854,911-Cell DCIS Atlas Maps Route to Invasive Breast Cancer

Researchers at The University of Texas MD Anderson Cancer Center have profiled 854,911 cells from 159 patients to build one of the most detailed maps to date of the tumor microenvironment surrounding estrogen receptor-positive ductal carcinoma in situ (DCIS), the most common preinvasive, stage 0 breast cancer. The team published its results in Science under the title "A cellular atlas of the preinvasive breast cancer microenvironment."

The study matters for a clinical decision problem that has grown sharper since screening became routine: DCIS now accounts for up to 25% of all diagnosed breast cancer cases, yet only 10–30% of untreated women with DCIS progress to invasive disease. Mammography has increased detection over the last three decades, but clinicians still lack reliable markers separating lesions that will turn invasive from those that will not, leading to both overtreatment and missed risk.

The atlas covers normal breast tissue, DCIS and invasive breast cancer samples. Co-lead Nicholas Navin, PhD, chair of Systems Biology at MD Anderson, framed the gap the work addresses: "DCIS is considered a snapshot before breast cancer becomes invasive, but not all DCIS cases progress to invasive disease, and the biology surrounding these lesions has been challenging to examine."

What did the researchers actually measure?

The team used single-cell RNA sequencing on the 854,911 cells from 159 patients and applied spatial transcriptomics to 39 samples. The combined analysis identified nine cell types and 60 cell states. Co-lead Savitri Krishnamurthy, MD, professor of Anatomical Pathology, led the effort alongside Navin; co-first authors were postdoctoral fellows Siyuan He, PhD, and Aatish Thennavan, PhD, both of Systems Biology.

The measurements extend beyond pure description. The authors stated in the paper that their data "greatly improve our fundamental understanding of the DCIS microenvironment and the progression of invasive disease." They documented extensive remodeling of specific immune cells, stromal cell types and myoepithelial cells in the DCIS microenvironment, together with changes in cancer cell metaprograms that form cellular ecosystems and spatial niches supporting DCIS growth and progression.

One structural finding stands out for translational work: previously unknown "cancer stromal loops," ring-like formations of supportive tissue and vascular cells that surround DCIS lesions and create conditions favoring tumor growth. The atlas also suggests that biological features associated with invasive breast cancer begin to emerge earlier in disease progression than previously established — a data point that could shift how risk windows are defined.

How does the study fit the overdiagnosis problem?

Genetic alterations inside DCIS cells are well characterized; the surrounding microenvironment — immune cells, blood vessels and tissue cells interacting with the lesion — has remained poorly mapped. Because DCIS cells remain confined within the basement membrane of breast ductal-lobular structures, as the authors note, the transition to invasive ductal carcinoma depends heavily on what happens at that boundary. Navin said the findings "may allow clinicians to better identify and treat patients with DCIS who are at higher risk of invasive progression while also avoiding unnecessary treatments for those at low risk."

For R&D managers in diagnostics and oncology drug development, the atlas provides named cell states and spatial structures — the stromal loops in particular — as candidate biomarkers or intervention targets. The scale of the dataset, covering the full normal-to-invasive spectrum in one framework, also sets a benchmark for comparable single-cell and spatial profiling programs.

What remains unvalidated?

The authors are explicit about limits. The findings require additional research to validate, and their use in diagnostic tools, prevention strategies or treatments remains untested. No clinical performance figures exist yet for predicting progression in individual patients; the 10–30% progression estimate applies to untreated populations, not to a validated classifier. The team said it will continue studying the cellular features identified in the atlas to determine whether they can predict which DCIS lesions are most likely to become invasive and to define the best strategies for individual patients.

The next milestone to watch is a prospective test of whether stromal loops or specific cell states separate progressors from non-progressors in independent cohorts.