Early breast cancer patients show tumour HR status specific L-selectin changes circulating neutrophils

Cancer mobilised neutrophils have been reported to undergo significant phenotypic changes in patients across a range of cancers [4]. However, most of the studies investigate circulating neutrophils in cancer patients with more advanced disease and following treatment interventions.

In a mouse model of breast cancer, mice harbouring spontaneous MMTV-PyMT breast tumours that induced neutrophilia [14], we found that systemically mobilised neutrophils showed reduced surface expression of L-selectin, (CD62Llow) in various tissues (Additional file 2: Figure S2). In this proof-of-concept study, we aimed to test if similar changes in circulating neutrophils could be detected in newly diagnosed patients with EBC. To do this, we compared neutrophils from blood samples of 44 treatment-naive patients with EBC (after biopsy, but prior to surgery or neoadjuvant chemotherapy) with those from 44 paired sex and age matched healthy volunteers (HVs) collected on the same day as each breast cancer sample. Moreover, to accommodate for any confounding factors such as the effect of adrenaline or cortisol released by patients experiencing stress in anticipation of treatment, we recruited an additional group of patients who were undergoing surgery for benign breast disease. Selected patients’ data and selection criteria are listed in Additional file 3: Table S1, Additional file 4: Table S2, Additional file 5: Table S3. The neutrophil gating strategy for flow cytometry is highlighted in Additional file 1.

We firstly analysed circulating neutrophils from EBC patients and paired HVs by FACS to assess whether a similar change in CD62L could be observed in patients. In line with the neutrophilia detected in the animal model of EBC [14], neutrophils were found to make up a higher proportion of total live cells in the blood (Fig. 1a). These data reinforce the notion that neutrophils are systemically responding to the presence of breast cancer. When we analysed the percentage of CD62Llow neutrophils in patients and compared to HVs, we found that, consistent with the murine HR negative MMTV-PyMT breast cancer model (Additional file 2: Figure S2), neutrophils from HR-negative patients had significantly elevated proportion of CD62Llow subpopulations compared to HVs (Fig. 1b). Conversely, HR-positive patients had significantly lower levels of CD62Llow subpopulations compared to matched HVs (Fig. 1c). These data indicate that perturbations in the surface expression of CD62L were detectable very early in cancer and are dependent on breast cancer’s hormonal receptor status (Fig. 1d, e).

Fig. 1

a. Flow cytometry analysis showing neutrophils (CD16+ CD66B+) as a % of total live cells in the blood for 8 patients with early breast cancer compared to HVs. Statistical analysis by unpaired T-test (two-tailed). Data are represented as mean ± standard error of the mean. **p = 0.0073. b–e Flow cytometry analysis of CD62Llow neutrophil in patients with: b HR negative breast cancer compared to HVs, Statistical analysis by paired T test (two-tailed). p** = 0.0331; c HR positive breast cancer compared to HVs, Statistical analysis by paired T test (two-tailed). p** = 0.0019; (d) HR positive compared to HR negative breast cancer, Statistical analysis by unpaired T test (two-tailed). p** = 0.0011; e HVs of HR positive compared to HVs of HR negative breast cancer, Statistical analysis by unpaired T test (two-tailed). p value not significant (p = 0.9455). Data is represented as mean ± standard error of the mean

Circulating neutrophils from patients with early breast cancer show an increase lifespan

As we could detect differences in markers associated with neutrophils aging via flow cytometry in circulating neutrophils of patients with EBC, we next assessed if a difference in their properties could be detected in isolated neutrophils ex vivo. Neutrophils are reported to have a short lifespan; however, this can be influenced in steady state by different tissue context [16] as well as by factors generated during inflammation [17, 18]. We therefore measured neutrophils’ lifespan, upon their isolation from the patients’ circulation. We calculated the half-life of neutrophils in culture for 32 h and found that neutrophils from patients with EBC (Additional file 6: figure S3a-b), regardless of subtype, had a longer half-life compared to paired HVs when cultured in their own plasma (Fig. 2a). Moreover, the fold increase in the half-life of neutrophils from EBC patients mildly correlated with the neutrophil to lymphocyte ratio (NLR) values (Additional file 6: figure S3c). In addition, when neutrophils from patients with EBC were cultured with plasma from the paired HVs we still observed a significant increase in half-life compared to neutrophils from HVs (Fig. 2b). Similarly, a difference was not observed when cancer patient-derived neutrophils were cultured in their own plasma or in HV plasma (Fig. 2c). This suggests that neutrophils are intrinsically primed to live longer when isolated from blood of cancer patients. Moreover, plasma from patients with EBC did not show any effect on neutrophils from HVs (Fig. 2d, e), reinforcing the idea that a change in neutrophil behaviour originates from a more complex in vivo priming.

Fig. 2

Functional assessment of neutrophil half-life as an indicator of lifespan for patients with breast cancer and HVs across a range of subtypes. a Neutrophil half-life as measured by 50% death of neutrophils in minutes for neutrophils from cancer patients (CN) or paired HVs (HN) which are cultured in their own plasma (CP or HP respectively). Breast cancer subtypes are colour labelled. Statistical analysis using Wilcoxon matched pairs signed rank test. p* = 0.0052; b Neutrophil half-life for neutrophils from cancer patients (CN) or HVs (HN) which are cultured in paired HV plasma (HP). Statistical analysis using Wilcoxon matched pairs signed rank test. p* = 0.0353; c Neutrophil half-life for neutrophils from cancer patients (CN) which are cultured in their own plasma (CP) or paired HV plasma (HP). Statistical analysis using Wilcoxon matched pairs signed rank test. p value not significant (p = 0.1205); d Neutrophil half-life for neutrophils from cancer patients (CN) or HVs (HN) which are cultured in paired plasma from cancer patient (CP). Statistical analysis using Wilcoxon matched pairs signed rank test. p* = 0.0012; e Neutrophil half-life for neutrophils from HVs (HN) which are cultured in their own plasma (HN) or plasma from paired cancer patients (CP). Statistical analysis using Wilcoxon matched pairs signed rank test. p value not significant (p = 0.7197)

Taken together, we could detect an intrinsic priming affecting neutrophils in EBC patients which display a prolonged half-life.

Circulating neutrophils in patient with early breast cancer showed perturbations in their overall intracellular kinase activity

Following the indications of an early perturbation detected by flow cytometry in circulating neutrophils from patients newly diagnosed breast cancer along with a prolongation of their lifespan, we aimed to assess if these phenotypic changes were reflected in appreciable overall signalling pathway activities. To do this we investigated their broader intracellular kinase network activation states using the Pamgene Kinome assay. To avoid unintentional activation due to antibody binding, circulating neutrophils were isolated from blood using a negative selection strategy. Total proteins were extracted from circulating neutrophils from EBC patients and their paired HVs and loaded onto kinome chips to functionally screen their kinases activity by quantitative measurements of phosphorylated target peptides specific to either serine/threonine or protein tyrosine kinases (STKs or PTKs respectively) (Additional file 7: Figure S4a and S4b). Specific peptides that increase or decrease output of phosphorylation can be linked to the kinase activity that typically recognised them. Collectively this analysis provides a functional readout of overall kinome activity in circulating neutrophils. Additional file 7: Figure S4c shows an example of the raw STK data represented as a heatmap for the HR-positive HER-2 negative group of patients. This heatmap represents the overview of phosphorylated peptides within neutrophils in patients and HVs samples prior to normalisation of the data. Only peptides which passed quality control were Log2 transformed, and combat corrected (normalised for batch and pair effect to account for experiments done on different days) prior to further analysis. The log fold change (LFC) of phosphorylated peptides between patients compared to paired HVs was calculated for each subtype of breast cancer (Additional file 7: Figure S4d, e). This showed there appears to be some variability between patients regarding the degree of phosphorylated peptides for individual patients compared to their paired HVs.

In order to identify global patterns in the data, we used Principal Component Analysis (PCA) to assess presence of clustering for the different subtypes of breast cancer compared to HVs for the two classes of kinases, STK and PTK family (Additional file 8: Figure S5). Overall, the results did not show obvious clustering of patient and HV groups, only in HR+ and triple negative patients some level of clustering could be observed compared to their paired HVs (Additional file 8: Figure S5b-c and S5f-g).

To get a better overview of the changes in kinase activity in the different breast cancer subtypes, we used Coral trees to plot all the kinases tested. Coral trees are used for visualization of the human kinome superfamily, representing both quantitative and qualitative data [19]. Using this analysis, the differences between the neutrophil kinome in all patients with a given breast cancer subtype compared to HVs became clearer (Fig. 3). Interestingly, the entire neutrophil kinome, and particularly the PTK family, appears to be more perturbed in patients with early HR-positive compared to HR-negative breast cancer (Fig. 3a). HR-negative disease is characterized by a milder increase in activity of the TK family compared to HVs while other kinase activity is overall reduced compared to the HVs (namely kinases within the CGMC family, AGC family and CK1 kinase) (Fig. 3b). Strikingly, the presence of HER-2 positive disease appears to revert this dominant TK activation increase and induced a reduction compared to HVs in both HR-positive and HR-negative patients (Fig. 3c, d). The rest of the kinome remains more activated compared to HVs.

Fig. 3

a-d. Coral trees to show the whole kinome family (including STK and PTK families) in the indicated breast cancer sub-types compared to HVs. Each coloured node represents a specific putative kinase which is upregulated (red) or down-regulated (blue) in activity in neutrophils in patients with breast cancer compared to HVs (each subtype of breast cancer compared to HVs is shown separately). Nodal size represents the median significance value for the individual kinase. a Coral tree to show differences in neutrophil kinome in patients with HR-positive HER2 negative breast cancer compared to HVs; b Coral tree to show differences in neutrophil kinome in patients with HR-positive HER2 positive breast cancer compared to HVs; c Coral tree to show differences in neutrophil kinome in patients with HRnegative HER2 positive breast cancer compared to HVs; d Coral tree to show differences in neutrophil kinome in patients with Triple negative breast cancer compared to HVs

Collectively these data show an early neutrophil cancer-specific priming in their overall intracellular kinome activity, that appears to be different accordingly to breast cancer subtype and deeply influenced by the presence of HER-2 positivity.

The kinome activity of circulating neutrophils could possess the potential to discriminate cancer from patients with benign disease

As shown in Fig. 3, the kinome activity signatures of circulating neutrophils from patients with breast cancer are perturbed compared to paired HVs and different accordingly to breast cancer subtype. However, there are differences between the EBC patients and HVs: only patients with cancer have previously undergone a biopsy and are under the stress of a surgery at the time of blood sampling. Therefore, to test if those differences are specific to cancer patients, we collected circulating neutrophils from patients that were diagnosed with benign breast disease after biopsy. Importantly, when analysing their kinome activity, while some perturbation could be observed, we did not detect alteration specifically in TK activity compared to paired HVs (Fig. 4a). Given this data suggested a potential breast cancer specific change, we tested if the changes in kinome activity in circulating neutrophils could retain the potential to predict characteristics of the disease in patients, by performing two types of predictive model analysis. The raw data used for the analysis are provided in Additional file 9: Supplementary File 1 and included the two variables measured for each kinase activity by comparing each patient to its paired HV measured using the Pamgene platform. The two variables indicate the degree of significance for each kinase for that given patient vs paired HVs (MFS) and the direction of the change, either up or down regulated compared to HVs (KS) respectively (Additional file 9).

Fig. 4

a Coral trees to show the whole kinome family (including STK and PTK families) in neutrophil kinome in patients with benign breast disease compared to HVs. Each coloured node represents a specific putative kinase which is upregulated (red) or down-regulated (blue) in activity in neutrophils in patients with breast cancer compared to HVs (each subtype of breast cancer compared to HVs is shown separately). Nodal size represents the median significance value for the individual kinase. b Outcome of predicted randomly selected patients for the benign or cancer group using a Random Forest classifier; c Outcome of predicted randomly selected patients for the HER2 positive or HER2 negative cancer patients using a Lightgbm classifier

Two machine learning algorithms were used for training the model according to the number of patients to compare in the two groups (Additional file 10: Supplementary File 2). Firstly, we compared the 29 samples from patients of all types with the 6 samples from benign lesions. In this case we use a Random Forest algorithm to train the model, which is better suited for imbalanced data (a much fewer patient in one group). Strikingly, randomly selected patients which were not used to train the model, could be predicted to belong to either the cancer or the benign group with a sensitivity of 82% and a specificity of 66% (Fig. 4b). Moreover, this model predicted that the kinase activity changes with the stronger weight were LCK and PKG1 (Additional file 11: Figure S6a). When looking at those two kinases in the Upstream Kinase Analysis (UKA) generated list of putative kinases based on peptides targeted for phosphorylation in patients with different breast cancer subtypes (Additional file 12: Figure S7), they are indeed among the kinases perturbed in all cancer patients but do not show changes in circulating neutrophils of benign lesions.

Secondly, given the marked effect of HER2 expression on PTK activity observed in patients of both breast cancer types compared to HV, we tested if HER-2 status could be predicted in cancer patients from the kinome perturbation in their circulating neutrophils. We compared the profiles of neutrophils isolated from both sub-types of breast cancer patients, either negative or positive for HER-2 mutation using Lightgbm algorithm to train the model since the number of patients in the two group are comparable (Additional file 10: Supplementary File 2). The results showed that we could identify the HER-2 status of patient with a specificity of 73% and a sensitivity of 57% (Fig. 4c). Interesting the model predicted that the kinase activity change with the stronger weight was p70S6K beta (Additional file 11: Figure S6b).

In conclusion, by training a predictive model with only the few patients analysed here, we could show that the differences in tumour-induced kinome activity of circulating neutrophils could have the potential to distinguish between EBC and benign lesions and it could detect the presence of HER-2 mutation in breast cancer.