(G-241) Highly sensitive circulating tumor cell (CTC) enumeration in multiple cancer types using a biomimetic and nanostructured capture surface

Introduction:: Introduction
Liquid biopsy is a promising approach for minimally invasive tumor tissue sampling and has the potential to revolutionize detection and treatment selection for cancer. Circulating tumor cells (CTCs) are a particularly valuable target for liquid biopsy because they potentially carry the entire genome, transcriptome, and proteome of the parent tumor and distant metastases. However, routine detection and analysis of CTCs is hindered by the low sensitivity of current methods. For example, immunomagnetic approaches typically isolate only a handful of CTCs from a tube of blood. Detection rates of immunomagnetic techniques from patients with confirmed, late-stage cancer diagnoses are below 75% in certain gastrointestinal cancers and below 50% in pancreatic and head and neck cancers.

We have developed a capture surface for highly sensitive CTC purification that incorporates biomimetic cell rolling on recombinant E-selectin and high-avidity capture by dendrimer-conjugated antibodies [1]. The nanostructured surface is enclosed in a flow chamber to mimic circulation. CTCs are recruited to the surface by rapid binding and unbinding on E-selectin, slowed, then captured by dendrimer-mediated multivalent binding. In this work, CTCs were enumerated from patients receiving treatment for rectal adenocarcinoma, pancreatic adenocarcinoma, and head and neck squamous cell carcinoma (HNSCC). A new machine learning approach was developed to efficiently enumerate the high numbers of captured CTCs.


Materials and Methods:: Materials and Methods
Samples were collected from patients receiving treatment for gastrointestinal cancer (n = 27) or HNSCC (n = 14) during and following treatment. In most cases, samples were collected prior to radiation or chemoradiation treatment. Typically, the buffy coat from 8 ml whole blood was purified by Ficoll Paque and resuspended in culture medium. CTCs were purified from other peripheral blood mononuclear cells in flow chambers according to previously-published methods [1]. CTCs were subsequently identified following immunocytochemistry as nucleated cells positive for pan-cytokeratin and negative for pan-leukocyte marker CD45. A convolutional neural network was trained on images of 1,521 CTCs and 8,252 particles defined as one of five classes of non-CTCs with signal in the cytokeratin channel. A logistic regression model incorporating several morphological characteristics along with the probability generated by the CNN was fit to a further 600 manually-classified cells.


Results, Conclusions, and Discussions:: Results and Discussion
The machine learning algorithm achieved a sensitivity of 82.9%, specificity exceeding 99.9%, and accuracy of 99.8% in identifying CTCs. A median 50 CTC ml^-1 whole blood (range 1-542) were detected in gastrointestinal cancer patients prior to receiving radiation therapy [2], and a median 73 CTC ml^-1 (range 4-333) were detected in HNSCC patients prior to receiving monoclonal antibody therapy [3]. Healthy donor samples had a median of 0.8 CTC ml^-1. These CTC numbers were more than an order of magnitude higher than abundance reported by size-based and immunomagnetic separation techniques in similar cohorts. Moreover, CTC abundance was predictive of disease recurrence in the gastrointestinal cancer cohort at 3 months follow-up (p = 0.045) with abundance generally tracking radiographic measures of treatment response (Figure 1). In HNSCC subjects, CTC trends predicted survival in 13 of 14 HNSCC subjects (Figure 2).

Conclusion

We have demonstrated exceptionally high CTC capture sensitivity in patients with gastrointestinal and HNSCC cancers. Additionally, post-treatment CTC numbers and trends were predictive of disease progression. Results suggest that CTC abundance may be an extremely useful tool for disease monitoring.


Acknowledgements (Optional): :

References (Optional): : [1] J.H. Myung, M.J. Eblan, J.M. Caster, S.J. Park, M.J. Poellmann, K. Wang, K.A. Tam, S.M. Miller, C. Shen, R.C. Chen, T. Zhang, J.E. Tepper, B.S. Chera, A.Z. Wang, S. Hong. “Multivalent binding and biomimetic cell rolling improves the sensitivity and specificity of circulating tumor cell capture.” Clinical Cancer Research 2018, 24(11):2539-2547.



[2] M.J. Poellmann, J. Bu, S. Liu, A.Z. Wang, S.N. Seyedin, C. Chandrasekharan, H. Hong, Y.S. Kim, J.M. Caster, S. Hong. “Nanotechnology and machine learning enable circulating tumor cells as a reliable biomarker for radiotherapy responses of gastrointestinal cancer patients.” Biosensors and Bioelectronics 2023, 226:115117.



[3] M.J. Poellmann, J. Bu, D.W. Kim, M. Iida, H. Hong, A.Z. Wang, D. Wheeler, R.J. Kimple, S. Hong. “Circulating tumor cell abundance in head and neck squamous cell carcinoma decreases with successful chemoradiation and cetuximab treatment,” Cancer Letters 2023, 562:216187.