Multidimensional Gene Expression Models for Characterizing Response and Metastasis in Solid Tumor Samples

Tumor Heterogeneity and Primary Versus Metastatic Evaluation of PD-L1

Logo (1)

Lorenzo Colarossi ( 1 ), Eleonora Aiello ( 1 ), Marie Cumberbatch ( 2 ), Cristina Colarossi ( 1 ), Chris Womack ( 2 ), Lorenzo Memeo ( 1 ).
( 1 ) Mediterranean Institute of Oncology, Viagrande, Italy; ( 2 ) Tristar Technology Group LLC, Washington, DC


Immunotherapy with checkpoint inhibitors allowing recovery of effector cell function, has been demonstrated to be highly effective in many tumour types and represents a true revolution in oncology.

However, current assays for the prognostic and/or predictive role of tumour PD-L1 expression are not fully standardized with respect to either quantity or distribution of expression, or regarding the concordance of expression in primary tumour versus metastasis.


In order to assess tumour heterogeneity, we evaluated PD-L1 expression using the Ventana SP-263 assay in paired formalin fixed paraffin embedded (FFPE) blocks from the same primary tumour for each of 22 lung squamous cell carcinoma (SCC), 6 lung adenocarcinoma (ADC), 53 gastric ADC, 55 colorectal ADC, 60 urothelial bladder carcinoma, 3 pancreatic ductal ADC and 4 head and neck (H&N) SCC. In addition, PD-L1 expression in primary tumour and synchronous metastasis was evaluated in 21 cases of lung neoplasm (10 ADC and 11 SCC), 30 cases of gastric adenocarcinoma, 55 cases of colorectal adenocarcinoma, and 15 H&N SCC, each with corresponding synchronous metastases. PD-L1 expression was determined by pathologist evaluation of the percentage of positive tumour cells, with those samples expressing ≥ 25% tumour positivity considered positive. Non-concordance was defined as those cases where the PD-L1 scores for the paired samples crossed the 25% cut-off.



In this limited cohort of tissues, the highest frequency of positivity for PD-L1 was observed for lung cancer and H&N cancer, with approximately 18% and 21%, respectively, exhibiting tumour membrane PD-L1 scores greater than or equal to 25% (Figures 1 & 2). Approximately 10% of gastric samples were considered positive for tumour membrane PD-L1, 8% of bladder cancer cases and 2% of colorectal ADC.

When PD-L1 tumour membrane staining greater than 1% is considered positive, then the frequency of positivity increases to 55% for lung cancer, 63% for H&N cancer, 30% for gastric cancer, 23% for bladder cancer and 4% for colorectal adenocarcinoma.


A cut-off of ≥ 25% tumour membrane PD-L1 was applied to define positivity. All paired tissues were then scored positive (≥ 25%) or negative (< 25%). Cases were recorded as non-concordant where the paired tissue outcome of positive or negative did not correlate. Concordance between the scores for each pair of blocks from the same tumour, or for primary versus synchronous metastases, can be viewed in the graphs displayed in Figures 1 and 2, respectively. Details for the non-concordant cases are summarised in Tables 1 and 2).

As summarised in Table 1, for lung carcinoma, no cases of ADC or SCC exhibited a change in PD-L1 classification when two blocks from the same tumour were evaluated. In gastric ADC we recorded a change in classification in 4 out of 53 cases, and for CRC ADC we found a difference between the two blocks evaluated in just 1 of the 55 cases. For bladder carcinoma, 1 out of 55 cases was non-concordant. For the remaining tumour cohorts (4 H&N SCC and 3 pancreatic ADC) 100% concordance was observed between two blocks from the same tumour. Evaluation of primary tumours and synchronous metastasis revealed no differences in PD-L1 scores for the 21 cases of lung carcinoma (Table 2). In the 30 gastric ADC cases, there was a discordant PD-L1 expression score for two pairs of tumours. Similarly, 1 of 55 colorectal ADC examined revealed a primary tumour PD-L1 score of 50% while the metastatic tissue was negative. One out of 15 H&N SCC cases was non-concordant.

Figure 1

Prevalence and Concordance of PD-L1 in Paired Blocks From The Same Tumour

ADC (Case 6)

SCC (Case 21)

Bladder Cancer (Case 60)

Table 1: PD-L1 Concordance in Paired Blocks

Type Diagnosis No. Cases Concordance B1 vs B2 Differences B1 vs B2
Lung ADC 6 6/6 -
SCC 22 22/22 -
Gastric ADC 53 49/53 3% vs 30%
20% vs 25%
25% vs 15%
30% vs 5%
Colon ADC 55 54/55 0% vs 50%
Bladder UROTH. CARC. 60 59/60 20% VS 30%
Pancreas ADC 3 3/3 -
Head & Neck SCC 4 4/4 -

Figure 2

Prevalence and Concordance of PD-L1 in Synchronous Metastases

Lung Cancer (Case 9)

Gastric Cancer (Case 28)

Table 2: PD-L1 Concordance in Matched Mets

TypeDiagnosisNo. CasesConcordance B1 vs B2Differences B1 vs B2
GastricADC3028/3020% vs 50%
30% vs 10%
ColonADC5554/5550% vs 0%
Head & NeckSCC1514/1525% vs 0%


  • These data demonstrate that PD-L1 expression between two sites in the same tumour, and between primary versus synchronous metastases is remarkably consistent.
  • Frequencies of non-concordance are remarkable similar whether 25% or 1% are used to define positivity.
  • The highest prevalence of tumour membrane PD-L1 positivity was observed for lung cancer and H&N cancer, with colorectal adenocarcinoma showing the lowest level of PD-L1 staining.
  • The conclusions drawn are that a representative PD-L1 score may be achieved from a single FFPE tumour block, and that for-cases where obtaining tissue from a primary tumour may be challenging, PD-L1 could be evaluated in metastatic tissue.