Neutrophils are short lived rapidly turned over immune cells. In homeostasis they are constantly replenished, with over 100 billion neutrophils per day produced in the bone marrow. Though their main role involves antimicrobial functions, they have emerged as important regulators of cancer. Here their function remains paradoxical with both pro- and anti-tumourigenic effects reported.
In health most neutrophils leave the bone marrow as mature cells which are no longer undergoing cell cycle and are doing limited transcription. Furthermore, neutrophils have capacity for long-term storage of proteins within granules, as such the transcriptomes cannot always predict effector functions. So despite the fact that single cell RNAseq has been transformational in defining different neutrophil populations, it is still true that both mature and immature neutrophil populations show a very weak mRNA to protein correlation.
This makes proteomics highly insightful to study neutrophils, but at the same time also highly complex. This is because neutrophils are small cells with low protein content (typically between 30-60 picograms), making low cell number and single cell analysis challenging. Furthermore, neutrophils display a very high dynamic range, with some proteins being extremely abundant such as S100A9 present in >200 million copies per cell, while other are present in a couple of copies. Finally, neutrophils are also loaded full of proteases such as neutrophil elastase and cathepsin G, which if activated digest proteins in a non-tryptic manner, leaving few tryptic peptides.
Yet despite the challenges, we knew there would be great value for a proteomics solution that could analyse from a low number of neutrophils down to single cells. Recent improvements in mass spectrometry (MS) based proteomics, involving sample processing, software tools and instrument sensitivity, have meant it was now possible to analyse the mini-bulk (< 1,000 cells) and single cell proteomes of primary human cells and not just cancer cell lines like HeLas. Hence, we focused on producing an MS-based workflow to perform both mini-bulk and single cell analysis of human neutrophils.
Mini-bulk is highly complementary to single cell proteomics (SCP), as it enables the proteomic ground truth signature of specific populations to be defined. Other groups have also done great work in this area showing the potential of mini-bulk (Ghosh et al., Mol Cell Proteomics). On our end, we pushed the limits to reduce the number of cells to be analysed down to 500 and showed how it can enable the study rare neutrophil populations, obtaining important insights. This workflow enabled us to stratify human neutrophils by both density and maturity. For the first time we analysed the proteome of mature and immature low-density neutrophils comparing to mature normal-density neutrophils (the normal neutrophils seen in circulation). It was hypothesised that low-density mature neutrophils had likely degranulated (explaining the low density), and thus would display much reduced granule content. Our mini-bulk data shows this is not the case. When comparing mature low-density to mature normal-density neutrophils we found no proteomic differences, suggesting that the changes in density are not protein but maybe lipid mediated. Mini-bulk showed the main component of variation for neutrophils was maturity, not density, and it enabled us to build a broad proteomic signature linked to maturity which we could then integrate into the single cell workflows.
The Mini-bulk data also suggested that the tumour associated neutrophils (TANs) in GBM much more closely resembled the mature neutrophil populations, suggesting immature neutrophils might not be infiltrating the GBM tumours. Interestingly the data also suggested the GBM TANs displayed an intermediate mitochondrial phenotype, with higher mitochondrial content than mature peripheral blood neutrophils. Whether these differences relate to adaptations to the tumour microenvironment or to the lipid rich environment of the brain remains to be determined.
We then took the leap and performed the highly complex SCP analysis of GBM TANs. Using flow cytometry-based cell sorting, individual TANs were sorted into 384 well plates containing the master mix. Directly after sorting the plate was spun and stored in the -80, meaning the samples were suitable for long term storage, with successful tests done after 10 months of storage. This is particularly important for human clinical samples, where it is not possible to predict when the next sample will be acquired. We envisage this type of SCP workflow being highly useful for other clinical samples.
In close collaboration with the Mechtler group within the Proteomics Tech Hub at the Vienna BioCenter, individual neutrophils were analysed on an Orbitrap Astral, coupled with IonOpticks columns and a FAIMS Pro Duo interface using a 50 sample per day method. The newer generation mass spectrometer was fundamental to analyse these small challenging cells. It enabled a increase of >400% sensitivity compared to the previous generation instrumentation, facilitating >1,100 proteins per neutrophil to be identified. This meant it was now really possible to probe neutrophil biology using SCP.
The SCP data was exciting, as it revealed distinct neutrophil populations which closely correlated to neutrophil functions. Importantly, many of the signatures were based on proteins that would not be detected at the transcriptomic level, such as histones and granules. Our SCP data enabled us to propose a progression of neutrophil states in the GBM tumour. Neutrophils would arrive fully loaded with granules, a population we termed ‘armed’, they would get activated in the tumour and begin degranulating, termed ‘engaged’. They could then progress to a controlled state of neutrophil extracellular traps (NETs), termed ‘vital NETs’, or an exhausted state where the main proteins required for neutrophil energy metabolism were highly reduced in abundance.
When we integrated the mini-bulk signature to the SCP data, it enabled us to identify an immature neutrophil population. The SCP data suggested this population was underrepresented in GBM tumours compared to the peripheral blood, as immature cells represented 8% of all neutrophils in GBM vs 33% of all neutrophils the peripheral blood. Furthermore, the SCP data highlighted a signature that matched a vascular restricted population, suggesting that the small number of immature neutrophils were likely stuck in the blood vessels and not the tumour itself.
(Reproduced from the main paper: https://www.nature.com/articles/s41467-025-67367-3)
The data also revealed another neutrophil population with a vascular signature. This population also had a high abundance of complement and coagulation proteins. Neutrophils are phagocytes capable of up-taking extracellular proteins and using them as fuel (Watts et al., J Clin Invest ), these proteins can be detected with proteomics, enabling an overview of the the microenvironment neutrophils are in.
Complement has been shown to trigger NETs, and this population displayed the signs of an extreme phenotype, lytic NETosis. The lytic NETs signature was marked by sharp reductions in the abundance of histones, nuclear envelope proteins, granule proteins and others. Hence, our proteomic data suggested this population was NETosing in the vasculature in proximity of blood clots. Interestingly, this proteomic profile closely matched a pro-necrotic neutrophil population recently discovered in murine lung cancer models (Adrover et al., Nature). The Adrover population was actively driving tumour necrosis, suggesting our lytic NETs population could also be causing the same phenomena in GBM and could prove to be a promising therapeutic target. Importantly, this signature is completely invisible at the transcriptomic level.
Finally, our SCP data also discovered an immunosuppressive neutrophil population, which also displayed angiogenic properties. Excitingly this classification of an immunosuppressive and angiogenic population has been also recently mapped in the NeuMap paper (Cerezo-Wallis et al., Nature ). What was also very interesting about this population is that it had signs of phagocytic activity, with increased abundance of immunoglobulin proteins as well as lysosomal and proteasomal proteins. It suggested the potential for phagocytosis in the context of GBM to correlate to pro-tumorigenic immunosuppressive neutrophils.
As a summary this is the first time SCP is used to study primary human immune cells, though an exciting implementation in progenitor and stem cells already showed great potential (Furtwängler et al., Science). Our data shows SCP can detect functional states invisible to scRNAseq and can help to define neutrophil roles, states and functions in health and disease, in this case in particular in glioblastoma. We envisage SCP becoming instrumental and transformational for immunology both for innate and adaptive immunity. The revolution has only just begun.
Final point: this work was only possible because of a great collaboration. Neurosurgeons, research nurses, research colleagues and of course the patients were all vital parts of the team. We are incredibly grateful for all the contributions everyone made for this work.