ABSTRACT
The hybridization signature approach, using colony filters and labeled complex
probes, can provide high throughput measurement of gene activity. We describe
here the implementation of this method to follow the expression levels of 47
genes in resting and activated T cells, as well as in epithelial cells. Using 4-fold spotting of colonies, imaging plate detection and various correction
and normalization procedures, the technique is sensitive enough to quantify
expression levels for sequences present at 0.005% abundance in the probe.
Comparison with Northern blotting shows good consistency between the two
methods. Upon activation of a T cell clone by an anti-CD3 antibody variations ranging from 2- to 20-fold are measured, some of which had not been reported
previously. This `multiplex messenger assay' method, performed using available
commercial apparatus, can be used in many cases where simultaneous assessment of mRNA levels for many genes is of
interest.
Gene indexes with many thousands of entries have been constructed by tag
sequencing of randomly selected cDNA clones (
1
-
8
) and are widely available in repositories such as the db EST database (
9
). As more and more genes are identified, efforts are redirected towards
understanding the control of gene expression that occurs in a strictly ordered
time- and cell-dependent fashion. In order to analyze the expression profiles of a
large number of known genes in the tissue (or cell) of interest we have adapted
a large scale gene expression analysis system recently described by us (
10
). It is based on the hybridization of complex probes with high density colony
cDNA filters followed by quantitative measurement of the amount of hybridized
probe on each colony. A somewhat similar approach using PCR products of
Arabidopsis thaliana
cDNAs has recently been reported (
11
).
Our method, called MMA, for multiplex messenger assay, is applied to the
investigation of differential expression of a set of known T cell genes in
three cell types: a thymic epithelial cell line, a cytotoxic T cell clone and
the same T cell clone stimulated by an anti-CD3 antibody. The activation of T lymphocytes by antigen during an immune
response is mediated by the T cell receptor, which recognizes peptide antigens
bound to self major histocompatibility complex (MHC) molecules on the surface
of an antigen-presenting cell. This stimulation initiates a cascade of biochemical
events that culminate in cellular differentiation and proliferation (
12
). The activation by an anti-CD3 antibody simulates the events observed after activation. The effect of
the mechanisms used by cells at the transcriptional level to regulate the
numerous genes involved in activation (including alterations of transcriptional
rate, termination of transcription and mRNA stability) are quantified in a
single step by our method.
KB5.C20, a CTL clone of B10.BR origin specific for the H-2K
b
alloantigen, was maintained in long-term culture as described (
13
). Samples of 20 * 10
6
cells were grown in RMPI medium with 10% FCS alone or in the presence of
plastic coated anti-CD3[epsilon] (145.2c.11) mAb (
14
) for 3 h.
The MTE-1D epithelial cell line, obtained after MTE cell line subcloning as
described (
15
), was grown in standard DME with 20% FCS.
Most of the cDNA clones used were obtained from an adult mouse thymus cDNA
library (
10
) by hybridization of filters containing part of the library with probes
corresponding to known genes. Others were found among already sequenced clones from the same library. For some
additional clones (including the control
A.thaliana
cytochrome c554 gene) the cDNA insert was transferred from the original cloning vector to that used for the cDNA library (pcDNA1) and then
transformed into MC1061 p3 bacteria to obtain a coherent set of clones in the
same plasmid vector and bacterium. Three clones containing essentially a
poly(A) sequence (50, 60 and 90 bp) were obtained by appropriate digestion of
the poly(A) tail of sequenced cDNAs followed by cloning at the multiple cloning
site of the pcDNA1 vector. In all cases participation of the original insert in
addition to the poly(A) stretch is <20 bp.
Filters were prepared using a BIOMEK 1000 (Beckman) robotics workstation and a
96 pin tool. Colonies from freshly grown replica plates were spotted onto
Hybond N filters (Amersham) (
10
). Each colony was spotted in quadruplicate twice in two opposite symmetrical
areas of the filter (see Fig.
1
A). Filters were subsequently treated as described by Nizetic (
16
).
The
A.thaliana
cytochrome c554 was provided in the pHD-1 vector by Herman Hofte (INRA, Versailles, France).The messenger RNA of this gene was prepared from this cDNA cloned into Bluescript SK+
vector at the
Not
I restriction site and messenger RNA was synthesized from the T3 promotor using
the RiboMax large scale production system (Promega).
Total RNA was isolated from cell lines using the Trizol reagent (Gibco BRL).
Complex probes were prepared from total RNA with an excess of oligo(dT) (
25
) to saturate the poly(A) tails and ensure that the reverse transcribed product
does not contain long poly(T) sequences. Aliquots of 25 [mu]g total RNA, 8 [mu]g dT
25
plus 300 ng dT
12-18
and a defined amount (0.5-5 ng in different experiments) of cytochrome c554 mRNA were mixed, heated to 70oC to remove secondary structure in the RNA and progressively cooled
to 43oC to ensure annealing of oligo(dT) with the poly(A) tail. Complex probes
were then prepared in 25 [mu]l by simultaneous reverse transcription and labeling for 1 h at 43oC in the presence of 50 [mu]Ci [
32
P]dCTP, 5 [mu]M dCTP, 0.8 mM each dATP, dTTP and dGTP and 200 U RNase H reverse
transcriptase (Gibco BRL). RNA is removed by treatment at 68oC for 30 min with 1 [mu]l 1% SDS, 1 [mu]l 0.5 M EDTA, 3 [mu]l 3 M NaOH and then equilibrated at room temperature for 15 min.
Neutralization is with 10 [mu]l 1 M Tris-HCl plus 3 [mu]l 2 N HCl. Unincorporated nucleotides were removed by
purification on a G50 column. The probe (after 5 min denaturation at 100oC) was then incubated with 2 [mu]g poly(dA) (dA
80
) in 1 ml hybridization mix (5* SSC, 5* Denhart's, 0.5% SDS) for 2 h at 65oC. Pre-hybridization and hybridization were both performed for
20 h. After hybridization filters were washed in 2* SSC, 0.1% SDS for 20 min and twice in 0.2* SSC, 0.1% SDS at 65oC for 1 h.
Hybridization of oligomers (labeled at the 5'-end with [[gamma]-
32
P]ATP and kinase) was in 6* SSC, 5* Denhardt's mix, 1% SDS for 15 h at 42oC, followed by two short (2 min) washes in 6* SSC, 0.1% SDS at room temperature. The vector
oligomer sequence used was 5'-GCTTATCGAAATTAATACGACTCACTATAG-3'.
Quantitative data were obtained using an imaging plate device. The hybridized
filter was exposed to an imaging plate for 20-35 h and then scanned in a Fujix Bas 1000 (Fuji) system. Hybridization
signatures were determined with a modified version of the Bioimage software
(Millipore) running on a Unix workstation (
17
). The resulting quantified data were then analyzed on a microcomputer
(Macintosh Centris 650) using Excel software with macro commands that compute
average values for each colony.
Northern blot analysis was performed according to Maniatis
et al
. (
18
) with nylon membranes (Hybond N; Amersham). The same amount of KB5.C20 and MTE-1D total RNA (25 [mu]g) or serial dilutions (26, 13, 8.6, 6.5, 5.2 and 2.6 [mu]g) for sensitivity estimations were loaded on the gels. The
resulting Northern blots were hybridized with probes labeled by random priming
with [[alpha]-
32
P]dCTP using purified cDNA inserts (
19
). Northern blots were exposed to a Fuji imaging plate and to X-ray film and the hybridization signals were quantified using Fujix Bas
1000 software.
The first section describes the MMA filter and the hybridization conditions for a complex probe prepared from total RNA. Controls are detailed
in the three following sections, mainly the standardization provided by a plasmid containing an
A.thaliana
cytochrome gene that does not hybridize with mouse sequences, and an evaluation
of the sensitivity and reproducibility of signal intensity measurements. The
two final sections present results on differential expression of a set of genes
between two different cell types, an epithelial cell line (MTE-1D) and a resting cytotoxic T cell clone (KB5.C20), and same T cell clone
in a resting or activated state.
In the experiments reported here we used a filter containing 47 clones corresponding to a series of known genes (Table
1
), spotted in quadruplicate to improve the precision of the measurement. The filter was hybridized successively with a vector probe (Fig.
1
A) and a complex probe (Fig.
1
B). This was prepared with total RNA from the MTE-1D cell line containing as an internal standard a small amount of
in vitro
transcribed RNA of
A.thaliana
cytochrome c554 corresponding to an approximate abundance of 0.1% (see below).
Artefactual hybridization, via the poly(A) stretch present in some clones, can
be a significant problem, as already discussed (
10
,
20
). To avoid production of poly(T) tracts the probe was prepared by simultaneous
reverse transcription and labeling in the presence of [
32
P]dCTP, using a large excess of oligo(dT) primers (dT
25
) under stringent hybridization conditions (see Materials and Methods). Three
plasmids containing only poly(A) sequences were used as negative controls, for
which, as for the vector control, the signal must be close to the filter
background. This checks for elimination of this artefact.
Table 1
Hybridization with the vector probe allowed detection of all the colonies that
had grown (Fig.
1
A). With the complex probe (Fig.
1
B) approximately half of the colonies gave hybridization signals; after
quantification the intensity values ranged over more than two orders of
magnitude. Three sets of spots are indicated on the image obtained with the
complex probe (Fig.
1
B), corresponding to three levels in the diagram. Figure
1
C shows a graphical representation of the data obtained after imaging plate
exposure and quantification. Each hybridization signal was corrected by
dividing by the signal obtained after vector hybridization and normalized using the c554 signal (see below). Results are ordered by increasing hybridization signal
on a logarithmic scale. The hybridization intensity of EF1-[alpha] corresponds to an abundance of 1%, the abundance of [alpha] tubulin is between 0.01 and 0.1% and H2TL(d) is among the
lowest intensities, corresponding to an abundance of <0.01%.
To standardize hybridization intensities obtained in several experiments we used the
A.thaliana
cytochrome c554 cDNA sequence (1 kb insert), which has no homology with
mammalian DNA. To compare independent hybridizations more precisely, the same
amount of c554 RNA,
in vitro
transcribed with T3 polymerase from the corresponding cDNA clone, was added
before labeling to the total RNA of each cell type or tissue to be tested. The quantification of
corresponding colonies present on each filter (Figs
1
A and B and
5
A and B) allowed us to normalize each independent hybridization according to
this value, which corrects for differences in the labeling, washing, duration
of exposure and progressive degradation of the filters. These variations can be
taken into account so that the differential expression levels for each clone
can be compared with greater confidence.
Previous analysis indicated that, under our hybridization conditions, the signal
intensity is proportional to both the amount of target and the concentration of the hybridizing sequences in the complex probe (
10
,
17
). To control variability resulting from the different amounts of DNA bound to
the filter due to spotting, growth of bacteria and DNA binding efficiency we
performed the following experiments.
After hybridization with a complex probe prepared from the MTE-1D cell line, each clone giving a signal on the filter was quantified. The
intensity obtained for the first colony of a quadruplicate (see above) is
plotted against that obtained with the fourth. The diagram (Fig.
2
A) shows dispersion around the diagonal resulting essentially from variations in
the amount of DNA in each spot. The same filter was hybridized with a vector
probe and the signals quantified. The intensity obtained with the complex probe
was divided by that measured with the vector probe. The resulting plot,
displayed in Figure
2
B, shows a marked reduction in the dispersion.
To compare the sensitivity of hybridization with complex probes with that of
Northern blots, two reference genes were selected after hybridization with MTE-1D RNA (Fig.
1
C): the p31 invariant chain (estimated abundance 0.0057%, close to the
threshold) and Rab 5c (estimated abundance 0.0094%). Labeled probes corresponding to these two cDNA clones were successively hybridized with a Northern blot obtained from a gel loaded with serial
dilutions of MTE-1D RNA (25-0.25 [mu]g). As shown in Figure
3
, the hybridization signals with p31 and rab5c cDNA are detected in lanes where
6.5 and 5.2 [mu]g RNA respectively were loaded. The signals obtained with these two
reference clones are close to and twice the threshold defined with the complex
probes (made from 25 [mu]g total RNA) respectively. Thus the threshold on Northern blots is reached
with approximately four times less total RNA. In conclusion, the sensitivity of
Northern hybridization seems ~4-fold better than that of complex probe hybridization.
We used this system to compare mRNA levels in two different cell types, a
cytotoxic T cell clone, KB5.C20 (unstimulated), and an epithelial cell line,
MTE-1D. Among the 47 clones spotted on the filter, 39 gave a signal with
complex probes made with either MTE-1D or KB5.C20 RNA. As shown in Figure
4
, 12 genes expressed in the cytotoxic T cell clone are not detected in this
epithelial cell line, whereas only three genes are exclusively detected in the
epithelial cell. Expression of the other genes is comparable in both cell
types, with the exception of the 5-fold higher representation of thioredoxin mRNA in KB5.C20 cells. This
expression profile was expected, since candidate genes were chosen to analyze their variations in T cells. Indeed, interferon [gamma], components of the T cell receptor complex, such as CD3[epsilon] and CD3[delta], associated molecules, such as CD8[alpha], ZAP-70 and p59
fyn
, and activation markers, such as the IL-2R[alpha] receptor, CTLA-1 and CTLA-3, are expected to be transcribed in a T cell and not
in an epithelial cell. Some HMG2 (high mobility group 2) family genes, members
of the HMG transcription factor group, are known to be highly expressed in
lymphoid cells (
21
,
22
). A new member of the small G protein family, mu-Rho (whose sequence is homologous to canine and human rho), was identified
with consistent expression in this T cell line as EN-7 small G protein, in agreement with the literature (
23
). On the other hand, among genes detected only by hybridization with the MTE-1D probe we found cathepsin L, which has been described as expressed in
epithelial cells (
24
), and mu-CD63, described as expressed in kidney and in macrophages after activation (
25
). Finally, the preferential expression in epithelium of (nuclear encoded) NADP
isocitrate dehydrogenase is probably related to the high mitochondria content
of epithelial cells in comparison with T cells (
26
).
To analyze transcriptional events occurring upon activation of KB5.C20 cells
total RNA was prepared from T cells either in the resting stage or activated by
an anti-CD3 antibody.
The same MMA filter was then hybridized with each complex probe, containing the same amount of RNA transcribed from
A.thaliana
cytochrome c554 (Fig.
5
A and B; see above). Our positive control (CD8[alpha]) is expressed in the quiescent T cell clone (Fig.
5
A), while a negative control, CD4, was not found expressed in either
unstimulated or stimulated T cells (Fig.
5
A and B;
27
).
A graphical representation of the results obtained with the RNA from resting T cells is shown in Figure
5
C (lower graph). All detected cDNAs were ordered according to the relative
abundance of RNA in resting T cells. The ratio between hybridization signals from the same
cDNA obtained using probes from stimulated or unstimulated T cells is plotted
(stimulated/unstimulated) in Figure
5
C (upper graph). Interferon [gamma] (IFN[gamma]) shows a stimulation ratio of ~20. This result is in agreement with previous reports showing
IFN[gamma] induction both at the mRNA and at the protein levels after CD3-mediated triggering of the KB5.C20 T cell clone (
28
). Expression of a number of genes increased with a stimulation ratio of 2 for
IL-2R[alpha] and CD8[alpha], while H2TL(d) and CTLA-1 increased with stimulation ratios of 3 and 5
respectively (Fig.
5
C, upper graph). The fact that other cytokine encoding genes, such as IL-2 and IL-4, were not induced in response to anti-CD3-mediated activation is characteristic of this type of
CD8
+
cytotoxic T lymphocyte (CTL) clone (
29
).
Reliable means of quickly assessing expression profiles for sets of cDNA clones
representing hundreds or thousands of genes are needed to provide this
essential complementary information that represents a first step toward
functional analysis. Most primary gene activation in eucaryotes requires only ~15-20 min from the initial stimulus to the appearance of mRNA,
indicating that if one regulatory gene were to simply activate the next
regulatory gene plus a group of functional genes, there would be hundreds if
not thousands of such steps in a process as complex as activation or
differentiation, lasting many days (
31
)
Current methods used to determine multiple expression profiles use widely
different approaches. Systematic sequencing of a set of randomly chosen clones
from carefully constructed cDNA libraries, as implemented by Okubo and co-workers (
3
), provides frequency data that translates into expression information, but its
sensitivity is limited unless very large numbers of clones are analyzed. Rapid
PCR-based methods use variations on the original differential display
technique (
32
) to target it to a set of pre-determined sequences (
33
). While very sensitive, this approach is not quantitative and does not lend itself readily to simultaneous assay of many
diverse sequences. The SAGE method (
34
), in contrast, does provide quantitative information while minimizing the
amount of sequencing work through the ingenious use of short, concatenated
sequence tags, however, detection and quantification of transcripts present at
low levels still requires the analysis of very large numbers of tags.
Hybridization signature methods are inherently parallel and can provide
simultaneous expression information on many genes; their sensitivity can be
enhanced by modern detection methods and, possibly, by the use of linearly
amplified probes. Schena
et al
. (
11
) have demonstrated such a system using PCR products of
A.thaliana
cDNAs printed in microarrays on glass microscope slides; fluorescence-based detection allows simultaneous two color hybridization, which
minimizes the experimental variations inherent in the comparison of independent
hybridizations. This sophisticated technology requires specially developed, state of the art instrumentation for both spotting of the DNA targets and detection of the
hybridization signals.
Our MMA system, based on commercially available equipment, can be implemented in
any laboratory and is quite generally applicable. We had previously established
hybridization signature measurement on high density filters (
10
), shown quantitative correlation between the signals and the amounts of target
and probe DNA and eliminated major artefacts. In the present implementation,
designed for more precise expression measurement on a smaller number of clones,
the precision has been increased by spotting clones in quadruplicate and the
sensitivity reaches 1 in 20 000 using 25 [mu]g total RNA (corresponding to ~0.5 [mu]g mRNA), to be compared with 1 in 50 000 with 2 [mu]g mRNA in the fluorescent system. An external standard allows
normalization of signals from independent hybridizations and makes possible
precise comparison of different experiments. The preparation of the complex
probe from total RNA makes it possible to use relatively low numbers of cells
(5 * 10
6
) or small amounts of tissues. Expression levels for a set of 100 genes are
obtained in one step using 10-20 times less material than needed for the Northern blot hybridization
technique, which would also require much longer. We observed that Northern blot
hybridization seems to be four times more sensitive than complex probe
hybridization, however, the same reliability in detection of variations is
conserved, even for weak inductions. The expression data obtained by this
method, just as with a Northern blot, provides a global view of the amount of
mRNA at one precise moment, resulting from the balance between transcription
and degradation. Of course, gene families give rise to difficulties in this
system, just as in Northern hybridizations.
On the set of clones used in this study, a difference in the pattern of
expression was easily observed between two cell types (MTE-1D and KB5.C20). We observed consistent expression of a new mouse Rho-like gene in T cells. This gene did not appear to be transcribed at
a detectable level in MTE-1D cells, while the other small G proteins Rab5c and Ran were detected (
35
,
36
; for a review see
37
,
38
). We easily detected the variation of expression between two different states
of a given cell (resting versus activated). Here we found that anti-CD3-mediated stimulation, which is efficient at inducing both perforin- and Fas-based cytotoxic activities in clone KB5.C20 (
28
), led to an increase in CTLA-1 gene expression without affecting CTLA-3 gene expression. It should be noted that the role in cytotoxicity
of the latter granule component is not clear, since cytotoxic function does not
seem to be affected in mice rendered deficient in CTLA-3 expression (
39
). In contrast, CTLA-1 appears to be necessary to the lethal hit delivered by CTL (
40
).
The MMA method is adapted to analysis of the transcriptional level of a
relatively large number of genes in a kinetic context. In the near future the
majority of genes will be partially or completely sequenced and this method
will be useful for cell typing by expression profile in a number of normal or
modified contexts, i.e. during development, as well as in neoplastic or drug-treated cells
We wish to thank Malek Djabali and Philippe Naquet (Centre d'Immunologie de Marseille-Luminy) for many helpful discussions and Charles Auffray (Genexpress, Généthon) for tag sequencing. This research was supported by institutional grants to
CIML from Centre National de la Recherche Scientifique (CNRS) and Institut
National de la Santé et de la Recherche Mdicale (INSERM), as well as by specific grants from
Groupement de Recherches et d'Etudes sur les Génomes (GREG) and from Association Française contre les Myopathies (AFM). SG was supported by an AFM
post-doctoral fellowship.
REFERENCES
Return