E-cadherin is a cell-cell adhesion protein that plays a prominent role in cancer invasion. Inactivation of E-cadherin in breast cancer can arise from gene promoter hypermethylation or genetic mutation. Depending on their E-cadherin status, breast cancer cells adopt different morphologies with distinct invasion modes. The tumor microenvironment (TME) can also affect the cell morphology and invasion mode. In this paper, we used a previously developed microfluidic system to quantify the three-dimensional invasion of breast cancer cells with different E-cadherin status, namely MCF-7, CAMA-1 and MDA-MB-231 with wild type, mutated and promoter hypermethylated E-cadherin, respectively. The cells migrated into a stable and reproducible microfibrous polycaprolactone mesh in the chip under a programmed stable chemotactic gradient. We observed that the MDA-MB-231 cells invaded the most, as single cells. MCF-7 cells collectively invaded into the matrix more than CAMA-1 cells, maintaining their E-cadherin expression. The CAMA-1 cells exhibited multicellular multifocal infiltration into the matrix. These results are consistent with what is seen in vivo in the cancer biology literature. In addition, comparison between complete serum and serum gradient conditions showed that the MDA-MB-231 cells invaded more under the serum gradient after one day, however this behavior was inverted after 3 days. The results showcase that the microfluidic system can be used to quantitatively assess the invasion behavior of cancer cells with different E-cadherin expression, for a longer period than conventional invasion models. In the future, it can be used to quantitatively investigate effects of matrix structure and cell treatments on cancer invasion.Electronic supplementary materialThe online version of this article (10.1007/s10544-019-0450-5) contains supplementary material, which is available to authorized users.