|
Detection of Mycoplasma Infection in Cultured Cells on the Basis of Molecular Profiling of Host Responses
Adaptive responses to diverse microbial pathogens might be limited in relatively few types. Host cell responses to
pathogens are believed to be patterned or stereotyped along with species or class. We tried to compose the host response to Mycoplasma in terms of cellular gene expression. Although gene expression profile of two host HeLa and 293 cells were quite different each other, 30 genes were differentially expressed by mycoplasma infection in both of HeLa and 293 cells. Six of them (PR48, MADH4, MKPX, CRK, RBM7, NEK3) were related to cell cycle or proliferation. Another category of genes like IL1HY1, KLRF1, TNFSF14, GBP1 were host defense to elicit immune responses. With this set of genes, we establish the prediction model for mycoplasma contamination.
The Mycoplasma, with over 100 different species, are the smallest self-replicating organisms known at present and constitute a distinct class within the prokaryotes characterized by their lack of a rigid cell wall. They can be classified into fermentative strains, which gain energy by fermentation of carbohydrates and non-fermentative strains that are unable to metabolize carbohydrates via glycolysis. The mycoplasmas are extra cellular parasites usually attached to the external surface of cells, but can also penetrate these (Razin etal., 1998). In humans, M. pneumoniae is a frequent cause of respiratory infections, and is at the origin of approximately 20 % of all community- acquired pneumonias (Hammerschlag, 2001). Being
'minimal cells', Mycoplasma have also been used to investigate the machinery of self-replicating organisms (Razin, 1992).
Beside health problems, mycoplasma contamination constitutes one frequent problem when studying cultured cells (estimated frequency varying from 5 to 35 %). The strains M. hyorhinis, M. orale, M. arginini, M. fermetans, M. hominis and Acholeplasma laidlawii represent 90-95 % of the contaminating isolates (Hay et al., 1989). Contamination is initially often difficult to detect because the contaminated culture grows well and appears normal by ordinary light microscopy. In addition, Mycoplasma is highly contagious and can rapidly spread through the cell stocks. The possible consequences of mycoplasma infection for the host-cells are multiple and variable, ranging from no apparent effect to extensive changes with inhibition of cell proliferation, induction of apoptosis, induction of cytokines and oxidative radicals, and malignant transformation (Avron and Gallily, 1995; Sokolova etal., 1998; Tsai etal., 1995). There is also a possibility that mycoplasmal biological activities may erroneously be interpreted as being of host origin (Choi et al., 1998).
Understanding the molecular basis of a host’s response to microbial infection is essential for preventing
disease and tissue damage as a result of the inflammatory response. A better understanding of this process should allow for the design of drugs that can more specifically and
effectively target infected cells with reduced side-effects. The host-pathogen interaction can result in many changes to the host cell including modulation of RNA expression, target receptor induction, actin cytoskeletal rearrangements, signal transduction pathway activation, and vacuolar trafficking (Finlay and Cossart, 1997; Cummings and Reiman, 2000). DNA microarray enabled us to describe a unique biological phenomenon in terms of genome-wide gene expression analysis. It can provide a bird-eye view on a certain phenomenon as well as complete list on the significant genes. Gene expression profiling using DNA microarray has been offering potentials for defining the patterns especially on gene expression during normal biological or disease processes. Moreover, many of differentially expressed genes that may play an integral role in these processes were identified.
In this paper, we have utilized spotted oligonucleotide microarray to examine the expression of 10,416 known
regulatory genes upon mycoplasma infection. We compare the distribution of patterns of gene expressions in two different cell lines, HeLa and 293 cells. After that we select 30 genes that expressed differentially in both of HeLa and 293 cells, as biomarkers for Mycoplasma infection. Actual prediction model for the infection is also discussed.
We used HeLa and 293 cells were maintained in Dulbeccos modified Eagle medium (DMEM, JBI, Seoul, Korea) supplemented with 10 % fetal bovine serum (JBI),
1 % Penicillin-Streptomycin (Invitrogen). Cells were plated in concentration of 2-5x10 6 per T75 flask 18 h before mycoplasma infection. M. hyoriniswere harvested at log phase to treat onto cells at 10 8 per T75 flask. After 48 h, cells were washed with phosphate buffered saline and harvested for RNA preparation.
Total cellular RNA was prepared using RNAeasy total RNA purification kit (Qiagen, Valencia, CA) followed by treatment with RNase-free RQ DNase (Promega Corp., Madison, Wl). For cDNA synthesis, random hexamer primers (Gibco/BRL) were used to prime reverse transcriptase (RT) reactions. Using this method it was possible to use the same RT reaction (cDNA) for PCR amplification with different sets of gene-specific primers. The cDNA synthesis was carried out using Moloney murine leukemia virus (M-MLV) Superscript II reverse transcriptase (Gibco/BRL) following the manufacturer's instructions. To analyze relative expression of different mRNAs, the amount of cDNA was normalized based on the signal from ubiquitously expressed actin mRNA.
Total RNAs from control or Mycoplasma-infected HeLa or 293 cells used as to extract total RNA for quantipli- cated experiment including one dye-swapping set.
Total RNA was prepared by using TriZol reagent (Gibco-BRL). The array used in this experiment was MacArray (Oligo-Human 10K, Macrogen Inc., Seoul, Korea) according to the manual provided by manufacturer. Briefly, 100 /zg of total RNA was labelled by incubating it with Cyanine-3-dUTP (6.0 mM) (Perkin Elmer Life Sciences) or Cyanine-5-dUTP (4.0 mM) (Perkin Elmer Life Sciences), dNTP mixture, 0.1M DTT, RnaseOUT, inorganic pyrophosphatase, reverse transcriptase at 40 °C for 3 hours. Before hybridization, 8 m of Cyanine-3-CTP
labelled cRNA and 8 <zg of Cyanine-5-CTP labelled cRNA were mixed together with 2.5 id of Mouse Cot-1 DNA (Invitrogen), 2.5 id of Deposition control target (Operon Technologies), and 12.5 id of 2Xhybridization buffer (Agilent Technologies). The mixture was centrifuged at 12,000 rpm for 3 min after it was boiled for 2 min. About 24 id of the mixture was hybridized on the array at 65 °C for 16 h. After hybridization, the array was washed with solution I (0.5xSSC, 0.01% SDS, and 1 mM DTT in nuclease-free water), solution II (0.25xSSC, 0.01% SDS, and 1 mM DTT in nuclease-free water), and solution III (0.06xSSC and 1 mM DTT in nuclease-free water) respectively. Cy3 and Cy5 fluorescent intensities were determined using the GenePix scanner (Axon Instruments), and images were analyzed using the built-in program to calculate relative ratios and to determine confidence intervals.
Fluorescence intensity was processed and measured using GenePix Pro software (Axon Instruments). Intensity data were imported to an in-house microarray database. Variance stabilizing normalization by Huber et al. was applied with the ‘vsn’ package in Bioconductor using the R statistical package. After performing intensity-dependent global LOWESS regression, spatial and intensity dependent effects were managed by pin-group LOWESS normalization following by the approach of Yang et al. A variation filter was used to reduce the number of genes for the following cluster analysis. Specifically, we applied the non-parametric Kruskall-Wallis one-way analysis of variance to eliminate the genes that did not change significantly across samples.
Host responses to mycoplasma infection were tested in
two different human cell lines, HeLa and 293, which is usually used by molecular biology experiments. Subconfluent cells at 2-5x10 6 cells per 100 mm dish were incubated with mycoplasma pellet at the concentration of 10 8 per 10 ml media. Gross morphology of each cells was not significantly changed upon microbial infections. To stabilize the host pathogen interaction, we incubated cells with Mycoplasma for 48 h. The transcripts of infected cells were compared with uninfected control cells.
Overall expression similarities for each samples of microarray data is provided in Fig. 1. These measures in the table are calculated as a pattern-similarity of two samples. The absolute value of Pearson correlation coefficient is used for the pattern-similarity. According to
the figure, we observe that characteristics of host response to mycoplasma have remarkable difference in two different cell lines.
Selection of biomarkers for Mycoplasma infection
From the quadruplicated profile of 10K oligonucleotide microarray in each cells, we selected genes for the host
responses to mycoplasma infection under the following standards (Table 1):
i) Standard deviation in both of infected and control groups should be less than two fold in both cell lines, ii) Average fold activation of infected groups should exceed two fold, iii) Average fold activation of uninfected control group should be less than two fold, iv) Activation of gene
expression should be found in both of HeLa and 293 cells. According to the GO ( http://www.geneontology.org/ ), we could classify these 30 genes in their biological functions (Table 1). In view of biological effect of Mycoplasma infection, we found that two categories of genes like cell cycle/proliferation and immune response were changed upon the infection. Especially Mycoplasma leads to the growth retardation or arrest in vitro, which might be explained by the transcriptional activation of genes like PR48, MADH4 and MKPX. Although two cell lines were not from immunological tissues, we detected the increase of transcripts like IL1HY1, KLRF1, TNFSF14 and GBP1.
Prediction model for Mycoplasma contamination
The 30 genes selected significantly, listed in Table 1, are believed that they are strongly involved the process of mycoplasma infection in human cells. From this fact, we define the contamination index for a given sample of
microarray, as follows :
where {g1,..., g30} is the set of gene in Table 1, M(g,s)
is the log-ratio of expression of g under the sample s, and it is the
function t hat indicate the
probability of a gene with expression value x became a differentially expressed gene.
Since these genes are differentially expressed in both
cells, we may consider that the contamination index Cl(s) reflects that whether the unknown sample s is infected by Mycoplasma or not. To give a specific criterion for mycoplasma contamination, we calculated the contamination indices with respect to the randomly selected genes. Mean values and standard deviations of the indices for each samples are extracted by iterating these process 1,000 times. Parameters in random model are listed in Table 2.
We set a criterion to predict contamination by Mycoplasma
for a given sample s as follows:
Cl(s) > m + 3V,
where m and a are mean and standard deviation in random model, respectively. The threshold “m+3V” is taken to satisfy that the false positive rate is less than 0.01. That is, the probability of Cl(s) > m+3V when s is not contaminated by Mycoplasma is less than 0.01. According to the Table 2, only the case of sample “293 cell-replication 2” fails to predict the mycoplasma contamination out of 8 total samples. In the sample “293 cell-replication 2”, we observe that the contamination index (0.16773) is much larger than the mean contamination index (0.0857) for randomly selected genes, but the standard deviation (0.0431) is relatively large to the contamination index, so the log-expression profile in the sample is far from the normal distribution, which is a basic assumption of our predicting model.
Efforts in functional genomics related to cancer research have yielded major successes in the pursuit of gene expression signatures. Approaches for gene expression analysis, such as time-series analysis, pattern discovery, clustering, and class prediction, have been recently reviewed (Solnim, 2002). Expression-based criteria or class predictors have been defined based on neighborhood analysis (Golub, 1999), a supervised method based on a subset of genes whose expression strongly correlates with specific classes, as well as Bayesian regression models (West, 2001) and artificial neural networks (Khan, 2001). These predictors were successfully used to classify novel samples in a manner consistent with clinical assessments. In fact, classifications based on gene expression alone or class discovery have also been
demonstrated and suggesting that gene expression profiling have the capacity to identify subtypes that have not been previously defined (Golub, 1999). Although these results are promising, one should note that many of the previously conducted cancer line gene expression analyses are one dimensional; in contrast, a host expression profile evoked by pathogen exposure would be expected to be temporal and may also exhibit dose dependence. Comprehensive sets of gene expression profiles that explore temporal and dose ranges for pathogen exposure must be produced to map the continuum of gene expression changes.
In this paper, we suggested a list of genes that is believed to strongly related to the mycoplasma contamination in two different human cell lines, HeLa and 293. We also suggested a model to predict that a given sample is infected by the mycoplasma. This naive process has strong power to predict, and is suitable to the our data which has unique structure that consisted of expression of genes upon infection in various experimental conditions such as cell-lines or dose ranges. To get a more strong prediction power for overall-non-contamination of a concerned sample, we need a structural experimental profiling data for host response of infection by wide source of pathogens.
|
