Lymph node pooling: a feasible and efficient method of lymph node molecular staging in colorectal carcinoma

Colorectal cancer (CRC) is one of the most prevalent malignancies and the second cause of death from cancer in developed countries [1]. Importantly, up to 25% of curative-intended surgically treated CRC patients develop local recurrence or distant metastases within 5 years of surgery [2‐5]. Overall and disease-free survival rates after curative surgical treatment depend on the T and N stages of disease [6‐9], but lymph node (LN) status is the most powerful prognostic factor [4, 10‐12]. Thus, pathological analysis of at least 12 LNs from surgical colectomies is required to ensure a reliable pN0 stage and is among the key quality measures for CRC care [13]. Moreover, the accuracy and predictive value of assigning stage II are directly proportional to the number of LNs examined [14].

Molecular LN assessment to detect tumor burden is far more sensitive than the standard hematoxylin and eosin (H&E) analysis [15‐19]. Whereas molecular methods can analyze the entire LN, <0.5% of the LN is histologically examined. Thus, pathological nodal staging (pN) may not be representative of the LN status and can lead to false negative diagnoses [16, 19, 20].

The use of more sensitive molecular methods of LN staging, rather than H&E, would be strongly recommended in stage I–II CRC to help identify those patients at risk of recurrence [16‐21]. In addition, recently implemented CRC screening programs result in earlier stage tumor detection [22, 23]. In this setting, a more accurate diagnosis and staging is mandatory. However, given the high prevalence of CRC and the high number of LNs to be analyzed, systematic molecular LN analysis or additional diagnostic methods beyond routine H&E are far from being incorporated into pathological diagnosis, mostly because of the high cost of molecular techniques and the supplementary workload.

The molecular assay One Step Nucleic Acid Amplification (OSNA; Sysmex Corporation, Kobe, Japan) is an automated, quantitative and high-sensitive assay for detection of cytokeratin 19 (CK19) messenger RNA (mRNA). It uses the reverse transcription loop-mediated isothermal amplification method (RT-LAMP) to amplify CK19 mRNA from LN tissue lysates. The assay has been validated for breast and CRC lymph node assessment, providing results comparable to extensive histological and immunohistochemical (IHC) LN analyses [15, 24‐26]. Additionally, it makes it possible to obtain the total tumor load (TTL), defined as the sum of CK19 mRNA copies/µL from each positive LN. The TTL corresponds to the amount of tumor burden present in all the LNs of a surgical specimen [26‐28].

In this study we compared two methods of LN molecular analysis, the individual analysis of each LN, and the new approach consisting in pooling several LNs into microcentrifuge tubes for analysis (onwards pooling method). We aimed to demonstrate the efficiency of the new procedure in the analysis of colon cancer (CC) LNs in terms of time spent in the molecular analysis, as well as the feasibility of introducing this approach into pathology daily practice. Molecular LN staging provides a practical approach for a more accurate diagnosis, which would enable the use of molecular data in CRC patient’s therapeutic management.

In this study, we used a highly sensitive molecular assay for the detection of tumor burden in a large number of LNs from CC patients. We showed that the pooling strategy is an efficient and rapid method that makes it possible to perform molecular analysis of multiple LNs at the same time, saving time and laboratory resources. This method can be easily introduced into the daily practice of pathology departments.

The need to process CRC lymph nodes using more sensitive methods than H&E has been unmet so far, given the evidence of an increased risk of recurrence and poor survival associated with the presence of undetected tumor burden with H&E in regional LNs of CRC patients [16‐21]. It is time for an upgrade of the classical H&E LN diagnosis and staging, which is becoming obsolete due to its low sensitivity in detecting the presence of tumor burden within LNs, mostly as a result of sampling bias and the limited amount of LN tissue analyzed [5, 16‐20, 29]. As has been stated in various studies and meta-analyses, this problem could be solved by using molecular methods [16‐21].

Although stage I–II CRC patients without nodal invasion or distant metastases have a 93% 5-year survival rate, 15–20% of patients histologically staged as pN0 may recur within 5 years of curative-intended surgery [5, 12, 13, 30, 31]. The importance of a precise pathological nodal staging is critical in determining the most suitable treatment and identifying those individuals at risk of recurrence [32, 33].

The OSNA assay is an automated, quantitative, rapid, and standardized method used in pathology departments for breast sentinel LN analysis [27, 34‐36]. Authors such as Castellano et al. [36] showed that it can provide useful data to decide to perform or spare axillary dissection intraoperatively. Its main advantages are that LNs can be entirely analyzed, and that it allows the use of the TTL to quantify the total amount of tumor burden present in all LNs from a surgical specimen.

We propose the use of LN pooling for the assessment of molecular LN staging in CC as a practical and novel method of multiple LN analysis. In our hands, the OSNA pooling analysis yielded better correlation with H&E LN status than the individual approach; it showed a higher PPV than the individual cohort (33.3 vs 16.4%), with similar NPV (98.4 vs 100%). The concordance between the OSNA pooling method and the H&E LN status was above 80%.

OSNA positivity rates of pN0 CRC patients are variable in previous reported studies [15, 26, 28, 37‐41]. In our series, among pN0 cases, OSNA detected tumor burden in LNs of more than 50% cases from the individual cohort, with 20.8% being positive in the pooling group. The latter results were similar to other published studies [15, 26, 28, 37‐41]. Croner and colleagues analyzed 1594 LNs from 103 patients and found 25.2% OSNA positivity. In their study, they performed the same LN workout as in our work [28]. Güller et al. [37] analyzed 313 LNs from 22 patients with only 15.3% OSNA positive pN0 patients. They used only half of the LN for OSNA analysis and the other half for extensive H&E and IHC workout. In addition, LNs smaller than 3 mm were not analyzed by OSNA. Similar results were obtained by Yamamoto et al. [26] with an upstaging rate of 11.3% in 1925 LNs from 124 pN0 patients, analyzing with OSNA only 44.5% of all dissected LNs and also using half of the LN for molecular analysis. They also found that the molecular positivity rate increased with the pT stage, from 2% in stage I to 25% in stage IIC. The importance of the sampling bias is reflected in these studies, with higher positivity rates obtained with the increase of the amount of LN tissue and the number of LN analyzed.

In LN molecular staging, tumor load quantification can become an objective tool that could be used to stratify the risk of an individual patient. TTL values from OSNA positive cases were 5485 copies/µL in the pooling cohort and 1940 copies/µL in the individual cohort, being lower in pN0 patients from both cohorts, i.e. 1540 and 1550 copies/µL, respectively. Furthermore, Yamamoto et al. [26] found an identical TTL amount of 1550 copies/µL in pN0 patients. They also observed an increment in the TTL with the increase in the number of positive LNs, rising to 24,050 copies/µL in pN1 patients and to 90,600 copies/µL in pN2 patients. Hence, TTL represents a continuous value of the amount of tumor burden present in the LNs of a given patient, which may be more accurate than the number of positive LNs [26, 27, 42]. Accordingly, OSNA studies in breast sentinel LNs demonstrated that the TTL held higher correlation with the prediction of additional axillary metastases than the number of positive sentinel LNs [27, 42].

Other authors used a similar model with qRT-PCR and found CEA mRNA levels in LNs of stage II CRC patients as predictors of prognosis [43]. Yet, using GUCY2C RT-qPCR, Hyslop found occult tumor burden in 87.5% of pN0 LNs, although only 20.9% of patients developed recurrent disease [17]. In our study, the rate of pN0 cases that were OSNA positive in the pooling cohort was 20.8%, similar to the recurrence rates in stage II CRC patients published in the literature [5, 12, 13, 30, 31].

Establishing cutoff values that hold biological significance is mandatory in molecular assays. In CRC, Croner tentatively stated a TTL of 4250 copies obtained with OSNA as determinant of macrometastasis [28]. In breast sentinel LN models, TTL values of 15,000 to 1.2 × 10⁵ copies of CK19 mRNA have been established as the cutoff to help clinicians decide on axillary LN dissection [27, 42]. Some authors have designed nomograms correlating different predictive values, as the clinical significance of the TTL may be modulated by other pathological factors [44]. Although our results are preliminary and not yet validated, we observed that TTL values of at least 7500 copies may hold a predictive value of recurrence in CC (AUC = 0.722; personal communication). Nevertheless, clinically oriented studies with long-term follow up are mandatory to validate these data.

The TTL has also been associated not only to the LN status, [26] but to different histological risk factors. In the present study, the presence of perineural and lymphatic invasion were independent factors related to TTL in both cohorts (p < 0.001; data not shown). In addition, in a recent multicenter study, our group showed a relationship of the TTL with size and pT stage in low grade tumors [45].

The study of multiple LNs at once reduces the molecular analysis to a single or few tests. This may help lessen some technical drawbacks such as the molecular evaluation of a large number of LNs separately, which may increase the likelihood of potential false positives. We obtained a much better overall accuracy with the H&E LN status in the pooling cohort than in the individual cohort. Additionally, the TTL of both cohorts did not statistically differ, and among pN0 cases, the TTLs were almost equal between them, and also with pN0 cases from other studies [26]. Thus, the pooling method may help to better assess the correct status of LNs in terms of LN positivity rates, as well as to provide reliable TTL data.

LN pooling also overcomes the second and most important drawback involved in performing molecular LN analysis in CRC, that is the amount of time invested in the analysis and the high costs of personnel and associated materials required. Patients’ TTL can be assessed with few molecular tests, analyzing at least the 12 LNs required. In our hands, the analysis was reduced to a median of only 2 molecular tests per case, and the time spent in the analysis was also greatly reduced, saving about 3 h of LN analysis per patient. The use of the pooling method enables the analysis of all LNs from 7.5 patients in the time needed to analyze one patient using the individual LN analysis. This represents a drastic drop of more than 85% of time and technical workload saved in molecular tests.

Our study has some drawbacks. Firstly, as stated, the two cohorts were not completely similar in terms of stage and tumor characteristics. In the individual cohort, cases with grossly positive LNs were discarded and pN1 corresponded to microscopic findings. Contrarily, the pooling cohort also included cases with gross positive LNs, as the inclusion criteria was made on the basis that fresh LN procurement could be performed. But although the pooling cohort had more high-grade cases and higher LN yields, it showed better correlation between the OSNA assay and the H&E evaluation of the LNs than the individual cohort. Secondly, the lack of follow-up of the pooling cohort makes it impracticable to interpret the significance of its results. Further studies are needed to obtain a predictive high-risk scale for stage I–II CC patients based on the TTL. As we lack the predictive values of the OSNA assay in colon cancer, pN staging is still mandatory, and LNs cannot be entirely submitted for molecular analysis.

Including in situ neoplasms and colon adenomas can be regarded as an additional drawback. However, there are some reports that these neoplasms may exceptionally harbor risk of metastatic disease [46, 47]. In addition, studies on circulating DNA have shown that in situ lesions may also harbor systemic disease in plasma [48, 49]. One of them included 85 patients (26 colorectal adenomas, 24 CRC and 35 normal patients). They found that adenomas had higher levels of total DNA, mitochondrial DNA, and Alu 247 bp fragment than normal controls, with a ROC curve of 0.797 [48]. Another study performed BEAMing PCR to detect mutant APC circulating DNA in patients with adenomas and metastatic and localized CRC and found that CRC had higher mutant APC DNA than controls. Nevertheless, 1 of the 11 adenomas had higher mutant APC DNA than controls [49]. They hypothesized that these fragments may derive from necrotic or apoptotic tumor cell DNA engulfed by macrophages and later released to de media [49]. Hence, the presence of low tumor burden detected by OSNA or other molecular techniques may not hold prognostic significance. For this reason, molecular analysis of LNs from precursor lesions and in situ neoplasms in our study could provide valuable data, as it may help to compare these lesions with early colon cancer, and to elucidate predictive values in this setting.

We believe that molecular LN staging in CRC performed with the OSNA analysis can replace the current pN histological staging in the near future. In order to be able to incorporate this new technology into the current pathologic staging, two conditions should be accomplished; firstly, to demonstrate that the OSNA results correlate with disease recurrence, with adequate positive and negative predictive values. Secondly, although the OSNA assay is easy, automated, fast, reproducible, and is already used in pathology departments for breast sentinel LN analysis, the high number of LNs to be analyzed in CCR, its high costs and time invested in molecular analysis were major drawbacks. In this study, we introduce the LN pooling approach with the OSNA assay to overcome these problems.

In conclusion, our study shows that the OSNA pooling makes it possible to analyze a high number of LNs from surgical colectomies with few molecular tests per patient. It also enables quantification of the amount of tumor burden present in LNs, measured as the TTL. We propose the pooling strategy as a rapid method, which can be routinely performed in pathology departments. It enables the integration of more accurate molecular results into pathological LN staging. This would benefit the therapeutic management of patients.