Colorectal cancer (CRC) is the third most common cancer (10–15% of all cancers) and one of the leading causes of cancer-related deaths worldwide, with an estimated half a million deaths worldwide and over fifty thousand deaths in the United States alone.

CRC is a heterogeneous complex of diseases caused by destructive genetic/epigenetic alterations that accumulate in a sequential manner through a multistep carcinogenic process [1]. It is therefore likely that tumors with similar characteristics might originate in the same manner and have a similar molecular behavior. Since the molecular features of a given tumor reflect the mechanism(s) underlying disease development and progression, the implication for tumor classification is significant. For instance, molecular classification of leukemia and lymphomas has tremendously enhanced our understanding of these diseases [2], [3], [4]. The investigation of the molecular bases of two major syndromes, familial adenomatous polyposis (FAP) and hereditary nonpolypsis CRC (HNPCC), has led to the identification of two main pathways by which these molecular events can lead to CRC [5]. About 85% of CRCs arise from events that result in chromosomal instability (CIN), with aneuploidy and early inactivation of adenomatosis polyposis coli (APC). A further 15% result from processes that generate microsatellite instability (MSI), replication error or loss in the caretaker mismatch repair (MMR) function associated with HNPCC [6], [7], [8], [9]. Although we have improved our understanding of the molecular events underlying the development of CRC, no significant impact on patient care has resulted. Even though considerable progress that been made in the treatment of patients with CRC using folic acid (FA)-modulated 5-flurouracil (5-FU), about 50% of CRC patients eventually develop metastatic CRC (mCRC). However, the use of new chemotherapy agents, such as oxaliplatin and irinotecan, either alone or in combination with approved biological agents, such bevacizumab and cetuximab, promises to prolong survival [10], [11].

Therefore, in order to maximize the available treatments, it is critically important to gain even more insight into the molecular mechanisms underlying disease development and progression, as well as significantly improve our efforts to elucidate new therapeutically relevant targets and molecular markers. Such efforts will help expand and diversify disease management options. Studies have also shown that shifting disease detection to an earlier stage through mass screening and intervention at this stage can reduce the risk of death from CRC [12], [13]. These findings strongly demonstrate the clinical need for biomarkers for the early detection of CRC so that the disease can be effectively managed.

Genomic techniques, such as DNA microarray analysis, and proteomic methods, such as two-dimensional (2-D) electrophoresis and mass spectrometry, are now commonly used to elucidate the expression profiles of genes and proteins in cells, tissues and bodily fluids [14], [15]. Indeed, the identification of genes and proteins that are characteristically produced during the development of cancer can potentially uncover useful biomarkers that will aid in the management of CRC. Although proteomics have played a dominant role in the field of biomarker development [16] and will continue to do so, the current proteomic strategies have not generated enough markers for CRC.

Interestingly, CRC is one of the first cancers for which tumor markers were used to aid in disease management. For example, carcinoembryonic antigen (CEA) has been used extensively to determine prognosis and monitor both disease progression and therapy after curative resection. A high level of CEA in the serum is associated with cancer progression. However, even in the absence of cancer, high levels of CEA have been reported in conditions such as hepatitis, pancreatitis, inflammatory bowel disease and obstructive pulmonary disease. In addition, other cancers, such as pancreatic, gastric, lung and breast, have elevated levels of CEA, indicating the lack of specificity of this marker. Other markers, such as carbohydrate antigen 19-9 (CA19.9), CA242, metalloproteinases-1 (TIMP-1), Thymidylate synthase, p53, and APC gene, all lack the necessary sensitivity and specificity. To be clinically useful, a biomarker must be effective and have high predictive value [7], yet no such single biomarker exists. Although no consensus has been reached, it seems that the development of multiple biomarkers for detection and risk assessment of a single cancer is favored over single comprehensive biomarkers that can successively predict risk for any type of cancer [7]. This is even more important as multiplexing methods are becoming more of a norm than the exception.

Biomarkers that are directly associated with tumor cells as they are transformed from normalcy into malignancy will be of significant interest as these markers can be useful tools for mapping the molecular features of the diseased cells. The ability of probes to identify an important clinical specimen, such as exfoliated malignant colonocytes, rather than normal colonocytes, would be particularly important for CRC [16]. Specifically, once exfoliated cells from colonic mucosa have been sloughed into the stool, it has already been demonstrated that DNA from the stool can be isolated and subjected to a multi-target DNA analysis. This assay can currently assess 15 mutational hot spots, including k-ras, p53, APC, BAT-26 and L-DNA. Although DNA fecal markers are quite promising, they are not widely used in clinical settings and therefore probes that can specifically detect the cells will be important.

In the development of sensitive, selective molecular markers for CRC, cell-based aptamer selection holds significant promise by its potential to identify multiple useful markers in a relatively short time. During the last two decades, appreciable efforts have been made to develop markers for various types of cancers using a process known as the Sytematic Evolution of Ligands by Exponential enrichment (SELEX) [17], [18]. Through this process, numerous oligonucleotide probes (aptamers) have been generated that can bind specifically to proteins associated with membranes of different tumor cells [19], [20], [21], [22], [23], [24], [25], [26]. Aptamers are short, single-stranded oligonucleotides (DNA or RNA), typically <100 mer, that have the ability to bind to other molecules with high affinity and specificity. They have been generated against a variety of targets from small molecules [27], [28], [29], [30] and peptides/proteins [31], [32], [33], [34], [35], [36] to whole cells [19], [20], [21], [22], [23], [24], [25], [26] as well as bacteria, viruses and virus associated proteins [37], [38], [39], [40], [41], [42], [43]. The cell-based selection strategy generates aptamers that bind to unknown targets in their native state. Nevertheless, the target can still be identified through affinity extraction and mass spectrometry [44], [45]. To create more sensitive and selective probes for CRC, we have developed a panel of DNA aptamers that specifically recognize colorectal cancer cells. These probes were generated by cell-SELEX using colorectal cancer cultured cell lines DLD-1 and HCT 116. Initial binding studies by flow cytometry and confocal microscopy using these cultured cell lines show that most of our aptamers have high affinity and selectivity. Our data further show that these aptamers neither recognize normal colon cells (cultured), nor do they recognize a majority of the other cancer cell lines tested. Our findings clearly show that the probes identify specific membrane proteins associated with colorectal cancers. We believe that these probes could be further developed for early disease detection, as well as prognostic markers, of colorectal cancers.

Over the last decade, several aptamers have been developed for different targets, including purified molecules, as well as complex targets, such as the live cells of different cancers. We have used the cell-based SELEX strategy to generate a panel of DNA aptamers for colorectal cancers. A random ssDNA pool (approximately 10¹⁴) was subjected to sequential binding and elution to select from the pool DNA sequences having the ability to bind to surface markers of the target cell. DLD-1 and HCT 116 were used as targets with HCT1116 and HT-29 as respective controls in separate selections. The introduction of counter selection provided the opportunity to eliminate, to the extent possible, common surface markers, while at the same time enriching differential markers on the target cells. The DNA pool collected after each round of selection was amplified by PCR, and the product was used to prepare ssDNA for the next round of selection. In this selection strategy, the incubation of the DNA pool with the cells was performed in culture dishes (cell monolayers). The enrichment of the selection pool through successive selection was monitored by flow cytometry. In order to use the cells for flow cytometry, cells were cultured overnight and dissociated using short time (30 sec–1 min) trypsin treatment and/or non-enzymatic cell dissociation solution (MP Biomedicals). Short time treatment with trypsin did not have any observable effect with respect to the ability of the selection pool to recognize the cells, since no difference between the signal intensities of the trypsin and the non-enzymatic treatment options was observed (Text S1, figure S1). Peak shift (increase in the fluorescence intensity as compared to the library) is an indication of fluorescence intensity of the cell as a result of the labeled DNA sequence binding to the cells (Figure 1). With the increasing number of selection cycles, there was a steady increase in the fluorescence intensity of the target cells, indicating that DNA sequences with better binding to the target cells were being enriched. However, there was no significant peak shift with the control cells, especially in the early and mid rounds of selection. By the 14^th round of selection, there was a significant increase in fluorescence signal of the target as compared to the control cells. An additional two rounds (16^th round) led to an increase in the signal enhancement of the control, but no significant increase for the target. This is possible at the terminal end when selection is continued further even when there is no significant signal difference between the successive pools to the target. The highly enriched pools were cloned and the positive clones sequenced. Sequence analysis provided potential DNA aptamer candidates grouped into families based on their sequence homology. About 8 distinct homologous families and many randomized sequences were identified for the DLD-1 selection. The number of sequences in the distinct homologous families ranged from 6–110 (Table 1) with few base mutations within a family. Representative sequences from the different families were chosen to test their interactions with the target.

The rolling cycle amplification (RCA) products from sequencing were used for the initial screening. The products from the chosen wells were diluted with 100 µl H₂0 and used for PCR amplification. The PCR products were used to prepare ssDNA and then used for binding assays. Two sequences having at least one base mutation were chosen from families with sequences greater than 20 (Table 1). The use of the RCA product provided us with a faster method of screening for potential aptamer candidates. The sequences with binding signal greater than that of the control (Text S1, figure S2) were chemically synthesized and labeled with fluorescein isothiocynate (FITC) or biotin at the 3′ end. All synthesized sequences were purified by HPLC and then quantified. The binding assays were performed using flow cytometry, as described, and the binding signal was directly detected in FL1 for FITC probes, FL2 for streptavidin-conjugated PE or FL3 for streptavidin-conjugated PE-Cy5.5. The initial binding assays with the synthesized sequences named KDED1 to KDED20 revealed many sequences binding to the cell target (Figure 2). Some of the sequences belonging to the same family, and therefore envisaged to be binding to the same target, showed different binding signal strengths to the target. These include KDED2/KDED15; KDED1/KDED5; KDED9/KDED10, KDED3/KDED19 and KDED7/KDED18.

The co-current selections, using DLD-1 (without negative selection) and HCT 116 as targets, also generated the following aptamers: KC2D3, KC2D4, and KC2D8 for DLD-1: KCHA10 and KCHB10 for HCT 116 (Figure 3).

All the developed aptamers were further tested with all the colorectal cancer cell lines used in this study. The following aptamers showed recognition only to the cell line used for the selection: KDED2/KDED15, KDED7/KDED18, KDED9/KDED10 and KC2D3 (DLD-1) and KCHB10 (HCT 116). Figure 4 shows the pictorial representation of the interaction between the individual aptamers and the three different colorectal cancer cell lines. The height of the cone represents the percentage of cells that had fluorescence intensity above the control library with the threshold set at 5% fluorescence signal intensity.

We then determined the apparent dissociation constants (Kd) for the selected aptamers, which ranged between 0.68 nM and 302 nM (Table 2 and Figure 5). In order to enhance synthesis efficiency and also increase the flexibility of chemical manipulation, some of the aptamers were truncated, and the binding strength and apparent Kds of the truncated sequences were also assessed. Before each truncation, the possible structures of each aptamer sequence were predicted using Integrated DNA Technologies oligoanalyzer under the selection conditions. The most favorable hairpin structures were selected and the hang over bases at the 3′ and/or the 5′ end removed, each at a time. Series of truncations were made and each of these was finally tested with the positive cells to assess the binding. The truncation of KCHA10 (KCHA10a) did not change the properties of the aptamer. However, significant improvement of binding affinity was observed with KDED7 truncation (KDED7a) and the various forms of KDED2 (KDED2a, KDED2a-1 and KDED2a-3), as demonstrated in the improved affinities of the truncated versions (Table 2). For instance, the Kd of KDED7 improved from 157.3 nM to 46.8 nM, about 3-fold improvement, while KDED2 improved from 191.9 nM to 29.9 nM, a 6-fold improvement. The rest did not show observable binding to the target. In almost all the assays reported, we have used all the individual sequences (truncated and full- length) of KDED2, since the difference in binding affinity can influence the sensitivity of a particular assay.

We further examined the selectivity by testing the interaction between aptamers and cultured normal cell lines, as well as cell lines from other cancers, including leukemia, lung, ovarian, brain and cervical cancer. As shown in Table 3, among the tested aptamers, KDED19 and KC2D8 showed significant signal intensity above the control library to some of the cell lines. For instance, there was significant recognition (≥ +++) of KDED19 to CAOV3, TOV21G, CEM and H661, whereas Hela and U87MG showed reduced signal (Table 3). KC2D8 showed a recognition trend similar to that of KDED19 with the cell lines tested (Table 3). There was no observable signal from the other aptamers with any of the cell lines tested, including the normal colon cell lines (CCD18Co and FHC) and fresh human colon cells (Text S1, figure S3). This implies that most of the developed aptamers are specific to colorectal cancers (Table 3). This observation is significant as it provides many possible options for further development of these aptamers for colorectal cancer studies. Because of the similar binding pattern observed with KDED19 and KC2D8, especially the signal with CEM, we decided to use Sgc8 against these aptamers in the subsequent competition assays to preliminary verify the target molecule.

In general, the cell-SELEX strategy produces different kinds of aptamers for different targets and/or the same target present on the extracellular surface of the cells. We therefore performed competition assays to assess if any of these aptamers, especially those from different families, could influence the binding of the other. Obviously, it was reasonable to assume that sequences synthesized from the same family will compete; also, sequences that bind different cells will not compete among themselves. For instance, it is possible for KDED2 to compete with KDED15, but it is unlikely that it will compete with KDED5 or KCHA10. On this basis, we performed competition of KDED2 (FITC) against KDED7, KDED10, KDED18, and KC2D3 (unlabeled), and KCHA10 (FITC) against KDED5, KDED20, KC2D4, KDED19 and KC2D8 (unlabeled), ending with KC2D4, KC2D8 and KDED19 against CEM aptamer Sgc8 (unlabeled). Ten-fold excess of the unlabeled competitor was first incubated with DLD-1 before the introduction of the FITC-labeled aptamer. This ensured competitive advantage and the potential to saturate and block the binding of the second aptamer if they both bound the same target or if the binding of one influenced the binding of the secondary aptamer. Unlabeled KDED2 and KCHA10 were used in these assays as positive control. KDED2 binding was not influenced by any of the aptamers tested against it. Similarly, we did not observe any competition between KCHA10 and any competing aptamers. However, there was significant influence on the binding of KDED19, KC2D4, and KC2D8 in the presence of 10-fold excess of unlabeled Sgc8 (Text S1, figure S4). This result suggests that these aptamers may be binding to the same target as Sgc8. This further suggests that KDED19, KC2D4 and KC2D8 will compete among themselves, although we did not perform such competition assay. These aptamers were developed using DLD-1, which did not have the initial blocking using Sgc8. This supports the idea that it is practicable to block a known marker in order to allow the development of probes for targets of interest.

Immunohistological imaging and fluorescence microscopy have been widely used in the study of solid tumors and, in particular, colorectal cancers. Therefore, we also assessed if these aptamers could be used for tumor imaging with the positive cell line. In this preliminary study, we used cultured cell lines. Here we performed binding assays in culture dishes similar to the selection protocol, but with cell confluence of over 60%. After washing, the signal was detected with PE-streptavidin conjugate or streptavidin-Alexa Fluor 633. Figure 6 shows the confocal images of KDED2, KDED3, KDED5 and KDED7 detected with PE-streptavidin. There was significant signal strength of the tested aptamers compared with the unselected library. The signal pattern shows that the aptamers bound to the surface of the cells attached to the culture dish.

Similarly, significant fluorescence signal was observed from streptavidin-Alexa Fluor 633 conjugates when DLD-1 was used against the other aptamers, including KCHA10, KC2D8, KDED18, KDED19, and KDED20 (Figure 7), as well as HCT 116 cells with KCHA10 and KC2D8. As expected, KDED2 did not bind HCT 116. The intensity of the fluorophore signal followed the same pattern as the flow cytometry. Interestingly, KC2D8 showed a stronger signal with HCT 116 than with DLD-1.

It is usual to assume that aptamers selected against tumor cell lines bind to surface proteins. This has been demonstrated in most of the SELEX protocols involving tumor cell lines [44], [45]. In order not to wholly translate any one SELEX data to the other, we performed assays to preliminarily determine the molecules on the cell surface to which the selected aptamers would bind. In this assay, the DLDL-1 cells were treated with trypsin and/or proteinase K for 15 min at 37°C. After the incubation period, the protease activity was stopped with the addition of ice cold culture medium containing FBS. The cells were quickly washed twice by centrifugation and then incubated with the aptamers. The untreated cells were used as positive control. Figure 8 shows the response of the aptamer binding after the protease activity. Except for KDED5 and KCHA10, all the other aptamers lost recognition in both trypsin- and proteinase K-treated cells. The fluorescence signals reduced to the background, indicating that the treatment of cells with the proteases caused digestion of the target protein. For KDED5, there was significant reduction in signal intensity, but the KCHA10 signal reduced only marginally, indicating that the targets were not wholly affected by the treatment.

We envisaged two possible causes: 1) insufficient time of protease treatment and/or 2) target molecules having significant portions other than protein, such as carbohydrate or heavily glycosylated protein not wholly exposed to protease digestion. In response, we performed long time (30 min and 1 hour) protease treatment with higher concentrations of proteinase K, as well as glycosidase treatment. The increase in protease concentration, as well as long incubation time, did not affect the binding of these two aptamers. In addition, the incubation of the cells with O-linked and N-linked glycosidases followed by protease treatment did not significantly influence the binding of these two aptamers. We believe that the glycosidase assays cannot be conclusive since we did not find literature to support the efficiency of this assay using cultured cell lines instead of pure glycoprotein.

The development of aptamers that will bind to the target at varying conditions, especially at physiological temperature and in culture medium, is important. This will increase the flexibility with which these aptamers can be adopted and implemented in many assay platforms. We therefore assessed the binding of the aptamers at 37°C, in culture medium, and under both conditions simultaneously. As shown in figure S5 (Text S1), KDED2, KDED2a, KDED2a-1 (same aptamer, but different sequence lengths) and KCHA10 maintained significant binding to DLD-1 cells. The signal strengths were not significantly different from the assays at 4°C. On the other hand, the other aptamers had reduced signal intensity to DLD-1. The signal difference may be a result of the differential affinity of individual aptamers. For instance, KDED2 (better affinity) and KDED15 belong to the same family, but KDED15 did not show significant binding at 37°C. In the RPMI-1640 experiments, KEDE2, KDED2a and KDED2a-1 showed equal binding strength, even when the incubation was performed at 37°C. KDED5 and KCHA10 showed reduced, but significant binding, as opposed to the remaining aptamers (Text S1, figure S6), some of which showed almost no binding. This observation is important in the further development of aptamer-based assays, such as targeted therapy, apoptotic and viability assays at physiological conditions.

Systematic evolution of nucleic acid probes for molecular recognition has generated a large number of useful aptamers for various applications in diverse fields, including biotechnology, biomedicine, pharmacology, microbiology and chemistry. We have used cell-SELEX to generate a panel of DNA aptamers that have specific recognition to colorectal cancers. In this report, we used DLD-1 and HCT 116 as the target cell lines, and the selection was performed in the culture dish. We believe that this system is an accurate representation of the native state of the surface markers. In one of the selections with DLD-1, we introduced negative selection with the HCT 116 cell line with the aim of selecting a panel of aptamers with diverse recognition patterns to colorectal cancers. Traditionally, in order to ensure the efficiency of negative selection, the control cell line is often in 5- to 10-fold excess of the target cell line. However, in these schemes, because the selection was done in the culture dish, the incubation volumes restricted the size of culture dish we could use. Consequently, only about 2-fold excess of control cell line was used. We therefore envisaged that the ratio of the positive and the negative cell lines was insufficient to potentially eliminate a significant number of the sequences binding to the common markers. However, we believe that it provided the opportunity for us to enrich for aptamers that could have differential binding patterns to colorectal cancer cell lines.

Sequential binding and elution of the DNA library pool with positive and negative cells eventually produced DNA-enriched pools that potentially bind to the target with high fluorescence signal intensity, but with minimal recognition to the control. In all flow cytometry binding assays for selected pools and aptamer candidates, cells were removed from culture dishes with either short time trypsin treatment (30 sec –1 min) at room temperature or by using commercial non-enzymatic dissociation buffer. We noticed that mild short time enzymatic treatment did not have any observable effect on the binding of any of the aptamers and that the fluorescence signal was similar to that of the cells in the non-enzymatic buffer. Consequently, we preferred using the trypsin in most of the assays since the non-enzymatic buffer was slightly harsh on cells, especially when left for a prolonged time. We generated a panel of DNA aptamers with diverse binding patterns to colorectal cancer cell lines. All four aptamer families (KDED2/KDED15; KDED7/KDED18; KDED9/KDED10 and KC2D3) showed specific binding to only DLD-1 cells, but not to the other colorectal cancer cell lines or the other cancer cell lines tested. In addition, one aptamer, KCHB10, showed specific binding to only HCT 116 cells. The other representative aptamer families, KDED3/KDED19; KDED1/KDED5, KDED20; KCHA10, KC2D4 and KC2D8, however, showed recognition to other colorectal cancer cell lines (HCT 116 and HT-29). CRC is a heterogeneous complex of diseases [1]; therefore, it is useful to find aptamers that have differential binding patterns. The importance of developing such panel of disease probes is that a combination of them can give high predictive value of disease management procedures [7], [46]. Such approach has been successfully implemented by using proteomic profiling to detect the risk of ovarian cancer [47]. A similar approach was demonstrated by detecting DNA from CRC stool sample using a multi-target DNA analysis panel. We believe that these sequences, when further developed, will contribute to efforts in developing more effective and reliable CRC disease management regimens.

Our flow cytometry data correlated with confocal microscopy imaging. We observed that aptamers which had high signal intensity by flow cytometry also produced brighter fluorescence signal by confocal microscopy. This is very important because of the fast turnaround time with flow cytometry coupled with its high sensitivity compared with immunohistochemical staining of tissue sections. This is supported by the study carried out by [48] which showed sensitivity of tissue-dissociated cells by flow cytometry compared to immunohistochemical staining. In their report, the mean fluorescence intensity of anti-EP4, anti-HLA-ABC, anti-HLA-DR and anti-CD80 was not affected by the enzymatic dissociation solution. However, although the mean fluorescence of anti-CD54 was reduced by 30% after 1 hour of enzymatic treatment, the signal was still significantly higher above the control. Therefore, in the future, we envisage that clinical CRC specimens can be dissociated and that the cell suspension can be used in flow cytometry analysis for faster and easier disease assessment or diagnosis.

The role of membrane and membrane-associated proteins in invasive and metastasis potential of tumor cells can be a major prognostic indicator in many cancers. Cell-SELEX can play a major role in identifying such potential markers through the development of aptamer probes. Preliminary determination of the target surface molecules of these aptamers indicates that they bind to membrane proteins. With the exception of KDED5 and KCHA10, which marginally reduced in signal strength, the rest of the aptamers completely lost recognition after protease treatment, meaning that the target was digested by the treatment. We envision that KDED5 and KCHA10 might be associated with glycoproteins which might be shielded by an extensive glycosylation, although the results from the glycosidase assay do not support it. We believe that this assay was not effective enough when adopted for whole live cells.

CRC is one of the first cancers to use tumor markers to aid management. Generally, however, because of the lack of very specific tumor markers, the needed advancements have not been realized. While the leading marker, CEA, has been used extensively to determine prognosis and to monitor disease progress and therapy after curative resection, it is not sufficiently specific based on its elevated level in other conditions, such as hepatitis, pancreatitis, inflammatory bowel disease and obstructive pulmonary disease, as well as pancreatic, gastric, lung and breast cancer. Similarly other markers such as KRAS had been suggested as prognostic markers, but KRAS mutations have been observed in several cancers and its importance as prognostic marker is still controversial. It is therefore important to observe that most of the aptamers developed in this report show significant specificity to CRC. Except for KDED19, KC2D4 and KC2D8, which bind to most of the cell lines tested, the rest of the aptamers are very specific to only CRC, with some recognizing a particular CRC cell line, as demonstrated by the specificity assays, but none of them bound to the normal fresh colon and normal colonic epithelial cell line, demonstrating that the target markers of these aptamers are disease-related. This initial characterization of the selected aptamers is very important since it provides key information regarding the potential use of these markers, though further studies may be necessary to determine their diagnostic and prognostic significance.