have found a way to bypass your silenced protein or even more cells mutated for RNAi processing. Am I right? For me looks dangerous even more with off-target effects.
Knock-in and knockout mice are both genetically altered mice that assist researchers in understanding the genetic functions of the human body at a greater level of detail. While doing genetic research directly on humans and modifying genes in human tissues is still not an ethical, viable option, genetically modified mice can present researchers with a window of opportunity that other models may not offer. This is due to the fact that the mouse genome – albeit far simpler – is extremely similar to the human genome. It contains many of the same genes that scientists would need to knock out or knock in, depending on their research goals. Despite both methods being used for genetic research, that’s pretty much where the similarities between knock-in and knockout mice end. In fact, the processes of generating both types of model are quite different. Moreover, the entire approach to the study of knockout mice is very different compared to that of knock-in mice, as outlined below.
The most important difference between the two types of models is that, in the case of knockout mice, a gene is targeted and inactivated, or “knocked out.” On the other hand, generating knock-in mice involves the opposite technique: altering the mouse’s genetic sequence in order to add foreign genetic material in the form of a new gene housed by the specific locus targeted by the researcher. Knockout mice are far older and more vastly researched when compared to knock-in models. The first knockout mouse was generated in 1989, while generating knock-in mice is a process that has only been perfected during the past few years. However, this technique has been growing in popularity at an accelerated rate, as the methods of obtaining knock-in mice continue to increase in number.
Knock-in and knockout technologies are very different. While both can be used on laboratory mice to facilitate genetic research without having to experiment on human subjects, the technology and the specific methods used to delete or inactivate a portion of the DNA sequence are distinct and recognized as such by researchers who are familiar with both gene knockouts and knock-ins. Knock-ins create a one-for-one substitution of the DNA sequence within the genome. In particular, knock-in technology is focused on a specific locus of the sequence, compared to how knockout mouse generating technologies and methods focus on engineering an entirely new DNA sequence, which is very similar to the original one, but made to ensure that the gene becomes inoperable.
It’s also important to take note of what both techniques are designed to achieve. Their uses vary in significant ways, to the extent that researchers tend to use each for different types of studies. Knockout mice are designed to help scientists peer into the details of what makes genes work and what each gene does. Because they are able to stop specific genes from fulfilling their functions, they can be used to study the impact of a specific gene on the entire body’s various functions. This will help medical and genetic researchers study genetic diseases more easily, along with various disorders that are affected by the impairment of a certain gene. In contrast, knock-ins help to study the human genome more closely by introducing human genetic material into a mouse’s genome. The possibilities are endless, but most of the progress done by genetic scientists has been in the field of immunodeficiency and stem cell research, as well as in studying various diseases that required the activation of human genes.
Gene knockdown vs gene knockout vs knocksideways? - Biology
What are the pros and cons of shRNA knockdown vs. CRISPR knockout?
Either shRNA-mediated knockdown or nuclease-mediated knockout (e.g. CRISPR or TALEN) can be valuable experimental approach to study the loss-of-function effects of a gene of interest in cell culture. In order to decide which method is optimal for your specific application, there are a few things you should consider.
Knockdown vectors express short hairpin RNAs (shRNAs) that repress the function of target mRNAs within the cell by inducing their cleavage and repressing their translation. Therefore, shRNA knockdown vectors are not associated with any DNA level sequence change of the gene of interest.
CRISPR and TALEN both function by directing nucleases to cut specific target sites in the genome. These cuts are then inefficiently repaired by the cellular machinery, resulting in permanent mutations, such as small insertions or deletions, at the sites of repair. A subset of these mutations will result in loss of function of the gene of interest due to frame-shifts, premature stop codons, etc. If two closely positioned cut sites in the genome (i.e. within several kb) are targeted simultaneously, this can also result in the deletion of the intervening region.
shRNA-mediated knockdown will never completely repress the expression of the target gene. Even for the most effective shRNAs, some residual expression of the target gene will remain. In contrast, in a fraction of treated cells, CRISPR and TALEN can generate permanent mutations which may result in complete loss of gene function.
Consistency and uniformity
shRNA vectors generally provide high cell-to-cell uniformity within the pool of treated cells and very consistent results between experiments. In contrast, CRISPR and TALEN produce results that are highly non-uniform from cell to cell due to the stochastic nature of the mutations introduced. To fully knock out the gene of interest in a cell, all copies of the gene in the cell must be knocked out. Given that normal cells have two copies of any gene (except for X- or Y-linked genes) while cancer cells can have more than two copies, such full knockout cells may represent a very small fraction of all the treated cells. For this reason, nuclease-mediated knockout experiments require the screening of clones by sequencing to identify the subset in which all copies of the gene of interest have been knocked out.
Off-target effects have been reported for both shRNA-mediated knockdown and nuclease-mediated knockout. The off-target phenotype(s) can be estimated by using multiple different shRNAs to target the same gene. If a gene knocked down by multiple different shRNAs results in consistent phenotype(s), then it argues against the phenotype(s) being caused by off-target effects. For CRISPR- or TALEN-mediated knockout, multiple clones containing loss-of-function mutations should be analyzed in order to account for any phenotype(s) that may be due to off-target mutations. Additionally, bioinformatically identified off-target sites could be sequenced in the clones to see if they have been mutated.
Animal transgenesis: an overview
Transgenic animals are extensively used to study in vivo gene function as well as to model human diseases. The technology for producing transgenic animals exists for a variety of vertebrate and invertebrate species. The mouse is the most utilized organism for research in neurodegenerative diseases. The most commonly used techniques for producing transgenic mice involves either the pronuclear injection of transgenes into fertilized oocytes or embryonic stem cell-mediated gene targeting. Embryonic stem cell technology has been most often used to produce null mutants (gene knockouts) but may also be used to introduce subtle genetic modifications down to the level of making single nucleotide changes in endogenous mouse genes. Methods are also available for inducing conditional gene knockouts as well as inducible control of transgene expression. Here, we review the main strategies for introducing genetic modifications into the mouse, as well as in other vertebrate and invertebrate species. We also review a number of recent methodologies for the production of transgenic animals including retrovirus-mediated gene transfer, RNAi-mediated gene knockdown and somatic cell mutagenesis combined with nuclear transfer, methods that may be more broadly applicable to species where both pronuclear injection and ES cell technology have proven less practical.
Down and Out with RNAi and CRISPR
Genes can be knocked down with RNA interference (RNAi) or knocked out with CRISPR-Cas9. RNAi, the screening workhorse, knocks down the translation of genes by inducing rapid degradation of a gene target’s transcript.
CRISPR-Cas9, the new but already celebrated genome-editing technology, cleaves specific DNA sequences to render genes inoperative. Although mechanistically different, the two techniques complement one another, and when used together facilitate discovery and validation of scientific findings.
RNAi technologies along with other developments in functional genomics screening were discussed by industry leaders at the recent Discovery on Target conference. The conference, which emphasized the identification and validation of novel drug targets and the exploration of unknown cellular pathways, included a symposium on the development of CRISPR-based therapies.
RNAi screening can be performed in either pooled or arrayed formats. Pooled screening provides an affordable benchtop option, but requires back-end deconvolution and deep sequencing to identify the shRNA causing the specific phenotype. Targets are much easier to identify using the arrayed format since each shRNA clone is in an individual well.
“CRISPR complements RNAi screens,” commented Ryan Raver, Ph.D., global product manager of functional genomics at Sigma-Aldrich. “You can do a whole genome screen with either small interfering RNA (siRNA) or small hairpin RNA (shRNA), then validate with individual CRISPRs to ensure it is a true result.”
A powerful and useful validation method for knockdown or knockout studies is to use lentiviral open reading frames (ORFs) for gene re-expression, for rescue experiments, or to detect whether the wild-type phenotype is restored. However, the ORF randomly integrates into the genome. Also, with this validation technique, gene expression is not acting under the endogenous promoter. Accordingly, physiologically relevant levels of the gene may not be expressed unless controlled for via an inducible system.
In the future, CRISPR activators may provide more efficient ways to express not only wild-type but also mutant forms of genes under the endogenous promoter.
Choice of screening technique depends on the researcher and the research question. Whole gene knockout may be necessary to observe a phenotype, while partial knockdown might be required to investigate functions of essential or lethal genes. Use of both techniques is recommended to identify all potential targets.
For example, recently, a whole genome siRNA screen on a human glioblastoma cell line identified a gene, known as FAT1, as a negative regulator of apoptosis. A CRISPR-mediated knockout of the gene also conferred sensitivity to death receptor–induced apoptosis with an even stronger phenotype, thereby validating FAT1’s new role and link to extrinsic apoptosis, a new model system.
Dr. Raver indicated that next-generation RNAi libraries that are microRNA-adapted might have a more robust knockdown of gene expression, up to 90–95% in some cases. Ultracomplex shRNA libraries help to minimize both false-negative and false-positive rates by targeting each gene with
25 independent shRNAs and by including thousands of negative-control shRNAs.
Recently, a relevant paper emerged from the laboratory of Jonathan Weissman, Ph.D., a professor of cellular and molecular pharmacology at the University of California, San Francisco. The paper described how a new ultracomplex pooled shRNA library was optimized by means of a microRNA-adapted system. This system, which was able to achieve high specificity in the detection of hit genes, produced robust results. In fact, they were comparable to results obtained with a CRISPR pooled screen. Members of the Weissman group systematically optimized the promoter and microRNA contexts for shRNA expression along with a selection of guide strands.
Using a sublibrary of proteostasis genes (targeting 2,933 genes), the investigators compared CRISPR and RNAi pooled screens. Data showed 48 hits unique to RNAi, 40 unique to CRISPR, and an overlap of 21 hits (with a 5% false discovery rate cut-off). Together, the technologies provided a more complete research story.
RNA interference (RNAi) silences, or knocks down, the translation of a gene by inducing degradation of a gene target’s transcript. To advance RNAi applications, Thermo Fisher Scientific has developed two types of small RNA molecules: short interfering RNAs and microRNAs. The company also offers products for RNAi analysis in vitro and in vivo, including libraries for high-throughput applications.
Arrayed CRISPR Screens
“RNA screens are well accepted and will continue to be used, but it is important biologically that researchers step away from the RNA mechanism to further study and validate their hits to eliminate potential bias,” explained Louise Baskin, senior product manager, Dharmacon, part of GE Healthcare. “The natural progression is to adopt CRISPR in the later stages.”
RNAi uses the cell’s endogenous mechanism. All of the components required for gene knockdown are already within the cell, and the delivery of the siRNA starts the process. With the CRISPR gene-editing system, which is derived from a bacterial immune defense system, delivery of both the guide RNA and the Cas9 nuclease, often the rate limiter in terms of knockout efficiency, are required.
In pooled approaches, the cell has to either drop out or overexpress so that it is sortable, limiting the types of addressable biological questions. A CRISPR-arrayed approach opens up the door for use of other analytical tools, such as high-content imaging, to identify hits of interest.
To facilitate use of CRISPR, GE recently introduced Dharmacon Edit-R synthetic CRISPR RNA (crRNA) libraries that can be used to carry out high-throughput arrayed analysis of multiple genes. Rather than a vector- or plasmid-based gRNA to guide the targeting of the Cas9 cleavage, a synthetic crRNA and tracrRNA are used. These algorithm-designed crRNA reagents can be delivered into the cells very much like siRNA, opening up the capability to screen multiple target regions for many different genes simultaneously.
GE showed a very strong overlap between CRISPR and RNAi using this arrayed approach to validate RNAi screen hits with synthetic crRNA. The data concluded that CRISPR can be used for medium- or high-throughput validation of knockdown studies.
“We will continue to see a lot of cooperation between RNAi and gene editing,” declared Baskin. “Using the CRISPR mechanism to knockin could introduce mutations to help understand gene function at a much deeper level, including a more thorough functional analysis of noncoding genes.
“These regulatory RNAs often act in the nucleus to control translation and transcription, so to knockdown these genes with RNAi would require export to the cytoplasm. Precision gene editing, which takes place in the nucleus, will help us understand the noncoding transcriptome and dive deeper into how those genes regulate disease progression, cellular development and other aspects of human health and biology.”
Synthetic crRNA:tracrRNA reagents can be used in a similar manner to siRNA reagents for assessment of phenotypes in a cell population. Top row: A reporter cell line stably expressing Cas9 nuclease was transfected with GE Dharmacon’s Edit-R synthetic crRNA:tracrRNA system, which was used to target three positive control genes (PSMD7, PSMD14, and VCP) and a negative control gene (PPIB). Green cells indicate EGFP signaling occuring as a result of proteasome pathway disruption. Bottom row: A siGENOME siRNA pool targeting the same genes was used in the same reporter cell line.
The functional genomics tool should fit the specific biology the biology should not be forced to fit the tool. Points to consider include the regulation of expression, the cell line or model system, as well as assay scale and design. For example, there may be a need for regulatable expression. There may be limitations around the cell line or model system. And assay scale and design may include delivery conditions and timing to optimally complete perturbation and reporting.
“Both RNAi- and CRISPR-based gene modulation strategies have pros and cons that should be considered based on the biology of the genes being studied,” commented Gwen Fewell, Ph.D., chief commercial officer, Transomic Technologies.
RNAi reagents, which can produce hypomorphic or transient gene-suppression states, are well known for their use in probing drug targets. In addition, these reagents are enriching studies of gene function. CRISPR-Cas9 reagents, which produce clean heterozygous and null mutations, are important for studying tumor suppressors and other genes where complete loss of function is desired.
Timing to readout the effects of gene perturbation must be considered to ensure that the biological assay is feasible. RNAi gene knockdown effects can be seen in as little as 24–72 hours, and inducible and reversible gene knockdown can be realized. CRISPR-based gene knockout effects may become complete and permanent only after 10 days.
Both RNAi and CRISPR reagents work well for pooled positive selection screens however, for negative selection screens, RNAi is the more mature tool. Current versions of CRISPR pooled reagents can produce mixed populations containing a fraction of non-null mutations, which can reduce the overall accuracy of the readout.
To meet the needs of varied and complex biological questions, Transomic Technologies has developed both RNAi and CRISPR tools with options for optimal expression, selection, and assay scale. For example, the company’s shERWOOD-UltramiR shRNA reagents incorporate advances in design and small RNA processing to produce increased potency and specificity of knockdown, particularly important for pooled screens.
Sequence-verified pooled shRNA screening libraries provide flexibility in promoter choice, in vitro formats, in vivo formats, and a choice of viral vectors for optimal delivery and expression in biologically relevant cell lines, primary cells or in vivo.
The company’s line of lentiviral-based CRISPR-Cas9 reagents has variable selectable markers such that guide RNA- and Cas9-expressing vectors, including inducible Cas9, can be co-delivered and selected for in the same cell to increase editing efficiency. Promoter options are available to ensure expression across a range of cell types.
“Researchers are using RNAi and CRISPR reagents individually and in combination as cross-validation tools, or to engineer CRISPR-based models to perform RNAi-based assays,” informs Dr. Fewell. “Most exciting are parallel CRISPR and RNAi screens that have tremendous potential to uncover novel biology.”
Schematic of a pooled shRNA screening workflow developed by Transomic Technologies. Cells are transduced, and positive or negative selection screens are performed. PCR amplification and sequencing of the shRNA integrated into the target cell genome allows the determination of shRNA representation in the population.
The convergence of RNAi technology with genome-editing tools, such as CRISPR-Cas9 and transcription activator-like effector nucleases, combined with next-generation sequencing will allow researchers to dissect biological systems in a way not previously possible.
“From a purely technical standpoint, the challenges for traditional RNAi screens come down to efficient delivery of the RNAi reagents and having those reagents provide significant, consistent, and lasting knockdown of the target mRNAs,” states Ross Whittaker, Ph.D., a product manager for genome editing products at Thermo Fisher Scientific. “We have approached these challenges with a series of reagents and siRNA libraries designed to increase the success of RNAi screens.”
Thermo Fisher’ provides lipid-transfection RNAiMax reagents, which effectively deliver siRNA. In addition, the company’s Silencer and Silencer Select siRNA libraries provide consistent and significant knockdown of the target mRNAs. These siRNA libraries utilize highly stringent bioinformatic designs that ensure accurate and potent targeting for gene-silencing studies. The Silencer Select technology adds a higher level of efficacy and specificity due to chemical modifications with locked nucleic acid (LNA) chemistry.
The libraries alleviate concerns for false-positive or false-negative data. The high potency allows less reagent use thus, more screens or validations can be conducted per library.
Dr. Whittaker believes that researchers will migrate regularly between RNAi and CRISPR-Cas9 technology in the future. CRISPR-Cas9 will be used to create engineered cell lines not only to validate RNAi hits but also to follow up on the underlying mechanisms. Cell lines engineered with CRISPR-Cas9 will be utilized in RNAi screens. In the long term, CRISPR-Cas9 screening will likely replace RNAi screening in many cases, especially with the introduction of arrayed CRISPR libraries.
Validating Antibodies with RNAi
Unreliable antibody specificity is a widespread problem for researchers, but RNAi is assuaging scientists’ concerns as a validation method.
The procedure introduces short hairpin RNAs (shRNAs) to reduce expression levels of a targeted protein. The associated antibody follows. With its protein knocked down, a truly specific antibody shows dramatically reduced or no signal on a Western blot. Short of knockout animal models, RNAi is arguably the most effective method of validating research antibodies.
The method is not common among antibody suppliers—time and cost being the chief barriers to its adoption, although some companies are beginning to embrace RNAi validation.
Overcoming Drawbacks of Gene Silencing with RNAi
Gene silencing by RNA interference has become a key tool in research and drug discovery since its discovery by Andrew Fire and Craig Mello. RNAi allows for the sequence-specific knockdown of target genes with a subsequent phenotypic analysis, representing a straightforward method for ascribing functions to genes.
The advantages of RNAi include the high efficiency of the gene knockdown, the ability to easily target the gene of interest, as well as stable and long-term silencing by expressing shRNAs. This makes for a powerful tool that has been successfully applied to answer many questions in cell biology. There are also risks, however, when using RNAi that need to be considered carefully.
Researchers have demonstrated that RNAi sequences do not only bind to one target. These changes in the gene expression pattern of the cell and potentially of the phenotype give rise to an off-target signature. The unidentified effects of such a signature bear a high risk to create false-positive outcomes that can bring a complete project into jeopardy. Furthermore, even most up to date algorithm-based sequence designs show a knockdown efficiency that is generally at 80% or less.
The effects of such specific but low knockdown can be masked by the off-target signature with phenotypic changes being undetectable. These issues can make RNAi unpredictable, slow, and risky, in particular in drug discovery, where speed and reliability of results are crucial factors. Sirion Biotech has developed technologies that help overcome the drawbacks of RNA interference.
Figure 1. Screening of 10 shRNA sequences (shRNA1-10) for the silencing of Hsp90b: The shRNA5 was identified with a knockdown efficiency of over 95%. Sequence efficiency was confirmed on both mRNA and protein level in NIH-3T3 cells that were transduced with an adenoviral vector expressing shRNA5.
Screening for Active shRNAs
One particular challenge with shRNAs is finding sequences that are highly effective with a validated efficiency of over 80%. Sirion has developed RNAiONE™ for fast and reliable identification of the most efficient sequence for a specific gene.
RNAiONE has been used in the silencing of murine Hsp90b, a difficult target due to its abundance. Ten validation vectors containing different bioinformatically evaluated shRNA sequences targeting Hsp90b were transfected in an optimized cell system expressing the GFP-tagged Hsp90b protein from the same vector.
Initial microscopic analysis provided first evidence of shRNA efficiency based on the GFP readout. Validation of the knockdown was then carried out by measuring the mRNA level with real-time PCR and confirmed by Western blot analysis. The results show that the efficiency can vary tremendously for different target sequences with a knockdown between 0 and over 90% (Figure 1), demonstrating the need for an appropriate validation protocol.
The data also illustrates that a highly effective sequence could be successfully identified, despite the high abundance of Hsp90b in the cell. Furthermore, results with RNAiONE can be applied reliably to Sirion’s different viral vector platforms providing immediate access to subsequent gene analysis and assay cell model generation. The overall effectiveness of RNAiONE allows Sirion to identify shRNA for most of the genes and to issue warranties to its customer labs.
Figure 2. VariCHECK enables a simultaneous switch between a wild-type protein and the ectopic undegradable protein by simply adding Doxycycline into the cell media.
A Cell Model with RNAi Built-in Control
A prerequisite to successfully conduct meaningful shRNA experiments, however, is not only a high and validated knockdown, but also the use of appropriate controls. VariCHECK™ serves as such a perfect built-in control for RNAi experiments which allows confirmation of on-target phenotypes. This system enables the depletion of an endogenous targetX, while simultaneously overexpressing the ectopic undegradable form of targetX, which then rescues a specific phenotype (Figure 2).
The basis of VariCHECK is Sirion’s All-In ONE vector inducible lentiviral system based on Clontech’s latest 3G technology. The constitutive TET-dependent transactivator is simultaneously expressed together with the TET-inducible gene of interest or the validated shRNA. Cells are transduced with a first vector that contains the inducible shRNA targeting the wild-type gene of interest (GOI) and a second vector with the inducible ectopic undegradable form of the GOI.
Figure 3A. VariCHECK serves as a perfect built-in control for RNAi experiments. Depletion of the oncology target GOIx significantly reduces cell proliferation in A549 cells.
Without Doxycycline (Dox), neither shRNA nor ectopic GOI expression takes place and the endogenous wild-type protein is present in the cell. In contrast, with Dox, wild-type transcripts are degraded by the expressed shRNA, while the ectopic protein is present.
The validity of VariCHECK has been successfully demonstrated for the functional analysis of an oncology target (GOIx). A stable cell line was generated that expressed the TET-inducible shRNA against GOIx, previously validated with RNAiONE. Gene knockdown was quantified, and the best performer cell clone with near quantitative knockdown was then transduced with lentiviral vector 2 that encoded the inducible ectopic undegradable GOIx.
Expression of GOIx in the absence and in the presence of Dox was quantified by real-time PCR and Western blot analysis, respectively, confirming an almost quantitative switch from wild type to an ectopically expressed form of GOIx on mRNA as well as protein level. Moreover, the inhibitory effect of GOIx knockdown on cell proliferation was fully rescued by ectopic expression of the undegradable GOIx, demonstrating the reduced proliferation as a real on-target phenotype (Figures 3A–C).
Figure 3B. Simultaneous overexpression of the ectopic undegradable GOIx fully restored proliferation, confirming a real on-target phenotype.
Beside its function as RNAi built-in control, VariCHECK can be also applied to reversibly switch expression between the wild-type protein and a defined mutant version in a single cell line. This approach can be used to undertake functional analysis of proteins and their mutations in a fast and equally reliable manner, which is of particular interest for the analysis of escape mutants from cancer drugs.
Since its discovery in 1998, RNAi silencing has become the method of choice for phenotype studies however, there are some risks that need to be considered including an insufficient knockdown and potential off-target effects. This can be highly critical in drug discovery and development, and add another question mark in compound and target validation, a process that is already lengthy and risky. Technologies such as RNAiONE and VariCHECK make the use of RNAi for scientists faster, more predictive and reliable in particular in drug discovery.
Figure 3C. Western blot confirms the near quantitative switch from wild type to ectopic expression of GOIx after Dox treatment.
Results and discussion
Overview of the MAGeCK algorithm
A schematic of the MAGeCK algorithm is presented in Figure 1. Briefly, read counts from different samples are first median-normalized to adjust for the effect of library sizes and read count distributions. Then the variance of read counts is estimated by sharing information across features, and a negative binomial (NB) model is used to test whether sgRNA abundance differs significantly between treatments and controls. This approach is similar to those used for differential RNA-Seq analysis ,,. We rank sgRNAs based on P-values calculated from the NB model, and use a modified robust ranking aggregation (RRA) algorithm  named α-RRA to identify positively or negatively selected genes. More specifically, α-RRA assumes that if a gene has no effect on selection, then sgRNAs targeting this gene should be uniformly distributed across the ranked list of all the sgRNAs. α-RRA ranks genes by comparing the skew in rankings to the uniform null model, and prioritizes genes whose sgRNA rankings are consistently higher than expected. α-RRA calculates the statistical significance of the skew by permutation, and a detailed description of the algorithm is presented in the Materials and methods section. Finally, MAGeCK reports positively and negatively selected pathways by applying α-RRA to the rankings of genes in a pathway.
Overview of the MAGeCK algorithm. Raw read counts corresponding to single-guided RNAs (sgRNAs) from different experiments are first normalized using median normalization and mean-variance modeling is used to capture the relationship of mean and variance in replicates. The statistical significance of each sgRNA is calculated using the learned mean-variance model. Essential genes (both positively and negatively selected) are then identified by looking for genes whose sgRNAs are ranked consistently higher (by significance) using robust rank aggregation (RRA). Finally, enriched pathways are identified by applying the RRA algorithm to the ranked list of genes.
CRISPR/Cas9 knockout screen datasets
We examined three recently published CRISPR/Cas9 knockout screen experiments ,,. The first experiment (or 'ESC dataset') performed negative selection on mouse embryonic stem cells (ESCs) to screen for essential genes. The second experiment (or 'leukemia dataset')  performed similar negative selection experiments on the chronic myeloid leukemia cell line KBM7 and the acute promyelocytic leukemia cell line HL-60. The controls for these studies were cells before Cas9 activation. The third experiment (or 'melanoma dataset')  was based on one human melanoma cell line (A375), which harbors a V600E mutation in the BRAF protein kinase gene. In this study, positive selection was performed to identify genes whose knockouts resulted in resistance to 7-day and 14-day treatment with the BRAF inhibitor vemurafenib (PLX), and the controls were cells treated with dimethyl sulfoxide (DMSO).
MAGeCK outperforms other methods in detecting significantly selected sgRNAs and genes
We compared MAGeCK with two different categories of methods, including methods for statistical evaluation of high-throughput sequencing read counts using NB models (edgeR and DESeq), and methods originally designed for ranking genes in genome-scale RNAi screens (RIGER and RSA). A summary of the comparisons between MAGeCK and these algorithms is presented in Table 1.
We first compared MAGeCK with edgeR and DESeq. All three algorithms model the high variance of sgRNAs with higher mean read counts (Figure S1 in Additional file 1). The variance models of MAGeCK and DESeq are similar, while edgeR has a lower variance estimation when read counts are low. We also evaluated the FDR of different algorithms by making comparisons between control samples and between replicates of the treatment samples in the ESC and melanoma datasets (there were no replicated treatment samples in the leukemia dataset). Since the CRISPR/Cas9 knockout system should show no difference in selection preference between control samples or between replicated treatment samples, a good method should not detect many significantly selected sgRNAs and genes between these samples. MAGeCK identified fewer significantly selected sgRNAs using the NB model than edgeR and DESeq (see Section A of Supplementary materials in Additional file 1 for more details). The distribution of the calculated P-values for all the sgRNAs approximates a uniform distribution (Figure S4 in Additional file 1), which indicates that our model controls the specificity for comparisons where we expect no true positives.
Next we compared the performance of MAGeCK with two RNAi screening algorithms, RIGER and RSA, at both the sgRNA and gene level. MAGeCK ranks sgRNAs based on the NB P-values, while the ranking of RIGER is based on the signal-to-noise ratio. RSA ranks sgRNAs based on their fold change between treatment and controls, but this approach introduces bias towards sgRNAs with fewer read counts (Figure S5 in Additional file 1). At the gene level, RIGER’s sensitivity was lower, and it identified less than 30 significantly selected genes in all datasets, and missed many of the essential genes (for example, ribosomal genes) in two negative screening studies , (Figure 2a). RSA had low specificity and reported high numbers of genes, even when comparing controls or replicates (Figure S6 in Additional file 1, Table S1 in Additional file 2). In contrast, MAGeCK was able to detect significant genes when comparing treatments with controls, while giving very few false positives when comparing controls or replicates (Figure 2a Table S1 in Additional file 2).
A comparison of MAGeCK with two other RNA interference screen algorithms, RIGER and RSA. (a) The numbers of significantly selected genes identified by MAGeCK, RIGER and RSA in different comparisons. For comparisons between control samples or between replicates of the same condition (highlighted in yellow), ideally no significantly selected genes should be detected. Comparisons between treatments and controls are highlighted in green. See Table S1 in Additional file 2 for a complete comparison. (b) The overlap of top-ranked genes between CRISPR/Cas9 knockout screening and RNAi screening on the melanoma dataset. The positive screening experiment was performed in the same way as for the melanoma dataset , except that pooled shRNA screening was used instead of CRISPR/Cas9 knockout screening.
Finally, we compared the screening results from the melanoma dataset with those from an independent study which used pooled shRNAs to screen PLX-treated A375 cells . We applied MAGeCK, RIGER and RSA to both the CRISPR/Cas9 knockout screens and shRNA screens and checked the consistency of the top-ranked genes (Figure 2b). Although the overall consistency of genes called from the different screens was low (fewer than 5% overlap), MAGeCK always identified more consistent genes than RIGER and RSA at different cutoffs. This shows that MAGeCK can be used for both RNAi screens and CRISPR/Cas9 knockout screens, and that MAGeCK identifies more consensus hits between different screening technologies than other methods (Table S2 in Additional file 2).
MAGeCK reports robust results with different sequencing depths and different numbers of sgRNAs per gene
Both sequencing depth and the number of targeting sgRNAs per gene affect the CRISPR/Cas9 knockout screening experiment outcomes substantially. To study the effect of sequencing depth on performance, we randomly sampled sequencing reads in one negative screening dataset (the leukemia dataset) and one positive screening dataset (the melanoma dataset), and used MAGeCK to identify significantly selected sgRNAs and genes. We compared the numbers of significantly selected sgRNAs and genes that are identified for different numbers of down-sampled reads (Figures 3 and 4 see Materials and methods for more details). At the sgRNA level, less than 10% of the sgRNAs could be detected in the datasets with one million reads (or 3.3% and 5.7% of the reads in the leukemia and melanoma datasets, respectively) compared with the full datasets. At the gene-level, however, MAGeCK could still detect, on average, over 40% and 80% of the genes in the full leukemia and melanoma datasets, respectively. This suggests that the robust rank aggregation approach makes MAGeCK robust to sequencing depth. Interestingly, MAGeCK could detect over 30% of the significantly positively selected sgRNAs in the melanoma dataset using only 1 million reads (Figure 4), a much larger fraction compared with the negatively selected genes in both datasets. This is because the reads corresponding to these sgRNAs dominate the library population (Figure S7 in Additional file 1), and the sequencing depth required to detect positively selected sgRNAs is much less in the positive selection screens.
MAGeCK is robust against sequencing depth and the number of targeting sgRNAs per gene. (a) The number of significantly selected sgRNAs and genes in the leukemia dataset using various sequencing depths. The maximum sequencing depth for all samples is 30 million. See Materials and methods for sampling details. (b) The number of significantly selected sgRNAs and genes in the melanoma dataset using various sequencing depths. The maximum sequencing depth for all samples is 17.5 million. See Materials and methods for sampling details. Error bars represent the standard deviation from three independent sampling experiments.
The number of identified positively and negatively selected sgRNAs at different sequencing depths. (a,b). The numbers of positively and negatively selected sgRNAs in the leukemia dataset (a) and melanoma dataset (b) under different sequencing depths are shown. The numbers are normalized by the number of identified sgRNAs at the maximum sequencing depths (30 million for the leukemia dataset, 17.5 million for the melanoma dataset).
We next evaluated the performance of the different algorithms after reducing the number of sgRNAs in a CRISPR/Cas9 knockout screen. The leukemia dataset was used since, on average, >10 sgRNAs were designed to target each gene. As the true essential genes are unknown, we selected 168 'reference' genes that are consistently ranked among the top 5% by all three methods using 10 sgRNAs/gene. We then tested whether the algorithms can detect these 'reference' genes using fewer sgRNAs (Figure 5 see Materials and methods for more details). Both MAGeCK and RSA detected more reference genes than RIGER, and could still identify over 80% of these 'reference' genes with four to six sgRNAs per gene (Figure 5). This suggests that when there are fewer sgRNAs available for some genes, MAGeCK and RSA can still make robust calls.
MAGeCK is robust to the number of targeting sgRNAs per gene. This figure shows the effect of different numbers of targeting sgRNAs per gene. Each bar indicates the percentage of top-ranked, 'reference' genes that are identified by MAGeCK, RIGER and RSA using different numbers of sgRNAs per gene. 'Reference' genes are those that are in the top 5% of ranked genes in all three methods when using 10 sgRNAs per gene. See Materials and methods for sampling details. Error bars represent the standard deviation from three independent sampling experiments.
MAGeCK identifies known and novel biologically interesting genes and pathways
We applied MAGeCK to the original CRISPR/Cas9 knockout screen studies to identify positively and negatively selected genes and pathways. Genes in pathways from the KEGG (Kyoto Encyclopedia of Genes and Genomes) and REACTOME databases were evaluated for pathway enrichment (Tables S3 to S10 in Additional file 2 Tables S11 to S18 in Additional file 3). In the leukemia and ESC CRISPR/Cas9 knockout screen studies, negatively selected genes were enriched in many fundamental pathways (Tables S9 and S10 in Additional file 2 Tables S10 to S14 in Additional file 3) ,. Pluripotency genes and genes that are well known to be essential for ESC proliferation were also negatively selected, consistent with the observations reported in the original study (Table 2). In the melanoma dataset, the oxidative phosphorylation pathway was negatively selected in the normal condition (treated with 14-day DMSO versus 7-day DMSO), supporting the hypothesis that melanoma cells are addicted to oxidative phosphorylation . Under the PLX treated condition, in addition to the genes that were reported before  (Table S7 in Additional file 2), MAGeCK also identified several new positively selected genes (Table 2), such as CDH13 (FDR = 1.7e-2, ranked 9th out of 17,419) and PPT1 (FDR = 8.5e-2, ranked 14th out of 17,419). Loss-of-function mutations of PPT1 cause neuronal ceroid lipofuscinosis and are resistant to apoptosis induction -,. CDH13, a tumor suppressor that negatively regulates cell growth, is frequently hyper-methylated and contributes to tumorigenesis in melanoma, lung and colorectal cancers ,,,. Interestingly, these cancers often harbor a BRAF V600E mutation that can be treated with the BRAF inhibitor PLX, and this mutation is also present in the melanoma cell line used in this CRISPR/Cas9 knockout screen. Our results imply that tumor patients harboring BRAF V600E mutations might have suboptimal response to PLX treatment if their tumors have CDH13 hypermethylation.
MAGeCK allows bi-directional screening and cell-type-specific screening
Although the original leukemia and ESC studies are negative screens and the melanoma study is a positive screen, MAGeCK is also able to perform bi-directional analysis to search for both positively and negatively selected genes simultaneously. This functionality allows MAGeCK to gain biological insights beyond the original screen design. For example, MAGeCK identified several positively selected genes from both negative-selection screens (the leukemia and ESC datasets), and negatively selected genes in the positive-selection screen (the melanoma dataset) (Table 2 Tables S4 and S8 in Additional file 2 Table S12 in Additional file 3). In the leukemia dataset, MAGeCK identified 23 positively selected genes, whose knockout induces cell proliferation. They include MAP4K3 (FDR = 0.14, ranked 9th out of 7,115), a tumor suppressor kinase in the mitogen-activated protein kinase (MAPK) pathway which induces apoptosis -,, and EPM2A (FDR = 0.14, ranked 10th out of 7,115), another protein phosphatase that negatively regulates cell cycle progression ,. From the ESC dataset, TRP53, a mouse ortholog of the human TP53 tumor suppressor gene ,, was ranked first out of the three positively selected genes identified. The negative regulator functions of these genes are consistent with our results that knocking them out confers a selective advantage for cell growth. From the melanoma dataset, MAGeCK only identified one negatively selected gene, RREB1, in the 14-day PLX treatment. RREB1 (FDR = 0.05, ranked 1st out of 17,419) is a transcription factor and a downstream activator in the RAS-RAF signaling pathway ,,, which is closely related to the BRAF mutation found in A375 cells ,. Interestingly, MAGeCK also found EGFR (FDR = 0.025, ranked 6th out of 17,419) and its associated pathways to be negatively selected in the 7-day PLX-treated samples, implying that PLX-treated cells are more dependent on EGFR. Our finding is consistent with recent studies linking ectopic EGFR expression in melanoma cells to PLX resistance , and with the improved efficacy of BRAF and EGFR combination inhibition in colorectal cancer cells with the BRAF V600 mutation ,.
Finally, we applied MAGeCK to identify cell type-specific essential genes and pathways that differ between the chronic myeloid leukemia cell line KBM7 and the acute promyelocytic leukemia cell line HL-60, which are part of the leukemia dataset ,,, (Tables S15 to S18 in Additional file 3). MAGeCK identified the KEGG 'chronic myeloid leukemia' pathway as essential in KBM7 (FDR = 9.00e-4, ranked 6th out of 181), correctly distinguishing the cell type differences between KBM7 and HL-60. At the gene level, IGF1R (FDR = 1.98e-3, ranked 1st out of 7,115) was found to be specifically essential in HL-60, which is consistent with the observation that an IGF1R inhibitor reduces proliferation and induces apoptosis in HL-60 cells ,. In addition, MAGeCK identified BCR (FDR = 1.60e-3, ranked 7th out of 7,115) and ABL1 (FDR = 1.98e-3, ranked 18th out of 7,115) as specifically essential in KBM7, which is consistent with the presence of the BCR-ABL fusion in this cell line ,,,. The ability of MAGeCK to identify cell type-specific essential genes will be very useful as more CRISPR/Cas9 knockout screening data become publicly available.
Genetic robustness is the ability of a living organism to maintain its viability and fitness despite genetic variations, including perturbations. Genetic perturbations play an important role in evolution however, organisms require buffering systems to ensure similar developmental outcomes despite minor differences in genetic makeup or environmental conditions, a process known as robustness or canalization [1, 2]. In 1932, dosage compensation was reported as the first example of genetic robustness. Male fruit flies were reported to have a twofold increase in transcription from their single X chromosome, resulting in the same gene expression levels as females with two active X chromosomes [3, 4]. In contrast, in mammals, females undergo inactivation of one of their X chromosomes through heterochromatization, allowing for similar developmental outcomes in both sexes [5–7]. The concept of genetic robustness was further supported by several recent studies: for example, only 20% of the protein-coding genes in yeast were reported to be essential for growth in laboratory conditions , and a lack of phenotype was reported for several mouse , zebrafish , and Arabidopsis  mutants.
Genetic robustness may arise from redundant genes, whereby the loss of one gene may be compensated by another with overlapping functions and expression pattern, as reported for several mutants in a range of model organisms [12–19] (reviewed in ). Another form of robustness arises from tightly regulated cellular networks including metabolic, signaling, and transcriptional networks. Perturbation of a particular gene’s function in a network may alter the expression of other genes within the same network, thereby maintaining cellular wellness [21, 22]. Additionally, in response to a gene knockout, organisms such as yeast may accumulate mutations in one or more genes modulating the affected pathway, thereby partially or fully rescuing the final outcome [23, 24].
While the above-mentioned modes of genetic robustness may occur as a result of the loss of function of a specific protein, a number of studies suggest a different form of genetic robustness, one that is triggered upstream of protein function (hereafter referred to as genetic compensation or transcriptional adaptation [Table 1]) [25–27]. The increasing use of recent advances in reverse genetic tools have revealed phenotypic differences between knockouts (i.e., mutants) and knockdowns (e.g., antisense-, including morpholino [MO]-, treated animals) in a number of model systems including Arabidopsis [28–30], mouse [31–34], Drosophila , zebrafish [10, 36], and human cell lines [37–39]. While some studies attributed these phenotypic differences to toxicity or off-target effects of the knockdown reagents [40–43] (reviewed in ), a recent study in zebrafish proposed gene expression changes and consequent compensation in mutant but not knockdown animals as the reason for the observed differences . While knockdown of egfl7, an endothelial extracellular-matrix (ECM) gene, leads to severe vascular defects, most egfl7 mutants exhibit no obvious defects. This discrepancy was attributed at least partly to the upregulation of other ECM proteins, specifically Emilins, in egfl7 mutants but not antisense-injected embryos. Moreover, the authors observed minor or no vascular defects upon egfl7 MO injections into egfl7 mutants, indicating that the phenotypic differences are not due to MO toxicity. In addition, this study reported upregulation of vegfab mRNA levels in vegfaa mutant but not knockdown animals. While the mechanisms triggering the transcriptional adaptation response in vegfaa mutant animals remain unknown, the authors propose that it lies upstream of protein function, as overexpression of dominant-negative Vegfaa, which causes a vegfaa mutant-like phenotype, did not lead to an increase in vegfab mRNA levels. In this review, we focus on studies reporting transcriptional adaptation and/or genetic compensation in higher eukaryotes and outline possible underlying molecular mechanisms.
Genetic compensation in response to gene knockout is a widespread phenomenon
Upregulation of related genes following a gene knockout may be a direct consequence of the loss of protein function. For example, mice lacking the ribosomal gene Rpl22 show no defects in translation owing to the upregulation of its paralogue, Rpl22l1, the expression of which is normally inhibited by RPL22 . Upregulation of related genes due to the loss of a negative feedback loop may be the first hypothesis to test when a mutant fails to show a phenotype, and a knockdown approach may help test it. For example, human RBL2 mutant T lymphocytes proliferate normally and exhibit normal immune function due to RBL1 upregulation, an upregulation also detected upon RBL2 knockdown in human breast cancer cell lines [47, 48], suggesting a negative feedback loop. Similarly, both knockouts and knockdowns of HDAC-1 lead to the upregulation of HDAC-2 in several human and mouse cell lines and tissues, and vice versa [49–51].
In contrast, lack of a compensatory response in knockdown animals compared to their corresponding mutants indicates that a trigger upstream of protein function is at play, perhaps the genomic lesion itself or the mutant mRNA (Table 2). For example, small interfering RNA (siRNA)-mediated depletion of TET1, an enzyme that converts 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), in mouse embryonic stem cells (mESCs) leads to a significant reduction in 5hmC levels and a loss of undifferentiated morphology in contrast, Tet1 mutant mESCs exhibit only a slight decrease in 5hmC levels and maintain an undifferentiated morphology , suggesting possible compensation by the closely related enzyme, TET2, in mutant but not knockdown mESCs . In addition, while knockdown of any of the three cyclin D family members was reported to inhibit proliferation in several cell lines [54–56], mice lacking a single isoform develop minimal defects, suggesting compensation by one of the other genes [57–59]. Indeed, knockout of two Cyclin D genes in mouse leads to the upregulation of the third Cyclin D gene. Accordingly, double knockout mice show minor phenotypes only in tissues that fail to upregulate the third Cyclin D gene . In addition, mouse Cyclin D2 mutant B lymphocytes exhibit no obvious proliferative phenotype due to the upregulation of Cyclin D3 . Furthermore, short hairpin RNA (shRNA)-mediated knockdown of Importinα5 was reported to inhibit neural differentiation of mESCs cells  however, Importinα5 mutant mice display normal brain development, possibly due to the upregulation of IMPORTINα4 expression . siRNA-mediated knockdown of Kindlin-2, which encodes an integrin coactivator, in mouse embryonic fibroblasts (MEFs) was reported to decrease INTEGRIN β1 activation and prevent INTERLEUKIN 1β–mediated increase in focal adhesion number . However, Kindlin-2 mutant cells were able to form focal adhesions due to the upregulation of KINDLIN-1 .
In another example, antisense-mediated knockdown of Tau was reported to inhibit axonal elongation in cultured neuronal cells [77, 78]. However, axonal elongation was not affected in cultured neurons from Tau mutants, possibly due to the upregulation of microtubule-associated protein 1A (MAP1A) . Interestingly, such upregulation was not detected upon Tau knockdown in mouse oligodendrocytes . Dystrophin mutant mice have been reported not to develop a severe muscular dystrophy phenotype due to the upregulation of a number of genes including that encoding the dystrophin-related protein UTROPHIN [81, 82]. Interestingly, UTROPHIN upregulation was not detected in Dystrophin knockdown mice .
Furthermore, β-Actin mutant mice were reported to display transcriptional upregulation of several other Actin genes, including γ-Actin and α-Actin [27, 84, 85]. Interestingly, restoration of β-Actin expression in β-Actin mutant MEFs did not lead to a reduction in the γ-Actin transcriptional upregulation response, implying that this transcriptional adaptation response is triggered upstream of β-ACTIN function . In addition, γ-Actin knockout, but not knockdown, in MEFs leads to αsm-ACTIN upregulation . Moreover, while siRNA-mediated depletion of the centrosomal protein AZI1 in MEFs leads to a significant decrease in ciliogenesis, MEFs derived from Azi1 mutant mice display no defects in ciliogenesis . The authors also reported that Azi1 mutant MEFs were resistant to Azi1 siRNA, ruling out off-target effects of the siRNA and leading them to hypothesize the existence of a compensatory response in the mutant MEFs. Interestingly, this potential compensation is not observed during sperm flagella formation. This approach of testing the antisense reagent in mutant cells was subsequently used by Rossi et al. in zebrafish  and can be a powerful tool to identify cases of compensation in mutants versus nonspecific effects of the knockdown reagents.
Global versus conditional loss-of-function studies
Reduction or absence of a phenotype in several germline mutants compared to their conditional counterparts has been reported in a number of studies in mouse. For example, germline mutants for Pkm2 are viable and fertile  however, conditional deletion of Pkm2 in MEFs limits nucleotide synthesis, leading to cell-cycle arrest . Similarly, Sirt1 mutant mice have no obvious liver defects, while hepatocyte-specific Sirt1 mutant mice develop fatty liver . Mice with conditional Fgfr3 deletion in chondrocytes exhibit more severe (and a higher incidence of) chondrona-like lesions compared to global mutant mice . Moreover, conditional loss of the RETINOBLASTOMA (RB1) tumor suppressor enables cell-cycle reentry of quiescent primary MEFs, while quiescent MEFs derived from global Rb1 mutant animals are unable to reenter the cell cycle, due at least in part to the compensatory upregulation of p107 . In addition, while Cd44 global mutant mice display only mild phenotypes [91, 92], keratinocyte-specific mutant mice display reduced epidermal stiffness and delayed wound healing, as well as reduced keratinocyte proliferation in response to 12-O-tetradecanoylphorbol-13-acetate . While cell nonautonomous effects may underlie some of these discrepancies, an alternative hypothesis is that a compensatory network becomes established during germline maturation or embryonic development, allowing the organism to adapt to the mutation. Recent data in zebrafish, however, suggest that a mutation does not need to go through the germ line to induce a compensatory response , indicating that multiple mechanisms may underlie this process.
Mechanisms underlying the transcriptional adaptation response
Based on the observations reported thus far, one can identify at least two possible triggers of the transcriptional adaptation response: (1) the DNA lesion and (2) the mutant mRNA. We will first speculate about how each of these potential triggers might lead to transcriptional adaptation and then briefly review other potential triggers including some that might induce posttranscriptional adaptation.
DNA lesion as the trigger for the transcriptional adaptation response.
This section will focus on the DNA lesion being the trigger for the transcriptional adaptation response and will mostly explore the potential role of epigenetic changes following DNA damage.
Following DNA damage, global chromatin reorganization and decondensation are detected [94, 95], actions mediated by several chromatin remodelers and histone-modifying enzymes (reviewed in ). One possibility is that in response to a mutation, global chromatin reorganization may positively affect chromatin accessibility around the compensating gene(s), thereby leading to increased expression levels (Fig 1A). Part of such a model is consistent with the process of dosage compensation in Drosophila where the male-specific lethal (MSL) proteins, together with other proteins, form a complex on the male X chromosome leading to H4K16 acetylation and subsequent induction of an open chromatin configuration, which is more accessible for transcription . Along these lines, a Caenorhabditis elegans study attributed the incomplete penetrance of intestinal phenotypes in skn-1 mutants  to the high variability in expression of the compensating gene end-1 . Interestingly, this variability in end-1 expression was attributed to differences in chromatin remodeling at loci controlling end-1 expression. It will thus be interesting to compare chromatin accessibility at the upregulated genes’ regulatory regions in wild-type, mutant, and knockdown samples.
(A) DNA damage response can induce chromatin reorganization, increasing chromatin accessibility at the compensatory genes’ regulatory regions. (B) Mutations can lead to transcripts that are targeted for degradation through mRNA surveillance pathways. The resulting RNA fragments may trigger the compensatory response. As a secondary effect of the mutated gene’s mRNA degradation, RBPs or miRNAs normally acting on the mutated as well as the compensating genes’ mRNAs become more available to exert their stabilizing effects on the compensating genes’ mRNAs. Abbreviations: miRNAs, microRNAs RBPs, RNA-binding proteins TFs, transcription factors.
Chromatin reorganization may be accompanied by changes in DNA looping and nuclear organization , which may also affect gene expression. Interchromosomal interactions are well documented (reviewed in ), and different kinds of stress, such as temperature, have been shown to increase interchromosomal interactions in Drosophila . DNA damage-induced stress could similarly lead to modifications in interchromosomal interactions, including those between the mutated gene and certain other loci leading to specific gene upregulation. Chromosome-capture studies in wild-type and mutant samples to identify changes in interchromosomal interactions may help test this model.
Leading to a different potential model, a number of studies have reported the generation of small non-coding RNAs (ncRNAs) from regions spanning a double-stranded break (DSB), termed DSB-induced RNAs (diRNAs) [103, 104] (reviewed in ). The authors proposed that such diRNAs are essential for DNA-damage repair (DDR), possibly by acting as guides for chromatin remodelers or proteins important for DDR. Thus, diRNAs might also guide specific transcription factors (TFs) or chromatin remodelers to regulatory regions of compensating genes through homology-based interactions, leading to increased transcription. Such a model of small ncRNAs guiding specific transcription factors or chromatin remodelers to modulate gene expression is consistent with publications describing that roX1 and roX2 RNAs are essential for the dosage-compensation response in Drosophila males by guiding the assembly of the MSL protein complex on the X chromosome and subsequent histone modifications [106–108] (reviewed in ).
One question for these models that involve chromatin remodeling concerns the transmission of the transcriptional adaptation response to the next generation. Genomic imprinting via histone modification [109–111] (reviewed in ) is a possible mechanism.
In addition, induction of GADD45A expression following DNA damage has been reported to induce global DNA demethylation in HEK293T cells, leading to increased activation of methylation-silenced promoters . Thus, one should also assess the methylation status of regulatory regions of the upregulated genes. However, to our knowledge, no links have been established to date between DNA lesions and changes in DNA methylation patterns at specific (i.e., compensating) loci.
Since all these models are based on DNA lesions, it will be important to assess transcriptional adaptation after inducing different types of mutations. One would expect the upregulation of the same genes following all types of mutations, including non-deleterious ones.
Mutant mRNA as the trigger for the transcriptional adaptation response.
This section will focus on the mutant mRNA being the trigger for the transcriptional adaptation response. After reviewing a few examples, we will focus specifically on how RNA fragmentation by different mRNA surveillance pathways could trigger such a response.
Mutations often lead to mRNAs with a premature termination codon (PTC), secondary structures that stall ribosomal translocation, or, less frequently, mRNAs that lack a stop codon. The presence of such mRNAs triggers the nonsense-mediated decay (NMD), no-go decay, or no-stop decay pathways, respectively, which results in mRNA degradation (reviewed in [114–116]). A recent study in zebrafish reported that two different mutations in the same exon of mt2 cause different degrees of phenotypic severity. Surprisingly, the mutant allele with the milder phenotype exhibited a higher degree of NMD. Antisense-mediated knockdown of the NMD pathway and consequent decrease in mutant mRNA degradation led to a more severe phenotype, consistent with the possibility that NMD triggers a compensatory response that decreases the severity of the mutant phenotype . One hypothesis is that the RNA fragments resulting from the mRNA surveillance pathways function to regulate gene expression. While the current understanding in the field is that the mRNA surveillance pathways lead to processive mRNA degradation [114, 118], it is possible that short-lived and relatively rare degradation intermediates are present.
If the fragments are long enough, one can hypothesize that they act in a fashion similar to long noncoding RNAs (reviewed in ) and, for example, guide specific transcription factors or chromatin remodelers to the regulatory regions of compensating genes through homology-mediated base pairing (Fig 1B). Other studies have reported that injection of short (20–22 nt) RNA fragments from a specific mRNA leads to increased transcription of the corresponding locus [120, 121]. Mechanistically, the authors report that the injected sense RNA fragments can form double-stranded RNA (dsRNA) duplexes with short antisense transcripts normally produced from the locus. The resulting dsRNAs may then be utilized by the RNA interference (RNAi) machinery in an ARGONAUTE-dependent manner to induce chromatin modifications at the locus and increase euchromatin histone marks or decrease heterochromatin histone marks. Although the exact machinery underlying such dsRNA-induced epigenetic changes remains unknown, this model is consistent with several other studies reporting transcriptional activation through histone modification following targeting of dsRNA to the promoter region of various genes [122–125]. Previous analyses of the mouse and human transcriptome have identified several antisense transcripts that can participate in forming sense/antisense pairs [126–130]. Thus, one could hypothesize that RNA fragments act in a similar fashion and form dsRNA duplexes with antisense transcripts from the compensating loci, leading to transcriptional upregulation.
RNA-binding proteins (RBPs) can also regulate gene expression in a number of ways (reviewed in ), one of which is by increasing gene expression through stabilizing mRNAs . The highly dynamic binding of RBPs is regulated by cellular conditions therefore, regulating RBP interactions following genotoxic stress may be a mechanism for the cell to compensate for a lost gene. Along these lines, mRNAs that encode functionally related proteins tend to be coregulated by specific RBPs, forming what is known as RNA operons or RNA regulons [133–135] (reviewed in ). Thus, the mutant and compensating genes might be regulated by the same RBPs, and if the mutant mRNA is subjected to degradation or if its secondary structure is affected by the mutation (thereby affecting RBP binding), RBPs would become available to stabilize the compensatory genes’ mRNAs (Fig 1B).
Besides their well-known function in silencing gene expression , micro-RNAs (miRNAs) can enhance gene expression through several mechanisms. Although miRNAs normally target mRNAs, miRNA-373 was reported to bind promoter regions of CDH1 and CSDC2 in PC3 (a human prostate cancer cell line) cells and induce their expression through an unknown mechanism . miRNAs can also increase the translation of certain mRNAs for example, under amino acid starvation conditions, miRNA10a was reported to bind the 5′UTR of ribosomal protein mRNAs and enhance their translation . miRNAs have multiple target mRNAs [140, 141], and, thus, if a mutation leads to mRNA degradation, the miRNAs targeting the affected gene will become available to modulate other targets (Fig 1B).
Since these models rely on the generation and potential degradation of mRNAs from the mutated locus, one would not expect upregulation of potentially compensating genes in the absence of active transcription of the mutant mRNA. It will thus be important to assess transcriptional adaptation in alleles where an mRNA is not produced.
Other potential mechanisms for the compensatory response.
This brief section will focus on increased translational response following the mutational loss of specific genes and will evoke processes such as mRNA modifications and upstream open reading frames.
In response to stress (such as heat shock), pseudouridylation or N6-methylation of adenosines (m6A) was reported to be enriched on certain mRNAs, thereby increasing their stability or promoting their translation [142, 143] (reviewed in ). However, as is the case for DNA methylation, there has been no report thus far about mRNAs from selective loci being modified in this manner.
Upstream open reading frames (uORFs) are regulatory elements present in the 5’UTRs of around 50% of vertebrate mRNAs [145, 146]. They may act as translational repressors, as the translation of the uORFs can occur at the expense of that of the mRNA’s coding sequence [147, 148]. Under cellular stress conditions, there is a tendency to inhibit global translation by phosphorylating eIF2α, which then acts as a competitive inhibitor of the translation initiation factor eIF2B, thereby reducing translation reinitiation rates . This mechanism may allow the increased translation of certain mRNAs under cellular stress. For example, the yeast transcription factor gene GCN4 has 4 uORFs and under normal conditions, the 4 uORFs are translated with less reinitiation at the main ORF. Under nutritional stress, the first uORF is translated efficiently however, due to eIF2α phosphorylation, the remaining uORFs are poorly translated, and reinitiation only occurs at the main ORF, thereby increasing GCN4 production . It is thus possible that certain gene mutations induce cellular stress, allowing for uORF skipping and increased translation of compensating genes. However, as is the case for the DNA and RNA methylation modifications mentioned above, it is not clear how specificity, in terms of selective proteins being upregulated, would arise.
The mechanism of gene silencing makes genes inactive, now we know it. But it may sometimes cause adverse effects even in plants. That is why the gene manipulation, gene editing and gene silencing techniques must be used by prior permission.
Scientists are now almost ready with the novel gene silencing approach mediated by RNA interference for Huntington’s disease. Also, various approaches to different genetic diseases are now under the research phase.