Part VI · Heterogeneity, evolution, and metastatic biology · Chapter 27

Spatial heterogeneity

What varies across a tumour, at what scale, and what a needle can therefore establish.

1 · Regional variation within the primary tumor

A primary tumour is not one population sampled repeatedly. Multiregion whole-genome and targeted sequencing of 303 samples from 50 primary breast cancers found that the extent of subclonal diversification varied between cases and followed spatial patterns¹. Some tumours were close to uniform across regions. Others carried distinct subclones in different parts of the same mass.

One number from that study belongs in any discussion of molecular profiling. In 13 of the 50 cancers, a potentially targetable mutation was subclonal rather than clonal¹. A targetable alteration reported from one region may therefore be present in a minority of the tumour. The report does not distinguish the two situations, because a single specimen cannot.

The structure of that variation is usually coarse rather than continuous. Single-nucleus sequencing of 1,000 cells from tumours in 12 patients with triple-negative disease found one to three major clonal subpopulations per tumour, together with a minor population of non-clonal cells². Regional variation is therefore better pictured as a small number of large populations distributed unevenly than as a smooth gradient.

The same study found that landmarks of progression, including chemoresistance and invasive potential, arose within subclones detectable in antecedent lesions¹. Regional variation is not decoration. It is where the clinically important events are already sitting.

2 · Multifocal and multicentric disease

Separate foci raise a question that a single-focus tumour does not. Are they one cancer sampled twice or several cancers in one breast.

Targeted sequencing of 360 genes across 171 tumour samples from 36 patients with ductal multifocal disease answers it partially³. The lesions in each patient were selected to share grade, ER status and HER2 status, so phenotype was held constant by design. Twenty-four of the 36 patients had substitutions or indels shared by all their lesions, and 11 of the 36 carried the same mutations in every lesion. The remaining 12 patients shared no substitution or indel between lesions, with differences in oncogenic mutations affecting genes including PIK3CA, TP53, GATA3 and PTEN. Genomically heterogeneous lesions tended to lie further apart in the gland than homogeneous ones.

Identical pathology therefore does not establish clonal identity. In one third of these phenotype-matched cases the lesions were unrelated at the resolution of that panel.

The phenotype is not always matched either. One series assessed ER, PR, HER2 and Ki-67 on every focus in 387 patients with multifocal or multicentric disease. Of those, 93 patients, or 24.0%, were heterogeneous for at least one of the four markers⁴. Disease-free survival at a median follow-up of 36 months was 81.2% in the heterogeneous group against 96.5% in the homogeneous group. The adjusted hazard ratio was 2.95, with a 95% confidence interval of 1.04 to 8.37. The study was retrospective, the follow-up was short and the interval is wide, so the size of the effect is uncertain. The direction is consistent with the genomic finding.

Assigning a tumour's phenotype from its largest focus is a convention. It is not a result.

3 · Spatial variation in receptor expression

A receptor result is the product of a chain of reductions. Staining is assessed per cell. Cells are summarised over a field. Fields are summarised over a section. The section becomes a percentage and then a category. Each step discards spatial information, and only the last number reaches the report.

That chain matters most near a threshold. The ER boundary sits at 1% of tumour nuclei staining, with a separate category for 1% to 10%⁵. A tumour staining 3% in one block and 0% in another crosses a category boundary without any large change in underlying biology. The category moved further than the tumour did.

The scale of genuine regional variation in receptor expression can be measured. Scoring ER, PR and HER2 patterns across 280 microdissected regions from 33 tumours produced a receptor heterogeneity score that was lower in grade 3 tumours than in grade 2 tumours⁶. Higher grade was associated with more uniform receptor expression rather than less.

HER2 is the receptor whose spatial variation has been characterised in most detail, including the clustered, mosaic and scattered arrangements and their separate consequences for sampling and for drug delivery. That material is in HER2 heterogeneity and is not repeated here. Its general lesson transfers to ER and PR. A percentage reported without an arrangement has discarded the information that separates a sampling problem from a delivery problem.

4 · Spatial variation in immune context

The immune infiltrate varies across a tumour. The useful question is at what scale, because that determines whether more sampling would help.

One study addressed the question directly. Multiplexed quantitative immunofluorescence for CD3, CD8 and CD20 was performed on 93 samples taken from different areas of 31 surgically resected primary breast carcinomas, and the variance was decomposed rather than described⁷. Between 66% and 69% of the variance sat between fields of view within a single section. Between 30% and 33% sat between biopsies from different regions of the same cancer. Differences between sections of the same biopsy were negligible.

Two conclusions follow, and they point in different directions. Immune infiltration is highly variable at the scale of a microscope field. Because most of that variance is local rather than regional, the mean from one biopsy estimates the whole-tumour mean reasonably well. Concordance for high against low marker status between a single biopsy and all three biopsies combined gave kappa values of 0.705 for CD3, 0.655 for CD8 and 0.603 for CD20⁷.

A mean-based immune score is therefore comparatively robust to where the needle went. Anything defined by arrangement is not. An excluded phenotype, an invasive front and a cellular neighbourhood are properties of local architecture rather than of an average, and they are developed in Spatial organization and what it adds. Sampling adequacy depends on which immune quantity is being asked for.

5 · Primary and metastatic discordance

Comparing a primary with a metastasis introduces time and treatment alongside space. The three cannot be separated by a paired sample, and pretending otherwise is the commonest error in this literature.

The genomic picture is reasonably clear. Sequencing of 299 samples from 170 patients with locally relapsed or metastatic disease addressed the timing directly⁸. The clones seeding metastasis or relapse disseminate late from the primary. They then continue to acquire mutations, largely through the same mutational processes already active in the primary. Most distant metastases carried driver mutations not seen in the primary, drawn from a wider repertoire of cancer genes than the early drivers. Some of those were clinically actionable, and they included inactivating events in the SWI-SNF and JAK2-STAT3 pathways.

A synchronous comparison partly isolates the spatial contribution, because almost no time and no treatment separate the two specimens. In 148 primary breast carcinomas assessed alongside their synchronous axillary nodal metastases, overall HER2 concordance was 95.28%, with discordance in 7 of the 148 cases⁹. Three of the discordant cases were negative in the primary and positive in the node, and four were the reverse.

That figure is the useful comparator. Set it against the much higher discordance rates for distant metachronous metastases in Discordance rates for ER, PR, and HER2 between primary and metastasis. The difference roughly measures what time, treatment and site add to the effect of sampling a different part of the disease.

6 · Intermetastatic discordance within one patient

Deposits in one patient can differ from each other, and the reason matters for what a single biopsy represents.

Whole-exome sequencing of multiple metastases obtained through a rapid autopsy protocol in 5 patients found two distinct patterns¹⁰. In 3 patients the data supported a monoclonal origin, with the metastases descending from one founding population. In 2 patients the metastases arose from at least two distinct subclones of the primary. Those 2 primaries showed mixed histological and pathological features, which suggests divergence that was already present in the breast and was carried into more than one lineage.

The series is small. Its value is that it identifies the two situations rather than averaging them. Where seeding was monoclonal, one deposit is a reasonable proxy for the others. Where seeding was polyclonal, it is not, and no routine test distinguishes the two at the bedside.

Receptor discordance between sites in the same patient follows the same logic and is quantified in Discordance across metastatic sites within the same patient. The clinical expression of intermetastatic divergence is mixed response, where some lesions regress while others progress under one regimen. Read through this section, mixed response is an expected consequence of polyclonal seeding rather than an anomaly, and it is developed as divergent evolution in Convergent and divergent evolution and what each implies for sequencing.

7 · Sampling limitations and what a single core can and cannot tell you

It is worth being explicit about what a core supports and what it does not.

A core can support the presence of what it detects. A clonal alteration found in a core is probably clonal. A strongly positive receptor result is unlikely to be an artefact of which region was sampled.

A core cannot support the absence of a subclone. It cannot give the proportion of the tumour carrying a phenotype, because it is not a random sample of the tumour. It cannot give the spatial arrangement. And it cannot establish whether a detected targetable alteration is clonal or subclonal, which was the case in 13 of 50 tumours in the multiregion series¹.

Whether additional sampling helps depends on the scale at which the variance sits. For immune markers most of the variance was finer than the biopsy, so taking more biopsies would not have removed the dominant source⁷. For subclonal genotype the variance is regional, so more regions add real information. The two questions have different answers, and a general rule about how many cores to take does not exist.

References

1. Yates LR, Gerstung M, Knappskog S, et al. Subclonal diversification of primary breast cancer revealed by multiregion sequencing. Nat Med 2015 21:751-759. PMID 26099045
2. Gao R, Davis A, McDonald TO, et al. Punctuated copy number evolution and clonal stasis in triple-negative breast cancer. Nat Genet 2016 48:1119-1130. PMID 27526321
3. Desmedt C, Fumagalli D, Pietri E, et al. Uncovering the genomic heterogeneity of multifocal breast cancer. J Pathol 2015 236:457-466. PMID 25850943
4. Li S, Wu J, Huang O, et al. Association of molecular biomarker heterogeneity with treatment pattern and disease outcomes in multifocal or multicentric breast cancer. Front Oncol 2022 12:833093. PMID 35814416
5. Allison KH, Hammond MEH, Dowsett M, et al. Estrogen and progesterone receptor testing in breast cancer: ASCO/CAP guideline update. J Clin Oncol 2020 38:1346-1366. PMID 31928404
6. Mardamshina M, Karagach S, Mohan V, et al. Integrated spatial proteomic analysis of breast cancer heterogeneity unravels cancer cell phenotypic plasticity. Nat Commun 2025 16:10482. PMID 41290667
7. Mani NL, Schalper KA, Hatzis C, et al. Quantitative assessment of the spatial heterogeneity of tumor-infiltrating lymphocytes in breast cancer. Breast Cancer Res 2016 18:78. PMID 27473061
8. Yates LR, Knappskog S, Wedge D, et al. Genomic evolution of breast cancer metastasis and relapse. Cancer Cell 2017 32:169-184. PMID 28810143
9. Ieni A, Barresi V, Caltabiano R, et al. Discordance rate of HER2 status in primary breast carcinomas versus synchronous axillary lymph node metastases: a multicenter retrospective investigation. Onco Targets Ther 2014 7:1267-1272. PMID 25050068
10. Avigdor BE, Cimino-Mathews A, DeMarzo AM, et al. Mutational profiles of breast cancer metastases from a rapid autopsy series reveal multiple evolutionary trajectories. JCI Insight 2017 2:e96896. PMID 29263308