Gene regulation at the single-cell level

From Wiki FKKT
Revision as of 22:57, 8 February 2015 by Katarina Uršič (talk | contribs) (New page: Rosenfeld, N., Young, J. W., Alon, U., Swain, P. S., & Elowitz, M. B. (2005). [http://www.sciencemag.org/content/307/5717/1962.full Gene regulation at the single-cell level.] Science 307 (...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search

Rosenfeld, N., Young, J. W., Alon, U., Swain, P. S., & Elowitz, M. B. (2005). Gene regulation at the single-cell level. Science 307 (2005), 1962–1965].

In this paper I will present you »Gene regulation at the single-cell level« written by Nitzan Rosenfeld, Jonathan W. Young, Uri Alon, Peter S. Swain and Michael B. Elowitz. Article was published in Science in year 2005. Currently it can only be accessed after free registration on the Science web page.


REGULATION OF GENE EXPRESSION


Regulation of gene expression is necessary for maintaining cell autonomy. Organism from bacteria to mammals use it for adjusting to different internal and external environmental conditions while multicellular organism also use gene regulation for developing different cell types and tissue formation. Regulation of gene expression is a demanding task, the right genes have to be expressed at the right time, rate and at the right place. In prokaryotes gene expression is usually managed by regulation of transcription. Cells can regulate the activity of already made proteins or they can alter the production rate by switching gene on or off. Operators act like a switch, they can enable the binding of RNA polymerase to promoter or not and therefore determine the fate of a gene transcription. Operons are under the influence of proteins made by regulatory genes. Repressor proteins can switch the operon off by binding to operator and preventing the beginning of transcription. There is a high level of specificity between regulatory proteins and operators, not all regulatory proteins can bind to all operator binding spots. Usually regulatory proteins are under control of constitutive promoters and are always expressed at least in small amounts. Binding or regulatory proteins is reversible, bigger expression rate of active regulatory proteins result in longer duration of operator state whether it be on or off state. Regulatory proteins are also often allosteric, present in active or inactive shape. Some need additional small molecules (corepressors or coactivators), that bind on them and then together can make a switch in expression. Roughly we can divide gene regulation into positive and negative. In positive gene regulation regulatory protein acts like an activator and its direct binding to DNA triggers transcription of a gene by increasing the affinity that RNA polymerase has for promoter. An example would be catabolite activator protein and its effect on lac operon. In negative gene regulation regulatory gene acts as a repressor and by binding to operator it turns the transcription of a gene off. Negative gene regulation can be preformed via repressible or inducible operons. Example of a repressible operon is trp operon which is usually active but it can be repressed when trp repressor and corepressor tryptophan bind together on its operator. Example of inducible operon is again lac operon. Repressor protein lacI binds to operator in the absence of lactose and prevents transcription of lac operon genes. When allolactose is present it acts as an inducer that inactivates the repressor protein (Campbell & Reece, 2008). Eukaryotic gene expression regulation also strongly relies on regulation of transcription, based on the function of transcription factors similar to that in prokaryotes. In addition gene expression can also be regulated by changes in chromatin structure organization via histone modification and DNA methylation. Tails of histones can be acylated which loosens chromatin structure and enables transcription since transcriptional factors have easier access to DNA. Acetylaton neutralizes the positive charge of lysins and prevents binding of nucleosomes together. Other chemical groups can be attached to histone tails as well for example attachment of methyl group can have the opposite effect as it can promote chromatin to condensate. Methylation of cytosine can also help inactivate DNA and therefore lowers the expression rate of certain genes. Experiments also showed the effect of condensed heterochromatin on repression of gene expression. Transcription however is not the only factor determining the rate of gene production where we measure the actual amount of functional protein. When dealing with eukaryotes we have to take into consideration RNA processing such as alternative RNA splicing, speed of mRNA degradation in cytoplasm, initiation of translation which can be blocked by regulatory proteins that prevent binding of mRNA and ribosome and lastly the processing and degradation of protein (Campbell & Reece, 2008).

Not only is gene regulation important for adequate cell functioning, it is also a major factor in maintaining the proper functioning of synthetic genetic circuits. There is still a great lack of knowledge for building and maintaining complex, long term sustainable genetic networks that could have a potential use in real life applications. It is important to understand and characterize new parts and modules, model the behavior of synthetic networks and study the in vivo interactions between parts in a synthetic network. Many biological parts and interactions between them are still not well understood so synthetic networks often base on similar parts. New parts that show great promise are engineered zinc fingers and nucleic-acid-based parts. At single molecule level advanced technology enables direct ways to observe the state of each molecule. In vivo measurements cannot be performed that way because of the lack of technology. Currently the most promising development is in the field of microfluidic devices but the problem is the absence of reliable sensors (Lu, Khali3, & Collins, 2010).


INTRODUCTION TO EXPERIMENT


Rosenfeld et al. generated »lambda cascade« strains of E. coli. Bacteriophage lambda is a temperate phage meaning it can reproduce inside the bacteria via lysogenic or lytic cycle. They decided to characterized its PR promoter. In natural conditions, cI repressor protein maintains lysogenic cycle. It represses transcription from PR promoter by binding to one of the three binding sites OR1, OR2 or OR3 and preventing the expression of lytic genes. cI has the highest affinity for OR1 site but after cI dimerisation the affinity for OR2 increases too. Affinity for site OR3 is low and cI only binds when present in high concentrations (Bakk, Metzler, & Sneppen, 2004). At the time that Rosenfeld et al. decided to performed the experiments models predicted gene regulation function (GRF) with the help of in vitro data. GRF is usually presented in a form of a graph with active transcription factor on the x axis and production rate on y axis. Due to many disadvantages authors decided to measure GRF in vivo, using lambda cascade strains of E. coli and a point mutation OR*-lambda cascade strain of E. coli. This designed lambda strain contained a fusion protein made from yellow fluorescent protein (YFP) and a repressor protein cI under control of a constitutive Ptet tetraycylin promoter. Ptet promoter can be repressed by TetR under PC promoter control and induced by tetracyclin or anhydrocyclin (aTc) (Fig. 1B). cI acts as a repressor of PR promoter, that controls transcription of a cyan fluorescent protein (CFP). By constructing this genetic circuit they were able to simultaneously monitor and measure the amount of repressor protein and its target gene protein in time simply by measuring fluorescence (Rosenfeld et al., 2005).


REGULATION DILUTION EXPERIMENT


GRF tells us how production rates of a protein vary at different repressor concentrations therefore authors of the original paper had to find a way to systematically change it's concentration. To answer that problem they used a regulator dilution method. During the cell growth repressor production shuts off and as the volume of that cell increases, concentration of repressor decreases. Repressor also gets diluted by cell division. Fig. 1C)shows a graphic representation of regulator dilution experiment, showing us the fluctuation in concentration of cI-YFP and CFP over time (stated in cell cycles). Under normal conditions, TetR blocks cI-YFP production and cell express CFP under control of PR promoter. By adding aTc cells start to produce cI-YFP, which represses production of CFP as shown in Fig. 1C. Once aTc is no longer added, TetR represses production of cI-YFP so its concentration decreases and we reach additional dilution with each cell cycle as the cells grow and divide. Less of cI-YFP means less repression of CFP production and once cI-YFP levels decrease to a certain value, it can no longer efficiently repress cfp gene, CFP concentration indeed rises with each cell cycle. Fig. 1D shows regulator dilution experiment with OR2*-lambda cascade strain, that contains an OR2 point mutation which was made with the help of site-directed mutagenesis of the PR promoter (Fig. S3). Red represents cl-YFP and green represents CFP. Each snapshot shows time in minutes. With the help of fluorescent time-lapse microscopy (Fig 1D, Fig. S1, Movie S1 and S2) and computational image analysis they managed to form family trees for cells in microcolonies (Fig. S2). They then quantify the level of cI-YFP and CFP over time for each cell lineage and then calculated at what rate CFP was being produced (shown in Fig 2A) with level of cI-YFP on x axis and rate of CFP production on y axis. With the use of fluorescence information and cell family trees they managed to calibrate biochemical parameters with a new technique using binomial errors in protein partitioning. They compared levels of fluorescence of a sister cell pair just after the division. Since there is an equal possibility in cI-YFP molecule going in each sister cell, distribution of CI-YFP fluorescence was binomial. With the help of a single parameter fit they managed to estimate how much of detected fluorescence belonged to a single CI-YFP molecule. By calculating partitioning errors in CI-YFP distribution calibration was possible despite cellular autofluorescence. Binomial errors in partitioning helped find the apparent fluorescent intensity of one independently segregating fluorescent particle vy 2B). If a cell contained Ntot fluorescent particles, total fluorescence is Ytot=vy*Ntot. When a cell divides, sister cells gets N1 and N2 particles (Ntot=N1+N2). Based on the assumption that the segregation of particles is independent the segregation becomes binomial and satisfies the formula, used in mathematical model. Fig. 3 shows results (3A shows mean GRF and 3B shows individual data) for PR and PR (OR2*) point mutation strain. Unknown regulation functions are often represented by Hill functions f(R) = β/[1 + (R/kd)n]. β is the maximal production rate, kd is a concentration of repressor that yields half-maximal expression and n is the degree of effective cooperativity in repression. When compared to data, measured in vivo, function fitted well (Fig 3, table 1). In vivo kd value was comparable to kd value estimated in previous studies. The effective cooperativity value (n>1) could be explained with cI repressor molecules dimerisation and interactions of repressor at neighboring sites. They also compared biomechanical parameters in PR and PR(OR2*) strain and found some differences. In PR strain, n is significantly increased and kd is significantly reduced (table 1). PR (OR2*) strain contained a mutation of a binding site, which could prevent cI from binding or lowered its affinity for cI and therefore we would need higher repressor concentration to achieve the same results as in PR strain. After observing the mean regulation function of wild type PR and its mutated variant OR2* they wanted to further investigate deviations from the mean GRF. Standard deviations of CFP production rates was about 55% of the mean GRF for any given cI-YFP concentration (Rosenfeld et al., 2005).


STANDARD DEVIATION TESTS AND CELL NOISE


There are many possible reasons for variations like this, from differences in environment of the cells, changes in gene copy number or noises. Noise causes differences even in identical cells or organisms. With molecules that are present within a cell in a large number, reactions can be predicted. That is however not possible with molecules like DNA or some proteins that are presented in a fewer copies per cell and can be influenced by random stochastic fluctuations. Noise could be a reason for reduced preciseness of cellular processes by effecting gene expression. Variation of gene expression can have its source in four main factors: differences in environment, mutations, differences in inner cell environment and stochasticity related to molecules with small copy number per cell. Noise is usually measured with the help of fluorescent proteins and is defined as the ratio between standard deviation and the mean value. Amplitude of noise is controlled by genetic factors, rate of transcription and regulatory dynamics (Raser and O'Shea, 2005). Intrinsic noise comes from biochemical reactions at genes, consequently changing the expression rates of two identical copies of gene. Extrinsic noise comes from differences in rates of cellular components. The difference in expression proteins can be contributed to intrinsic noise such as protein degradation, fluctuations in the amount of mRNA or different levels of promoter-biding. Extrinsic noise has the same effect on expression of both proteins inside a cell but effects different cells differently (differences in environment, concentration and activity of trancriptional factors etc.) extrinsic noise can be further divided into global and specific noise. Global noise comes from fluctuations in basic reactions that effect the whole cell. Specific noise originates in fluctuations of a factor that only effects a specific gene or pathway (Raser and O'Shea, 2005). Measuring of noise and determining whether its intrinsic or extrinsic could be done by comparing expression of two fluorescent proteins under identical regulation in the same cell. This two-reporter method was developed by Michael B. Elowitz. With coworkers they observed protein expression under control of identical promoters on the same E. coli chromosome which enabled them to distinguish between the two main types of noise. Intrinsic noise was measured by comparing the amount of both fluorescent proteins, CFP and YFP, produced in a single cell. Effect of extrinsic noise was observed when production rates of both proteins were the same for a single cell but varied among multiple cells (Elowitz et al., 2002). Studies confirm that the dominant effect of global extrinsic noise over intrinsic noise and is a major source of variability between cells. Overall effect of noise on cells depends on the degree and the recurrence of noise. Time is an important factor, the longer the endurance, the greater the effect can be. Evolutionary speaking, noise can have a positive or a negative impact, intrinsic noise can increase the expression of different allele combinations and therefore increases phenotypic variability. In a stressful situations or change in environment, some cells could potentially adapt better than other or there could be opposite response (Raser and O'Shea, 2005). Noise can effect an entire genetic network and can be amplified. A study, made by Hooshangi et al. showed how effect of noise increases with the complexity of genetic circuit. They designed three circuits, each with one additional transcriptional step in a regulatory cascade. What they concluded is that the longer the cascade, the bigger the cell variability and coordination of cell responses reduced (Hooshangi, Thiberge, & Weiss, 2005). Rosenfeld et al. tested the environment possibility by comparing different microcolonies that were induced at different cell densities but no significant changes in GRF were detected (Fig. S6). Next they analyzed how the increased gene copy number during DNA replication/cell division would affect GRF. The results showed strong correlation, newly divided cells produced less CFP than cells just prior to division. Even with that taken into consideration, standard deviation was still around 40 % of mean GRF after normalizing production rates of CFP to an average phase of a cell cycle. Lastly they tested effects of intrinsic and extrinsic noise.


SYMMETRIC BRANCH STRAIN TEST


To find out the origin of fluctuations, they designed a »symmetric branch« strain (Fig. 4D, movie S3), similar to one mentioned before, to help them distinguish between intrinsic and extrinsic noise. Symmetric branch strain of E. coli produced YFP and CFP under control of two identical PR promoters, both regulated by cl-YFPY66F repressor protein (Fig. 4D). cI-YFPY66F gene contained a mutation made with the help of site-directed mutagenesis. They introduced a single point mutation which changed the tyrosine at YFP position 66 to phenylalanine which prevented repressor fluorescence so the only YFP fluorescence detected came from the YFP under PR control and not the fusion repressor protein. After measuring the expression of both fluorescent protein genes by measuring fluorescence for each cell, results concluded that the difference in production rates is bigger because of extrinsic noise (intrinsic noise played a role too). Extrinsic noise has bigger effect on the difference in production rates than intrinsic noise (about 35 % versus 20 %). Because they quantified the extrinsic noise at known repressor concentration, the only fluctuations measured are those of global cellular components. Analysis of extrinsic noise is also more precise because of its dynamic observation – »measuring in the rate of protein expression« (Rosenfeld et al., 2005). Different paced fluctuations affect genetic networks in different ways. One way to characterize fluctuations in by τcorr, their autocorrelation time. If a fluctuation is longer than the length of a cell cycle, it can accumulate and produce significant effects whereas short term fluctuations (much shorter than a cell cycle) have smaller effects as they even out. Authors divided their measurements into three categories (Fig. 4A-C): fast fluctuations similar to intrinsic noise; periodic DNA replication related fluctuations and aperiodic fluctuations, similar to extrinsic noise. When comparing mean GRF with a selected single cell lineage lambda, we can see irregularities (Fig. 3B) with autocorrelation time around 40 minutes (Fig. 4E) which is close to the length of a cell cycle (around 45±10 minutes). Therefore most fluctuations persist for about one cell cycle (whether it be overexpression or underexpression of CFP). These observed fluctuations are too slow to be classified as a consequence of intrinsic noise (which has autocorrelation time under 10 minutes). As we're modeling genetic circuits whether they be natural or synthetic, we must take extrinsic fluctuations into consideration (Rosenfeld et al., 2005).


CONCLUSION


Authors of original article successfully characterize lambda promoter, in vivo calibrated biochemical parameters using novel techniques and characterized the effect of extrinsic and intrinsic noise. GRF efficiency of a cell is determined by various factors and when constructing a genetic network it is crucial to take that into consideration.


References:



Campbell, N. A., & Reece, J. B. (2008). Biology 8th edition. San Francisco: Pearson Benjamin Cummings. 351-364.

Bakk, A., Metzler, R., & Sneppen, K. (2004). Sensitivity of OR in phage lambda. Biophysical Journal, 86(January), 58–66. doi:10.1016/S0006-3495(04)74083-7

Elowitz, M. B., Levine,A. J., Siggia,E. J., & Swain P. S. (2002). Stochastic gene expression in a single cell. Science (New York, N.Y.), 297(5584): 1183-1186. doi:10.1126/science.1070919

Hooshangi, S., Thiberge, S., & Weiss, R. (2005). Ultrasensitivity and noise propagation in a synthetic transcriptional cascade. Proceedings of the National Academy of Sciences of the United States of America, 102(10), 3581–6. doi:10.1073/pnas.0408507102

Lu, T. K., Khali3, A. S., & Collins, J. J. (2010). Next-Generation Synthetic Gene Networks. Nat Biotechnol, 27(12), 1139–1150. doi:10.1038/nbt.1591.Next-Generation

Raser, J. M., & O'Shea, E. K. (2005). Noise in gene expression: origins, consequences and control. Science (New York, N.Y.), 309(5743): 2010–2013. doi:10.1126/science.1105891

Rosenfeld, N., Young, J. W., Alon, U., Swain, P. S., & Elowitz, M. B. (2005). Gene regulation at the single-cell level. Science (New York, N.Y.), 307(2005), 1962–1965. doi:10.1126/science.1106914