15 replies to this topic

[neuropath]

Hi Bioforumers,

I have a puzzling situation on my hands. I am working with embryonic fibroblasts derived from knockout (KO) mice for a particular gene. These cells showed clear differences in behavior compared to wild-type cells, e.g. they proliferate faster. In order to find out the basis of these differences, I performed microarray analyses to compare wild-type and KO transcriptional profiles. To my big surprise, the target gene for the knockout was not expressed in both wild-type and KO cells. However, I do see distinct differences in the transcriptional profiles between the two types of cells and this is reproducible in a knockout in a different exon created independently in a lab located in another country. If the target gene is not even expressed, what then could account for the differences in transcriptional profiles between wild-type and KO?

If any of you have encountered similar situations, hope you can share some wisdom.

Thanks.

Neuropath

Edited by neuropath, 08 December 2010 - 03:59 AM.

[bob1]

Where did the wild-type cells come from (same line of mice, siblings of the KO mice, skin...)?

[neuropath]

bob1, on 09 December 2010 - 11:32 AM, said:

Where did the wild-type cells come from (same line of mice, siblings of the KO mice, skin...)?

Hi bob1,

Yes, the wild-type cells came from the same line of mice as the KO. I did PCR genotyping and saw deletions in the KO cells are as they should be. Bear in mind that I am seeing the same phenomenon in 2 different lines of mice. The KO cells from the two lines share highly similar gene expression profiles. Any ideas?

[bob1]

Are the wild type cells MEFs as well?

Is the target gene cell cycle dependent?  Low level expression (check with qRT-PCR)?  How was the KO generated?

[HomeBrew]

Believe the biology.  Clearly, this is an anomaly of the  microarray analyses  --  how did you do them?  How many technical and biological replicates did you do?  What statistical analysis says the gene is not expressed?  Not expressed compared to what?

[neuropath]

bob1, on 12 December 2010 - 02:10 PM, said:

Are the wild type cells MEFs as well?

Is the target gene cell cycle dependent?  Low level expression (check with qRT-PCR)?  How was the KO generated?

Hi bob1,

Yes, the wild-type cells are MEFs as well. The wild-type and KO MEFs were prepared and cultured the same way.

So far, there is no evidence to suggest that this gene is cell cycle dependent and it is not known to be involved in any cell cycle regulated functions. However, it is something worth checking. The probe intensities for the gene is either at background value or even negative relative to background. Housekeeping genes levels are normal and all the QC analyses are ok so I do not doubt the quality of the microarray data. I have done semi-quantitative RT-PCR which but did not get the expected band size even at high cycle numbers. I am going to try doing RT-PCR on freshly isolated MEFs and see if the gene was shut down during in vitro culturing or passaging. However, it would still not answer the question of why there are differences between the transcriptional profiles of WT and KO.

KO for both lines (two differenct exons) were generated by cre-lox excision. The KO were not done in my lab but in 2 established labs, one in the US and the other in Japan.

Edited by neuropath, 13 December 2010 - 05:23 PM.

[neuropath]

HomeBrew, on 12 December 2010 - 05:45 PM, said:

Believe the biology.  Clearly, this is an anomaly of the  microarray analyses  --  how did you do them?  How many technical and biological replicates did you do?  What statistical analysis says the gene is not expressed?  Not expressed compared to what?

Hi HomeBrew,

I did two biological replicates and two technical replicates for each on the Illumina Mouse WG-6 Beadchip. Probe signals with detection p-value <0.05 was normalized and background subtracted using GenomeStudio. It is basic but it is sufficient to give me information about the expression of the gene. The signal intensities for the 4 different probes for the gene are at either background or below background levels. All the QC analyses checked out fine. From my experience with other Illumina beadchip experiments in the past, I know that the data tend to be quite reliable.

I was hoping someone out there would have encountered a similar phenomenon - i.e. KO target gene not expressed in both WT and KO yet there are phenotypic/functional/transcriptional differences between the two. My search of the literature has yielded nothing so far. It is even difficult to come up with a good search string

[HomeBrew]

neuropath, on 13 December 2010 - 08:00 AM, said:

...KO target gene not expressed in both WT and KO yet there are phenotypic/functional/transcriptional differences between the two.

This is a self-contradictory statement -- if a gene is not expressed in WT, how can knocking it out have any phenotypic effect?  That is why I was asking about how you concluded it was not expressed in wild type...

[neuropath]

HomeBrew, on 13 December 2010 - 09:31 AM, said:

neuropath, on 13 December 2010 - 08:00 AM, said:

...KO target gene not expressed in both WT and KO yet there are phenotypic/functional/transcriptional differences between the two.

This is a self-contradictory statement -- if a gene is not expressed in WT, how can knocking it out have any phenotypic effect?  That is why I was asking about how you concluded it was not expressed in wild type...

That is the whole purpose of my post......to shed some light on this paradoxical phenomenon. Signal intensities from WT are either zero or negative values, just like KO. The data has been normalized.

There has been a new development. I have found some papers to support my observation that this gene is not expressed in MEFs. This gives us some assurance about the validity of the microarray data. So the question remains......why are there differences in transcriptional profiles between WT and KO when the target gene is not even expressed in WT cells?

[HomeBrew]

Did the papers you found also rely on microarray expression profiles to conclude that this gene is not expressed in wild type?

I can't get past the fact that if a gene is not expressed in wild type, then it can have no influence on wild type phenotype, thus knocking it out would also have no effect.

Perhaps the Illumina probes for this gene are faulty?

[dpo]

How was the knock-out made? From what I read, I'm afraid your knock-out not only affected your gene of interest, but also has an effect on other genes. For instance, by deleting a part of the chromosome required for attachment to the nuclear lamina, you change the position of the chromosome in the nucleus, which can have an effect on the expression of other genes on that chromosome. Alternatively, you may have deleted an enhancer/repressor for another gene located some kilobases/megabases away from your gene of interest. Of course, this is all very hard to prove, but I would recommend checking the conservation of the intronic regions you deleted. If they are nicely conserved between different species, this may imply a functional role for these sequences...

[96well]

Perhaps the genomic region deleted in your model does not code only for a mRNA, but also for a miRNA. A differential miRNA expression could clearly influence your microarray results as well as your phenotype (i.e., different cell growth).

[neuropath]

dpo, on 14 December 2010 - 01:20 AM, said:

How was the knock-out made? From what I read, I'm afraid your knock-out not only affected your gene of interest, but also has an effect on other genes. For instance, by deleting a part of the chromosome required for attachment to the nuclear lamina, you change the position of the chromosome in the nucleus, which can have an effect on the expression of other genes on that chromosome. Alternatively, you may have deleted an enhancer/repressor for another gene located some kilobases/megabases away from your gene of interest. Of course, this is all very hard to prove, but I would recommend checking the conservation of the intronic regions you deleted. If they are nicely conserved between different species, this may imply a functional role for these sequences...

Hi dpo,

Two different knockouts were created by deletion of two separate exons. Your insights are perceptive. There is a gene that is transcribed in the reverse orientation to my gene of interest and they share a very short bidirectional promoter. Their expression patterns are quite similar in vivo. I've looked at the expression of this gene in my microarray data but unfortunately, it too is not expressed in both WT and KO. Your suggestions however now give me new leads to pursue.....I will see if any of those differentially expressed genes are located in or around the loci of my gene of interest. If your prediction is correct, it should show up in the data. I will look and keep you posted. Thanks for the great suggestions

Edited by neuropath, 14 December 2010 - 07:31 AM.

[neuropath]

96well, on 14 December 2010 - 05:50 AM, said:

Perhaps the genomic region deleted in your model does not code only for a mRNA, but also for a miRNA. A differential miRNA expression could clearly influence your microarray results as well as your phenotype (i.e., different cell growth).

Hi 96well,

Your suggestion is a great one too. I am now considering doing a microRNA microarray. I also recently learned about LINC - large intergenic non-coding RNAs. It shows how much remains to be discovered about gene regulation. A colleague of mine suggested another possibility......that the knockout effect was imprinted in the genome and passed on down the germline through methylation or acetylation of genes or chromosomes. This will be really tough to prove, although not impossible. Thanks for your thoughts  

[Radish]

well, it is not that hard to prove if it is an epigenetic event, you can just go with bisulfide sequencing or chip for repressive histone marks (if it is MEFs the raw data is already public, you just have to download it and analyse the epigenetic marks for your gene).

The thing is, if it is not expressed in MEFs why analyse the KO in the first place? Is it expressed in other cell lines?
It is ok for it not to be expressed in wt MEFs, it is just weird the phenotipical differences, we do a bunch of KO mice in our lab and I never seen anything like that (except when certain areas in the 3'UTR or other regions involved in regulation got removed by mistake).

best regards
Radish