Analysis of the immune microenvironment in pre-treatment non-small cell lung cancer (NSCLC) patients with follow-up response data to second line immunotherapy

Logo (1)

M Bhagat ( 1 ), N Elliott ( 2 ), S Warren ( 2 ), C Womack ( 1 ), L Memeo ( 1 ), L Colarossi ( 1 ), M Cumberbatch ( 1 )
( 1 ) TriStar Technology Group, LLC, Washington, DC 20006, USA, ( 2 ) NanoString Technologies, Inc, Seattle, WA 98109, USA.


Patients selected based on PD-L1 expression for second line immune checkpoint inhibitor (ICI) treatment are often identified using archival specimens collected months or years prior to starting immunotherapy. Such patients may have received and failed multiple lines of standard of care (SOC) treatments with subsequent potential impact on PD-L1 expression and immune microenvironment.


We have analysed formalin fixed paraffin embedded (FFPE) tissues in a cohort of NSCLC patients taken during resection performed as first line surgical treatment for which radiotherapy, SOC chemotherapy, and second line immunotherapy clinical follow-up data are available. The immune microenvironment was evaluated by immunohistochemistry (IHC; n=18 patients) plus digital image analysis (CellProfiler™) for CD3 (2GV6) and CD8 (SP57), PD-L1 (22C3) was scored by a pathologist for tumour proportion score (TPS) and combined positivity score (CPS). Samples (n=22 patients) were analysed by NanoString using the IO360TM gene expression panel. The aim of the study was to explore whether immune signatures predictive of response to ICI therapy may be identified in such samples.


Clinical follow-up data indicated objective response to ICI therapy for 4 patients, with time from first diagnosis to receiving second line ICI treatment ranging from 5 to 102 (mean 33.6) months (Table 1). During this time these patients failed various lines of radiotherapy and SOC chemotherapy prior to receiving immunotherapy.

wdt_ID Gender Age (years) Stage at diagnosis Sub-Type Immunotherapy Time from first diagnosis to immunotherapy (months) Overall survival (OS) (in months) Progression free survival (PFS) (in months) Alive/Dead Responder status
1 M 63 IIIB NSCLC Nivo 5 6 6 A R
2 M 63 IIIA ADC Durva 16 7 7 A R
3 M 57 IB ADC Pembro 88 No data 4 No data R
4 M 71 IIA ADC Nivo 102 17 17 A R
5 M 78 IIB SCC Nivo 35 5 5 D NR
6 - 62 IIIA SCC Nivo 25 2 2 D NR
7 M 59 IIIA ADC Nivo 12 0 0 D NR
8 F 57 IIIA SCC Pembro 6 2 2 A NR
9 M 59 IIIA micropapillary ADC Nivo 12 0 0 D NR
10 M 70 IA adenosquamous carcinoma Nivo 13 7 7 D NR

While the Tumour Inflammation Signature (Ayers et al., JCI 2017) was not predictive of response (not shown), gene expression analysis did identify several signatures associated significantly with response, including increased abundance of CD8 T cells, cytotoxic cells, cytotoxicity and MHC class II antigen presentation (Figure 1). PD-L1, whether measured by NanoString (Figure 1) or IHC (Figure 2), was found not to be significantly associated with response to immunotherapy in this small cohort of samples. The frequency of PD-L1 positivity by IHC was 5/18 ≥ 1% for TPS and 9/18 ≥ 1% for CPS. In support of the significant NanoString CD8/cytotoxicity signatures, CD8 T cells by IHC were elevated significantly in the responder versus non-responder populations (Figure 3).

Figure 1

CD8 T Cell and Cytotoxicity NanoString Signatures Associated with Response


Figure 1. NanoString IO360 analysis of response. A) Forest plot of scores vs. response showing the association of signatures with response (R: responder; NR: non-responder). Points represent mean log2 fold-changes for signatures between R and NR; lines show 95% confidence intervals; larger boxes indicate statistical significance, B) Boxplot of scores vs. response,  NR;  R, and associated ROC curves illustrating the predictive performance of a signature score; predictive signatures have a curve that reaches the top left corner; shading shows 95% confidence intervals.

(Disclaimer: For Research Use Only. Not for use in diagnostic procedures.)

Figure 2

No Association of PD-L1 Expression (TPS or CPS) with Response


Non Responder


Immunotherapy: Nivo
% PD-L1 TPS: 100
% PD-L1 CPS: 100
CD8/mm2: 928




Immunotherapy: Durva
% PD-L1 TPS: 20
% PD-L1 CPS: 23
CD8/mm2: 1528

Figure 2. PD-L1 expression. Samples were stained by IHC for PD-L1 (22C3) and A) scored by a pathologist for TPS and CPS, B) representative images of positive PD-L1 expression.

Figure 3

Tumour Infiltrating Lymphocytes are Significantly Associated with Response


Figure 3. Analysis of tumour infiltrating lymphocytes by IHC. Samples were stained by IHC for CD3 and CD8. Digital images were acquired using an Aperio scanner and analysed using CellProfilerTM to deliver the number of immune cells/mm2. Data are expressed as mean cells/mm2 for NR versus R populations (A; *p<0.05), or plotted as individual data (B). Mean data values (±SE) for CD3, CD8 and PD-L1 are summarised (C). Representative images for CD3 and CD8 for NR and R cases are shown (D)

wdt_ID Status CD3 CD8 TPS (PD-L1) CPS (PD-L1)
1 R 1743 ± 410 961 ± 261 5 ± 5 6 ± 6
2 NR 743 ± 105 329 ± 57 8 ± 7 9 ± 7


  • NanoString IO360™ identified CD8 T Cell abundance and cytotoxicity immune signatures related to response to immune checkpoint inhibitor therapy
  • NanoString analyses were supported by significant increases in CD3 and CD8 T cells in the viable tumour microenvironment in responders to immunotherapy
  • Responses to immune checkpoint inhibitors was not associated with PD-L1 expression measured by NanoString or IHC. The Tumour Inflammation Signature (TIS), an analytically validated diagnostic assay which measures a suppressed adaptive immune response in the tumour and enriches for response to immune checkpoint blockade, was not significantly associated with response for these cases.
  • Taken together these data demonstrate that despite various lines of previous radiotherapy and chemotherapy spanning several years, immune profiles associated with response to second line immunotherapy can be detected in surgical first line resection samples