Though colon cancer is the second leading cause of cancer deaths in the US, it is entirely preventable through early screening to detect and remove adenomatous polyps. Colonoscopy has long been regarded as the “gold standard” but is expensive, invasive, and uncomfortable, and only about half those considered at risk for colon cancer currently submit to colonoscopy or to less reliable alternatives such as fecal occult blood test. Here we describe the use of gene expression analysis to detect altered expression of certain genes associated with not only colon cancer but also polyps. The analysis can be performed on rectal swabs, with specimens provided in a routine doctor's office visit. The existence of this cheap and simple test, together with an active program to encourage individuals to submit to screening, could help eradicate colon cancer.
Colorectal cancer is the third most common cancer in the US and continues to be the second leading cause of cancer deaths. Each year, about 150,000 new cases will be diagnosed and 50,000 patients will die of this disease. These data have not changed much despite increased efforts over the last decade to persuade individuals at risk to undergo screening. In the US, colonoscopy remains the so-called gold standard, because it can detect not only early colorectal cancer but also significant precancerous polyps such as the serrated adenomas or the villotubular adenomas, which can be removed by the same procedure. In principle, then, colon cancer is entirely preventable.
Colonoscopy is expensive and invasive, however, and requires a colon preparation prior to surgery. Individuals without health insurance, and of lower socioeconomic status, are less likely to have a colonoscopy. Some of these individuals opt for other recommended screening procedures, including combination of fecal occult test for blood in the stool annually or flexible sigmoidoscopy every 5 years. These tests are also more common in many countries outside the US, due to financial limitations. Lately, with the US in recession and major cutbacks proposed on health care expenditures, even expert gastroenterologists have suggested that perhaps colonoscopy should not be the gold standard for colon cancer screening [
In February of 2010, the National Institutes of Health (NIH) organized a Consensus Conference, bringing together with a public representative a group of experts representing the fields of cancer surveillance, health services research, community-based research, informed decision-making, access to care, healthcare policy, health communication, health economics, health disparities, epidemiology, statistics, thoracic radiology, internal medicine, gastroenterology, public health, end-of-life care [ eliminate financial barriers to colonoscopy; promote interventions that have been shown to be effective in persuading people to be screened; develop more effective educational programs for targeted patient groups, such as those who lack health insurance and/or are of lower socioeconomic status; implement approaches that will ensure appropriate followup of positive screening results; develop systems to ensure the high quality of colorectal cancer screening programs; conduct studies to determine the comparative effectiveness of the various colorectal cancer screening methods currently in use.
The development of a cheaper and less invasive screening procedure that could detect both cancer and precancerous polyps as reliably as colonoscopy would address most of these issues. In particular, a cheap, simple, noninvasive test that could be performed by any physician would remove the financial and emotional barriers to screening. Several possible alternatives are currently being researched, including and stool tests for DNA mutations associated with colon cancer [
In this paper, we will discuss our research to develop a rectal swab test, without the need for a bowel preparation. Those individuals testing positive will be recommended for immediate diagnostic colonoscopy to remove any large precancerous polyps or resect an early cancer. Individuals repeatedly testing negative in the swab test should not require colonoscopy. Our aim is that the high sensitivity and specificity of this test will compare favorably with all existing screening modalities.
Tissues examined in our studies included the entire colon of mouse (C57BL/6J-min/+ mice, and the wildtype littermates were obtained from Jackson Laboratory, Bar Harbor, ME, USA), which was removed from the animals, opened longitudinally, and washed in cold phosphate-buffered saline, polyps and morphologically normal colon tissue removed from patients undergoing colonoscopy at the California Pacific Medical Center (CPMC), and colon cancer removed from patients undergoing surgical resection at CPMC. The appropriate procedure for obtaining formed consent was followed for all individuals participating in these studies. All samples from human patients were snap-frozen on dry ice as soon as possible within 30 minutes of surgery, then taken immediately to the laboratory for RNA preparation (see below).
Total RNA was extracted from tissues using RNAeasy kits from Qiagen (Valencia, California). RNA samples were treated with RNase-free DNase to remove any genomic DNA contamination and were reverse-transcribed. Fifty ng of cDNA from each sample were used as template for PCR amplification with specific oligonucleotide primers using the Applied Biosystems 5700 Sequence Detection System (PE Applied Biosystems, Foster City, California). PCR reactions were performed according to the manufacturer’s instructions using the SYBR Green PCR Core Kit (PE Applied Biosystems, Foster City, California).
We analyzed fifteen genes, all of which have previously been shown to be altered in expression in human colon cancer. They fall into four groups, including those involved in the (1) APC/
Specific primers against each gene were designed using the Primer Express Software (PE Applied Biosystems, Foster City, California). Primer length was 21–27 nucleotides, with a theoretical
All the cDNA samples were tested for genomic DNA contamination by using primers for
For quantitation of gene expression, the fluorescence of the SYBR Green dye bound to the PCR products was measured after each cycle and the cycle numbers were recorded when the accumulated signals crossed an arbitrary threshold (
We used the Wilks lambda criterion for a multivariate analysis of variance (MANOVA) to compare the patterns of expression levels of several genes from cancer versus normal subjects. This test takes into account correlations among gene expression levels and controls the false positive rate by testing the global hypothesis of no differences in gene expressions between cancer and normal subjects. If the test was significant; that is, there was evidence that expression patterns differ, then we used univariate
In some studies, we also determined the Mahalanobis distance (M-dist). This measure summarizes, in a single number, the differences in a pattern of gene expression, for any individual against the average of a pool of individuals, taking into account variability of each gene’s expression and correlations among pairs of genes. It is thus well suited both for comparing a control population with an experimental group, such as individuals with cancer, as well as determining the degree of similarity or fit that an individual of unknown characteristics has to a well-characterized group. The latter is what allows M-dist to be used to determine how significantly an individual value differs from a group of controls, which is necessary in screening.
To perform the calculations, first, for each control biopsy (total of 105), we calculated its M-dist from the multivariate mean of the other 104 control biopsies. We plotted ordered M-dist for the 105 control biopsies against the theoretic expected order statistics for the appropriate chi-squared distribution, to verify that control gene expression values (log base 2) were multivariate normal. Then we computed an M-dist for the gene expression data for each biopsy from each individual with polyps, where M-dist measured the individual’s multivariate distance (i.e., difference in pattern of expression) from the pooled mean of the 105 control biopsies.
M-dist can be converted to
Gene expression changes in colon mucosa of a mouse model of colon cancer. We began our studies by examining the APCmin mouse [
When we analyzed polyps that were removed from these animals at various ages, we observed a wide range of expression levels of these genes, ranging from several that were dramatically upregulated to several that were modestly upregulated, others that exhibited no significant change in expression level, and several that were downregulated. As shown in Table
Relative gene expression levels in colon polyps of APCmin mice (mean ± SE).
No. | Gene | Wildtype littermate | Individual polyp | |
---|---|---|---|---|
1 | OPN | <0.01 | ||
2 | MIP-2 | <0.001 | ||
3 | Gro- | <0.001 | ||
4 | CXCR2 | <0.001 | ||
5 | COX-2 | <0.001 | ||
6 | Cyclin D1 | <0.001 | ||
7 | SDF-1 | <0.01 | ||
8 | c-myc | <0.001 | ||
9 | M-CSF1 | NS | ||
10 | CD44V6 | <0.01 | ||
11 | COX-1 | <0.01 | ||
12 | PPAR- | NS | ||
13 | p21cip/waf1 | <0.05 | ||
14 | PPAR- | <0.05 | ||
15 | PPAR- | <0.001 |
Gene expression levels were determined using RT-PCR. In no. 1~5,
In studies like this that have been carried out previously by other investigators, it has been assumed that gene expression values in normal appearing mucosa in the mutant mice, in regions away from the polyp, would be similar to those in control mice without polyps. However, when we actually compared the two, we found it was not the case. In these experiments, polyps were removed from the intestines of APCmin mice at three different ages—6, 13, and 23 weeks old—and the polyp-free intestines compared with normal colon tissue from wildtype littermates. The intestines were divided into six equal segments of approximately 1.5 cm in length, colonic mucosa was isolated, and the expression of the five genes most altered in polyps analyzed.
While the expression levels of a particular gene in a particular segment at a particular age showed little variability from one wildtype animal to another, there was considerable variation in values for APCmin mice. As shown in Table
Multivariate analysis of gene expression in normal-appearing colon mucosa of 23 week old APCmin mice, as compared to colon mucosa of normal mice.
Colon segment | ||||||
Gene | 1 | 2 | 3 | 4 | 5 | 6 |
COX-2 | ++ | ++++ | + | ++++ | + | +++ |
CXCR-2 | + | ++ | — | ++ | + | ++++ |
MIP-2 | ++ | ++ | — | ++ | + | +++ |
Gro- | — | + | — | — | + | +++ |
OPN | — | — | — | — | — | — |
Colons were removed from animals and any polyps removed. The colons were then divided into 6 segments, colon mucosa isolated, and gene expression determined as described in the Material and Methods section, with values for APCmin mice compared to those for wildtype mice. Multivariate analysis was performed on these values as described in the Material and Methods section, in which the significance of the difference in expression between APCmin and wildtype mice was determined for each gene in the presence of all the other genes. For this analysis, 6 mice were used for each group (APCmin and wildtype), and 1 mucosa sample analyzed per segment per mouse, +:
Distribution of IL-8 expression in sigmoidal-rectal colon of a single cancer patient. The patient had a cancer in the sigmoidal-rectal colon, as indicated by the spot with jagged edges. Locations of other mucosa samples removed for analysis are indicated by the circular spots. Levels of IL-8 expression in each sample were determined by RT-PCR and are indicated roughly by color coding, as shown in the Figure. The mean level of expression of IL-8 in colon mucosa of noncancer patients was 2.25 (as shown in Table
We next analyzed colon samples from human patients who had previously undergone surgery to remove colon carcinomas [
As with the mice, we observed great variability of expression levels in morphologically normal mucosa from cancer patients (Table
Multivariate analysis of gene expression in normal appearing colon mucosa of colon of individuals with cancer and controls.
Sigmoidal-rectal colon
Normal subjects | Cancer patients | |||
Gene | Mean | Range | ||
1 | CXCR2 | 1.30 ± 1.11 | 0.81–210.11 | <0.01 |
2 | Gro- | 2.93 ± 6.93 | 0.78–104.69 | NS |
3 | IL-8 | 2.25 ± 2.63 | 1.22–82.14 | 0.0001 |
4 | COX-2 | 1.80 ± 2.63 | 0.91–66.26 | 0.001 |
5 | OPN | 1.55 ± 2.04 | 0.94–58.08 | 0.0001 |
6 | Gro- | 1.92 ± 3.34 | 0.80–36.50 | NS |
7 | M-CSF-1 | 1.54 ± 1.40 | 1.54–30.70 | 0.0001 |
8 | COX-1 | 1.22 ± 0.87 | 0.12–9.58 | NS |
9 | CD44 | 1.12 ± 0.56 | 0.54–6.52 | <0.05 |
10 | c-MYC | 1.24 ± 0.82 | 0.12–4.76 | NS |
11 | Cyclin D | 1.28 ± 0.84 | 0.43–4.44 | NS |
12 | PPAR- | 1.10 ± 0.62 | 0.02–2.87 | NS |
13 | PPAR- | 1.15 + 0.55 | 0.023–1.90 | <0.01 |
14 | P21 | 1.04 + 0.29 | 0.40–1.68 | <0.01 |
15 | PPAR- | 1.07 + 0.40 | 0.01–1.28 | <0.01 |
Ascending colon
Normal subjects | Cancer patients | |||
Gene | Mean + SD | Range | ||
1 | CXCR2 | 1.32 + 1.08 | 1.90–90.20 | <0.05 |
2 | Gro- | 1.60 + 2.08 | 0.46–29.90 | NS |
3 | IL-8 | 1.66 + 1.62 | 1.32–182.66 | <0.05 |
4 | COX-2 | 1.84 + 3.04 | 2.96–152.50 | 0.0001 |
5 | OPN | 1.53 + 1.31 | 9.24–152.98 | 0.0001 |
6 | Gro- | 1.40 + 1.41 | 0.63–11.16 | NS |
7 | M-CSF-1 | 1.68 + 1.62 | 4.01–40. 19 | 0.0001 |
8 | COX-1 | 1.17 + 0.75 | 0.84–44.90 | <0.001 |
9 | CD44 | 1.11 + 0.51 | 0.99–13.63 | 0.0001 |
10 | c-MYC | 1.16 + 0.63 | 0.39–10.82 | NS |
11 | Cyclin D | 1.38 + 1.08 | 0.12–13.15 | NS |
12 | PPAR- | 1.16 + 0.58 | 0.22–4.09 | NS |
13 | PPAR- | 1.13 + 0.55 | 0.02–7.08 | <0.05 |
14 | p21 | 1.09 + 0.40 | 0.04–2.66 | NS |
15 | PPAR-g | 1.08 + 0.42 | 0.01–1.14 | 0.01 |
Colon mucosa samples were isolated from (a) the sigmoidal-rectal region of noncancer subjects (78 samples from 12 individuals) and from the adjacent normal mucosa of patients with sigmoidal-rectal cancer (62 samples from 5 patients); or (b) from the ascending region of noncancer subjects (39 samples from 11 individuals) and from the adjacent normal mucosa of patients with ascending colon cancer (65 samples from 4 patients). Samples were analyzed for gene expression as described in the Material and Methods section. Means + standard deviations are given for noncancer subjects; ranges are given for cancer patients. Multivariate analysis was then performed on each gene taken in relation to all the other genes, to determine the significance of the difference between cancer and noncancer individuals. NS, not significant at
All together, seven genes appeared to be significantly upregulated in morphologically normal mucosa of sigmoidal-rectal cancer patients, relative to mucosa of noncancer patients: M-CSF-1, OPN, IL-8, COX-2, CXCR2, p21, and CD44. An additional two genes—PPAR
The samples of normal-appearing mucosa from cancer patients that were analyzed for the data in Table
These observations strongly suggest that the differently regulated areas of gene expression in normal-appearing colon mucosa of cancer patients did not result from a field effect of spreading cells from the original cancer. It appears that, in individuals with cancer, the normal-appearing colon mucosa has developed abnormalities that can be detected at the molecular level. Polley et al. [
To summarize these studies, morphologically normal colon mucosa in APCmin mice and in human cancer patients is not metabolically normal. Altered gene expression in this tissue does not appear to result from a field effect, because there was no correlation between extent of altered regulation and distance from polyp or tumor. Our data suggest that alterations of expression levels of certain genes may be an early event in carcinogenesis and may serve as a marker of risk to development of colon cancer.
We next examined whether these alterations in gene expression patterns could also be observed in morphologically normal colon mucosa of individuals with adenomatous polyps [
To distinguish any effects of personal/family history alone from the presence of polyps, we initially carried out three group-wise comparisons: (1) individuals with polyps and no personal/family history versus controls, (2) individuals with polyps and personal/family history versus controls, and (3) individuals with polyps and personal/family history versus individuals with polyps and no history. The first two comparisons, individuals with polyps and with or without history versus controls, were significant, whereas there was no significant difference in gene expression levels between individuals with history and without history.
Further analysis was carried out on individual biopsies, using the Mahalanobis measure. We compared the M-dist for controls and for individuals with polyps, plotted on a logarithmic scale. A log-rank test comparing the distribution of all biopsies from individuals with polyps versus all controls indicated a highly significant difference (
To summarize, this study found that normal-appearing colon in individuals with polyps, like that we had previously demonstrated in individuals with colon cancer, exhibited altered levels of gene expression. Thus these changes occur relatively early in the carcinogenetic process, before the appearance of an actual cancer.
Since our previous studies had indicated that the presence of either adenomatous polyps or colon cancer in humans is associated with significant alterations in the expression of certain genes in the normal-appearing portion of the colon, we next examined whether such changes exist even in individuals with no polyps but possibly at risk for cancer by virtue of a family history of the disease [
Twelve individuals with a family history of colon cancer in a first-degree relative and sixteen individuals with no known family history of colon cancer were included in the study. Biopsy samples of normal-appearing colon mucosa were obtained from the ascending, transverse, descending, and rectosigmoid regions of the colon (2–8 biopsy samples were obtained from each region). Relative to normal controls, the expression of several genes, including PPAR-
Multivariate analysis of the expression values of all sixteen genes indicated a significant difference in the biopsy samples from the rectosigmoid region (
The studies discussed above demonstrate that morphologically normal colon mucosa from individuals with colon cancer or at increased risk for colon cancer have altered gene expression patterns, which could be the basis for screening. However, all of the studies we have discussed to date involved removal of biopsies during colonoscopy. Since the point of developing a molecular screening process is to avoid the necessity of colonoscopy, we next sought to develop a more noninvasive way of obtaining colon mucosa samples. Using an anoscope, we inserted a soft brush about 2 cm. into the colons of individuals and gently swabbed to remove colon mucosal cells.
These cells were removed from the brush by dipping and swirling it in a buffer. This was followed by extraction of RNA, preparation of cDNA, and PCR. In this manner, we compared rectal swabs with biopsies from 90 patients, who included individuals with no polyps, but with family or personal history of cancer, individuals with adenomatous polyps (with or without history), control individuals with neither history nor polyps or colon cancers, and cancer patients.
Analysis of individuals with cancer, polyps, or family/self-history of cancer clearly showed that gene expression profiles of swab samples were very similar to profiles of samples obtained by biopsies (Table
Significance of three groups versus controls for gene expression levels.
Comparison of overall group | ||
Swabs | Biopsies | |
Cancers versus controls | <0.001 | NA |
Polyps versus controls | <0.01 | <0.01 |
FHSH versus controls | <0.01 | <0.01 |
Summary of patients with altered gene expression in three groups versus control group.
Number of patients with altered gene expression | ||
Swab samples | Biopsy samples | |
Cancer ( | 5/5 (100%) | NA |
Polyps ( | 21/25 (84%) | 20/25 (80%) |
FHSH ( | 26/37 (70%) | 25/37 (68%) |
Biopsy data showed comparable numbers (68–80%). These data indicate that the sensitivity of our gene expression analysis to detect individuals of cancer risk is quite high if multiple rectal swabs are analyzed. Some patients in our polyps’ group also had family history or self-history but no subject in the family/self-history (FHSH) group had polyps. This may have resulted in a higher percentage of significantly different individuals in the polyps group than in the FHSH group.
The cancer group was very small, consisting of just five individuals. But gene expression analysis indicated that not only did the cancer group differ significantly from the control group but each of five individuals was highly significantly different from controls as well. The M-dist values of 48 out of 50 (96%) of our total swab samples from these five individuals were above the 95th percentile line. This suggests a high sensitivity of this assay to identify individuals with colon cancer, higher than has generally been reported using stool analysis of gene mutations [
Furthermore, in one case, swabs were taken from an individual with cancer both before as well as after bowel preparation. The altered gene expression profile was highly significant in both instances. While further studies will be required to support this conclusion, this result suggests that the rectal swab procedure may be able to dispense with bowel preparation. This is another significant disadvantage associated with colonoscopy that undoubtedly contributes to poor patient compliance, so eliminating it should further increase the attractiveness of the swab procedure.
Our studies suggest that gene expression analysis may be suitable as a screening process to identify individuals at risk for developing colon cancer. While promising advances have been made in the use of both DNA stool tests [
As applied to colon cancer, as we envision it, gene expression would not replace colonoscopy but allow its limited resources to be focused on those individuals whom expression analysis indicates are most likely to have polyps. Given the large and growing variety of tests being explored, it may be that a combination of more than one type of test will prove to have the highest sensitivity to detection of cancer and polyps. If individuals who are free of polyps and cancer can be reliably identified without colonoscopy, it would result in an enormous reduction of needed resources, for both individuals and society.