CTC PD-L1 Expression in Patients with MBC at Timepoint-1
Blood samples from 124 patients with MBC were assessed for CTC PD-L1 expression at timepoint-1 and subsequent timepoints in selected cases (Supplementary Fig. 1). Timepoint-1 was at or close to any time that a patient was found to have progressive disease.
Of the 124 samples at timepoint-1, 52 (42%) had elevated CTC (≥ 5/7.5ml WB) (Table 1). Twenty-one (40%) of these 52 specimens had ≥ 1% CTC PD-L1 expression of 1–2+ [median 15.2% (range 1-100%); Fig. 1; Table 1], within a semi-quantitative grading system we developed (Supplementary Fig. 2). Although phenotyping data is typically reported only for patients with ≥ 5/7.5ml WB, for patients in this study with 1–4 CTC/7.5ml WB at least one PD-L1 positive CTC was also observed in 9/30 (30%) patients (Table 1).
Table 1
CTC PD-L1 expression at timepoint-1.
#CTC/7.5ml WB
|
CTC Enumerationa
|
PD-L1 Positiveb
|
0 CTC
|
42
|
N/A
|
1–4 CTC
|
30
|
9 (30%)
|
≥ 5 CTC
|
52
|
21 (40%)
|
Total
|
124
|
30 (24%)
|
a Number of patients assessed for CTC/7.5ml WB at timepoint-1. |
b Number (%) of patients with ≥ 1% CTC PD-L1 expression at timepoint-1. |
Platelet PD-L1 Expression in Patients with MBC
In the classic CellSearch® system, leukocytes are identified by staining with fluoresceinated anti-CD45, and platelets are not visualized. During CTC PD-L1 expression analysis using CellSearch®, we observed PD-L1 staining on what visually appeared to be platelets. In extensive evaluations of WB using CellSearch® standard antibodies, as well as investigational studies using labeled antibodies against a number of other biomarkers, we have not observed platelet staining. Platelet PD-L1 was not observed in over 70 WB samples collected longitudinally from 12 healthy donors spiked with MDA-MB-231 cells and processed through CellSearch® and stained with the 29E.2A3 PD-L1 antibody. Therefore, we further investigated this curious finding.
In the CellSearch® system, a platelet specific marker cannot be used simultaneously with anti-PD-L1, since only one additional florescence channel is available for phenotyping. Therefore, using cytospins, we demonstrated that the PD-L1 positive, non-nucleated objects observed in CellSearch® co-stained with anti-platelet antibodies, confirming that they were indeed, platelets (Supplementary Materials and Methods, Supplementary Fig. 3). Platelet PD-L1 staining was not an artifact of the CellSave tube fixative (Supplementary Fig. 4 and Supplementary Table 2). When PD-L1 positive platelets were visible, they were evenly distributed throughout the CellSearch® cartridge as is visualized in images of three different frames within one CellSearch patient cartridge (Supplementary Fig. 5). Additionally, we tested the specificity of the PD-L1 antibody we employed, 29E.2A3, by siRNA knockdown and cell line staining (Supplementary Fig. 6–7, see supplementary materials for details).
The CellSearch® system is designed to enrich for cells expressing the epithelial cell adhesion molecule (EpCAM). However, CellSearch® does not enrich epithelial cells to purification, and some leukocytes, which are EpCAM negative and CD-45 positive, are routinely carried-over during the enrichment process in all samples. Hence, the addition of fluoresceinated anti-CD45 in the CellSearch® system in order to distinguish the two. However, it was not previously recognized that platelets, as well, are carried over during the enrichment process, and since platelets would not mimic CTC, no added staining is included in the CellSearch® system. Therefore, we tested the carry-over of platelets in the CellSearch® assay.
For patient samples in which platelet PD-L1 staining was observed using the classic CellSearch® method (Fig. 2A), PD-L1 staining was maintained in the aliquot containing only the platelet pellet (Fig. 2B). Staining for CD-45 (APC) and CK (FITC) in this aliquot was negative, demonstrating that the sample was clear of white blood cell (WBC) and CTC carryover (Fig. 2B). Platelet poor plasma (PPP) did not have PD-L1 staining, confirming that the PD-L1 staining was on platelets (Fig. 2C). In both platelet pellet and PPP samples, the DAPI fluorescence was seen only on DAPI coated magnetic beads (Fig. 2B-E) added to permit scanning of the samples. For patient samples that did not have platelet PD-L1 staining present in the classic CellSearch® method, neither the platelet pellet nor PPP displayed PD-L1 staining (Fig. 2D-E). Taken together, these data confirmed that the non-CTC PD-L1 staining is on platelets and that platelets do indeed carry-over during the CellSearch enrichment process.
In a subset of patients, the number of platelets carried-over into the CellSearch cartridge was quantified using a Hemavet. Platelet PD-L1 staining was independent of number of platelets within the CellSearch® cartridge as well as routine clinical complete blood count determined on the same day as the research blood collection (Supplementary Table 3).
Inter-patient platelet PD-L1 expression was heterogeneous. Using a semi-quantitative scale (Supplementary Fig. 8), 41 (33%), 48 (39%), 24 (19%), and 11 (9%) of 124 samples at timepoint-1, had 0, < 100, 100-1,000, and > 1,000 PD-L1 positive platelets/frame of the CellSearch® cartridge, respectively (Table 2). Using an arbitrarily designated cutoff of ≥ 100 PD-L1 positive platelets/frame as positive, 35 (28%) samples were positive for platelet PD-L1 expression at timepoint-1.
Table 2
Platelet PD-L1 distribution in all samples
Platelet PD-L1 Scorea
|
Blood Draw Timepoints
|
|
1
N (%)
|
2
N (%)
|
3
N (%)
|
4
N (%)
|
5
N (%)
|
124
|
59
|
16
|
6
|
2
|
Negativeb
|
89 (72%)
|
39 (66%)
|
9 (56%)
|
2 (33.3%)
|
0
|
0
|
41 (33%)
|
16 (27%)
|
5 (31%)
|
0
|
0
|
< 100
|
48 (39%)
|
23 (39%)
|
4 (25%)
|
2 (33.3%)
|
0
|
Positiveb
|
35 (28%)
|
20 (34%)
|
7 (44%)
|
4 (66.6%)
|
2 (100%)
|
100-1,000
|
24 (19%)
|
12 (20%)
|
5 (31%)
|
2 (33.3%)
|
0
|
> 1,000
|
11 (9%)
|
8 (14%)
|
2 (13%)
|
2 (33.3%)
|
2 (100%)
|
aAverage platelet count/3 CellSearch Frames (see Supplementary Materials and Methods for details). |
bArbitrary classification of platelet PD-L1 staining (see Supplementary Materials and Methods for details). N = number. |
At timepoint-1, platelet PD-L1 expression was associated with elevated CTC levels, but not with CTC PD-L1 expression. Twenty-four of the 52 (46%) samples with ≥ 5 CTC/7.5ml WB, but only 11/72 (15%) samples with < 5 CTC/7.5ml WB had PD-L1 positive platelets (p = 0.0002) (Table 3). Platelet PD-L1 expression was independent of CTC PD-L1 expression for both samples with ≥ 5 CTC/7.5ml WB (p = 0.34) and < 5 CTC/7.5ml WB (p = 0.99) (Table 3).
Table 3
Association of platelet PD-L1 score and CTC enumeration and PD-L1 expression at timepoint-1
|
Association of Platelet PD-L1 Positivity with:
|
|
CTC Enumeration
(≥5/7.5ml WB)c
|
CTC PD-L1 Expression d, e
|
N0 of CTC/7.5 ml WB
|
|
N0 of CTC/7.5 ml WB
|
0
N
(%)
|
1–4
N
(%)
|
≥5
N
(%)
|
(0–4)
N
(%)
|
(≥ 5)
N
(%)
|
Total
|
CTC PD-L1 Scoreb
|
NA
|
NEG
|
POS
|
Total
|
NEG
|
POS
|
Total
|
PlateletPD-L1 scorea
|
NEG
(0 - <100)
|
61 (49%)
|
28 (23%)
|
89
(72%)
|
|
39 (54%)
|
15 (21%)
|
7 (10%)
|
61 (85%)
|
15 (29%)
|
13 (25%)
|
28 (54%)
|
POS
(100 - >1,000)
|
11 (9%)
|
24 (19%)
|
35
(28%)
|
3 (4%)
|
6 (8%)
|
2 (3%)
|
11 (15%)
|
16 (31%)
|
8 (15%)
|
24 (46%)
|
Total
|
72 (58%)
|
52 (42%)
|
124
|
42 (58%)
|
21 (29%)
|
9 (13%)
|
72
|
31 (60%)
|
21 (40%)
|
52
|
aArbitrary classification of platelet PD-L1 staining based on average platelet count/3 CellSearch Frames (see Supplementary Materials and Methods for details). |
bNumber (%) of patients with ≥ 5 CTC/7.5ml WB enumeration and < 1% (NEG) or ≥ 1% (POS) CTC PD-L1 expression. |
cChi-squared test for comparison Platelet PD-L1 NEG vs. POS according to CTC PD-L1 enumeration; p-value = 0.0002 |
dChi-squared test comparison Platelet PD-L1 NEG vs. POS according to CTC PD-L1 expression if 0–4 CTC/7.5 ml whole blood; p-value = 0.34 |
eFisher’s exact test comparison Platelet PD-L1 NEG vs. POS according to CTC PD-L1 expression if CTC ≥ 5 p-value= 0.99 |
Association of CTC PD-L1 and Platelet PD-L1 Expression with Clinical/Pathological Features
We assessed associations of CTC PD-L1 and platelet PD-L1 expression with clinical and pathological features (Supplementary Tables 4–8).
CTC PD-L1 and Clinical/Pathological Features. Only patients (n = 91) with ≥ 1 CTC/7.5ml WB at at least one timepoint were included in the analysis of association of CTC PD-L1 and clinical/pathological features. By univariable analysis, CTC PD-L1 positivity was less likely in patients with ER + compared to triple negative MBC at either the time of 1st clinical metastatic biopsy (RR = 0.33, p < 0.001) and at a subsequent, later metastatic biopsy (RR = 0.31, p < 0.001). Similarly, CTC PD-L1 was less likely in patients with HER2 + compared to triple negative MBC at the time of either metastatic biopsy (RR = 0.22, p < 0.001 for both) (Supplementary Table 5). CTC PD-L1 was also less likely in patients with bone only disease compared to patients without bone disease (RR = 0.14, p < 0.001). CTC PD-L1 was significantly increased in patients currently receiving or who had just progressed on either endocrine therapy (ET) (RR = 3.19, p < 0.001) or CDK4/6 inhibitors (RR = 4.11, p < 0.001) (Supplementary Table 5). CTC PD-L1 was not associated with anticoagulant drugs, although only 13 patients were on a dedicated anticoagulant medication (rivaroxaban, enoxaparin, apixaban, clopidogrel) (Supplementary Table 5).
In multivariable analysis, in contrast to the univariable analysis, CTC PD-L1 was significantly higher in patients with ER + compared to triple negative disease (RR = 2.56, p = 0.007) and HER2 + compared to triple negative disease (RR = 3.14, p = 0.04). In concert with the univariable analysis, it was associated with prior treatment or progression on CDK4/6 inhibitors (RR = 3.6, p = 0.008) (Table 4). Likewise, CTC PD-L1 was significantly lower in patients with bone only disease (RR = 0.09, p < 0.001) or with bone and other sites of disease (RR = 0.19, p < 0.001) (Table 4).
Table 4
Multivariable results of factors of interest with CTC PD-L1 positive rate a and Platelet PD-L1 positivity b
Characteristics
|
Category
|
Rate Ratioc or Odds Ratiod of PD-L1 Positivity (95% CI)
|
P-value e
|
CTC PD-L1
|
Most recent met hormone status
|
(Overall)
|
|
0.03
|
|
ER+, HER2- vs. Triple Neg
|
2.56 ( 1.298, 5.058)
|
0.007
|
|
HER2 + vs. Triple Neg
|
3.14 ( 1.040, 9.492)
|
0.04
|
Disease Site
|
(Overall)
|
|
< 0.001
|
|
Bone + other site vs. Other site (no bone)
|
0.19 ( 0.078, 0.462)
|
< 0.001
|
|
Bone only vs. Other site (no bone)
|
0.09 ( 0.034, 0.243)
|
< 0.001
|
CDK4/6 inhibitor
|
Yes vs. No
|
3.60 ( 1.403, 9.238)
|
0.008
|
Platelet PD-L1
|
RBC (M/ul)
|
Continuous variable
|
0.73 (0.642, 0.820)
|
< 0.001
|
a Only significant factors included in this table. See Supplementary Table 6 for full multivariable analysis. |
b Only significant factors included in this table. See Supplementary Table 8 for full multivariable analysis. |
c Rate ratio is calculated using Poisson GEE model assuming an independent correlation structure to explore the association between CTC PD-L1 and factors of interest. |
d Platelet PD-L1 positivity is binary positive (≥ 100 PD-L1 positive platelets) or negative (< 100 PD-L1 positive platelets) by CellSearch, odds ratio is calculated using GEE model assuming an independent correlation structure to explore the association between Platelet PD-L1 and factors of interest. |
e Statistical significance is any p < 0.05. |
Platelet PD-L1 and Clinical/Pathological Features. By univariable analysis, platelet PD-L1 positivity was higher in patients who had increased numbers of CTC (OR = 1.03 for each 100 CTC/7.5ml increase, p < 0.001) and in patients with ≥ 5 vs. <5 CTC (OR = 1.45, p < 0.001), but as noted above was independent of CTC PD-L1 status (Supplementary Table 7). It was significantly lower in patients with increased number of red blood cell counts (univariable OR = 0.72 for each M/ul increase, p < 0.001; multivariable OR = 0.73, p < 0.001) (Supplementary Table 7, Table 4). Platelet PD-L1 was also statistically lower in patients who were current vs. passive/never smokers by univariable analysis (OR = 0.76, p < 0.001)(Supplementary Table 7). Anticoagulant drugs did not appear to affect platelet PD-L1 expression. However, since only a single patient was on clopidogrel and NSAID or aspirin use was taken on an as needed basis and often not recorded, no association with specific platelet-affecting agents could be drawn. Platelet PD-L1 expression was not associated with any other identifiable pathological or clinical features (Supplementary Tables 7 and 8).
No attempt to associate either CTC or platelet PD-L1 expression and outcomes was performed, since patients were enrolled based on convenience and represented a vast heterogeneity of breast cancer subtypes, treatments, lines of therapy, and follow-up.
Serial Specimen CTC PD-L1 and Platelet PD-L1 Expression
Serial CTC PD-L1. Of the 124 patients enrolled, 59 had specimens assessed for CTC PD-L1 and platelet PD-L1 at multiple subsequent timepoints, ranging from 1.5 weeks to 27 months after timepoint-1. Time between subsequent timepoints as well as whether the blood was drawn when the patient was on treatment or at progression is detailed in Supplementary Table 9. Of these 59, 35 (59%) had ≥ 1 CTC/7.5ml in two or more subsequent specimens and 14/35 (40%) were CTC PD-L1 negative at all timepoints (data not shown). Five (14%) patients maintained CTC PD-L1 positivity (Fig. 3A), 9 (26%) patients converted CTC PD-L1 status from negative to positive (Fig. 3B), 4 (11%) patients converted from positive to negative (Fig. 3C), and 3 (9%) patients had CTC PD-L1 expression fluctuating from negative to positive to negative again at subsequent timepoints (Fig. 3D).
Serial Platelet PD-L1. Platelet PD-L1 positivity varied significantly over time in some but not all patients (p = 0.005). Of the 13 patients who had platelet PD-L1 positive at timepoint-1, 10 (77%) maintained positivity at subsequent blood draws (Fig. 4A). In contrast, 2 (15%) patients converted from platelet PD-L1 positive to negative (Fig. 4B) and 1 (7%) patient had platelet PD-L1 status fluctuating from positive to negative to positive again among subsequent timepoints (Fig. 4C). Of the 46 patients who had platelet PD-L1 negative at timepoint-1, 31 (67%) maintained platelet PD-L1 negativity at subsequent blood draws (data not shown), whereas 15 (33%) patients converted from having platelet PD-L1 negative to positive at a later timepoint (Fig. 4D).