Breast cancer metastasis progression is associated with elevated activity of kynurenine monooxygenase and kynureninase

Abstract

Introduction

Metastasis remains the major cause of death in breast cancer (BrCa) and lacks specific treatment strategies. The kynurenine pathway (KP) has been suggested as a key mechanism facilitating progression of BrCa. While KP activity has been explored in primary BrCa, its role in metastasis remains unclear. To better understand this, we examined changes in the KP of BrCa with no metastasis compared to BCa that produced local or distant metastases. Given that the cancer cell secretome plays a role in metastasis, we also investigated the relationship between changes in KP activity and serum proteins of patients with local or distant metastases.

Methods

To investigate changes in the KP in BrCa, with and without metastasis, we quantified KP metabolites in blood sera collected from patients with stage 1 BrCa (n = 34), BrCa with local metastases (n = 46), BrCa with distant metastases (n = 20) and healthy controls (n = 39). The serum protein profile of the BrCa patients with local or distant metastasis was determined before correlation analyses were carried out to examine the relationship between changes in the KP and cancer serum proteins using SPSS.

Results

We found that the KP was elevated in BrCa patients with local and distant metastasis compared to healthy controls and stage 1 BrCa patients. The activity of kynurenine monooxygenase (KMO) and kynureninase (KYNU) A was positively associated with disease stage and was higher compared to healthy controls. Proteome analysis in patients with local or distant metastasis revealed the dysregulation of 14 proteins, 9 of which were up-regulated and 5 down-regulated at the distant metastasis stage. Importantly, three of these proteins have not been previously linked to BrCa metastasis. In the correlation studies between the KP profile, cancer serum proteins and metastasis status, KYNU A had the greatest number of significant associations with cancer serum protein, followed by KMO.

Conclusion

Our findings reveal that the KP was regulated differently at various stages of BrCa and was more dysregulated in patients with local or distant metastasis. These KP activity changes showed a significant association with cancer serum proteins in BrCa patients with local or distant metastasis, highlighting the potential role of KP in BrCa metastasis.

Introduction

Breast cancer (BrCa) is the most commonly diagnosed cancer worldwide in females, representing 11.7% of all cancers and ranking first in incidence in 159 of 185 countries. BrCa also ranks fifth as the leading cause of global cancer mortality, with ~685,000 deaths yearly, accounting for 1 in 4 of all female cancer fatalities [1].

While treatment is highly successful if BrCa is detected at an early stage, current treatment strategies are not as effective in limiting and treating metastatic disease to locoregional lymph nodes or distant organs [2–4]. Distribution of BrCa metastases differs between secondary organs, with the greatest frequencies observed to bones and liver, less so to lungs/pleura and the central nervous system [5–7]. Localisation of BrCa metastases depends on the molecular subtype of the primary tumour, with cancer dissemination to bone observed in patients with all subtypes [8]. A combination of the molecular subtype of the primary BrCa (particularly, hormonal receptor status) and the organ-specific localisation of distant metastases is a strong predictor of survival time [9]. The molecular subtype of the primary tumour is not the exclusive predictor of BrCa organotropism; this is also affected by factors of the tumour microenvironment, such as cell-extracellular matrix interactions and the systemic and local immune status [5]. The challenge of BrCa is to detect metastases at the earliest possible stage [10–12]. Unfortunately, currently available imaging technologies have insufficient spatial resolution to detect micrometastatic lesions.

Blood sampling is one of the most accessible, minimally invasive approaches that potentially can be employed to address this problem. The value of serum biomarkers in evaluating the risk of metastatic recurrence in BrCa patients was clearly demonstrated in the pioneer work of Carlsson and co-authors [13]. In this longitudinal study, using a microrray of 65 serum proteins (the majority being immunoregulatory), a robust classifier 21-protein signature was identified allowing high-confidence stratification of BrCa patients with high and low risk of recurrence between 3–36 months after removal of primary tumour. Later, Noman et al. demonstrated the prognostic significance of two other elevated serum proteins (sonic hedgehog and interleukin-6) to detect metastatic progression of BrCa [14]. Considering these findings and our understanding of the role of inflammatory signalling in the BrCa tumour microenvironment in facilitating metastatic progression, we hypothesised that immune response-associated serum metabolites may be useful to non-invasively monitor BrCa progression and that kynurenine pathway (KP) metabolites could potentially be useful blood markers, due to close relationship to immune modulation [15]. There is some evidence demonstrating the involvement of the KP enzymes in BrCa biology [16–19], however, its role in the metastatic dissemination of this cancer remains unclear. It is also well documented that cancer-related secretomes (e.g., proteins) are associated with the priming of the potential secondary site(s) for metastasis [20, 21].

The KP is the major biochemical catabolic pathway of the essential amino acid tryptophan (TRP) (Fig. 1) [22]. In normal physiological conditions, the KP is tightly regulated. It plays a central role in production of nicotinamide adenine dinucleotide, the key substrate for cellular energetics [23]. However, the KP becomes dysregulated in disease-associated inflammation, including cancer [24]. Cancer cells have been shown to manipulate the KP to suppress local immune surveillance through at least three different mechanisms [25]. Firstly, overactivation of the KP enzymes, indoleamine 2,3-dioxygenase 1 (IDO-1) and/or tryptophan 2,3-dioxygenase (TDO), leads to rapid depletion of TRP in the local tumour microenvironment and consequently induces cell cycle arrest in T cells [26]. The second mechanism involves the direct interaction between T cells and KP metabolites such as 3-hydroxyanthranilic acid (3HAA), 3-hydroxykynurenine (3HK) and quinolinic acid (QUIN). This interaction has been shown to induce apoptosis and inhibit the activation of tumour-targeting T cells [27, 28]. The third mechanism consists of interaction between the KP metabolites kynurenine (KYN) and kynurenic acid (KYNA) and the aryl hydrocarbon receptor (AhR). The AhR is a critical regulator of the immune response and can be found in all immune cells [24, 29]. Activation of this receptor by KYN or KYNA suppresses the differentiation and activation of T cells [30]. Taken together, these pieces of evidence indicate that KP may be involved in the metastatic dissemination of BrCa. Although there have been 58 studies and 6 clinical trials to date examining the involvement of the KP and BrCa, only 4 studies have looked at the KP profile in BrCa metastasis [15, 29]. The first evidence of KP involvement in BrCa metastasis was provided in a 2012 study by Sakurai et al. where the authors reported that IDO-1 activity positively correlated to the number of secondary tumours. Furthermore, IDO-1 activity in BrCa patients with multiple secondary tumours was elevated compared to patients with only solitary secondary tumours [18]. In later studies Sakurai et al. also reported that the activity of IDO-1 in BrCa patients with metastases decreased after chemotherapy [16, 17]. Another study showed that elevated IDO-1 activity was not limited to tumour cells. Mansfield et al. reported that macrophages in lymph nodes of BrCa patients with sentinel lymph node metastasis showed higher IDO-1 activity which was strongly associated with the advanced stages of BrCa [19]. Furthermore, IDO-1 expression correlated positively with increased populations of immune suppressive T regulatory cells in the lymph node, strongly suggesting that IDO-1 expressing macrophages suppress the immune response in lymph nodes, allowing cancer propagation.

Fig. 1. A simplified diagram of the kynurenine pathway.

Majority of TRP is metabolised through KP to produce NAD⁺, an essential cellular energy factor. Enzymes of this pathway (red font) can be dysregulated to promote immune evasion through the overproduction of KP metabolites (black font) such as KYN, 3HK, 3HAA and QUIN that can affect anti-tumour immune response. Light blue arrows indicate enzymatic reaction.

In this study, we have examined the role of the KP in BrCa metastasis by profiling KP activity in BrCa patients with local or distant BrCa compared to BrCa patients with no metastasis. As circulating cancer-associated proteins are highly associated with cancer spread [20, 21], we also investigated the relationship between the changes in serum KP metabolites and the BrCa proteome profiles in patients with local or distant metastasis to provide insights in the role of KP enzyme(s) in BrCa progression.

Methods

Patient Cohort

Sera from BrCa patients and healthy controls was sourced from the Victoria Cancer Biobank and the Australian Breast Cancer Tissue Biobank. The study cohort comprised the following groups: 34 patients with stage 1 primary BrCa, 46 patients with locally metastatic BrCa (evidence of metastases in regional lymph nodes), 20 patients with distant metastatic BrCa (evidence of metastases in distant organs) and 39 healthy volunteers. The inclusion criteria for the stage 1 BrCa cohort included a primary diagnosis of BrCa with no history of other cancers, no current infections and no symptoms of irritable bowel syndrome, as these may affect the KP profile. The inclusion criteria for patients with locally advanced and metastatic BrCa cohorts included a history of BrCa-only diagnosis, i.e. no history of other cancers and no symptoms of irritable bowel syndrome. Samples used in this study were from female patients between 27 and 87 years old. The demographic and clinical characteristics of healthy and cancer cohorts are shown in Table 1.

Table 1.

Demographic and clinical characteristics.

	Healthy Controls, n (%)	Stage 1 Luminal BrCa, n (%)	Local metastases, n (%)	Distant metastases, n (%)
Total	39	34	46	20
Sex
F	39 (100)	34 (100)	46 (100)	20 (100)
M	0(0)	0 (0)	0 (0)	0 (0)
Age
Mean ± Standard Deviation	47 ± 5	63.7 ± 13.6	55.17 ± 14.2	48.75 ± 10.4
Healthy control vs cancer stage, p-value		p < 0.0001	p = 0.0338	p = 1
Stage 1 vs metastases, p-value			p = 0.0069	p = 0.0003
local vs distant metastases, p-value				p = 0.8021
<40	0 (0)	0 (0)	7 (15.22)	4 (20)
40–49	26 (66.67)	3 (8.8)	14 (30.43)	7 (35)
50–59	13 (33.33)	11 (32.35)	7 (15.22)	3 (15)
60–69	0 (0)	11 (32.35)	9 (19.57)	6 (30)
>70	0 (0)	9 (26.47)	9 (19.57)	0 (0)
Grade
1		34 (100)	2 (4.35)	0 (0)
2		0 (0)	8 (17.39)	1 (5)
3		0 (0)	23 (50)	5 (25)
Not available		0 (0)	13 (28.26)	14 (70)
PR status
Positive		34 (100)	27 (58.70)	8 (40)
Negative		0 (0)	9 (19.57)	8 (40)
Not available		0 (0)	10 (21.73)	4 (20)
ER status
Positive		34 (100)	27 (58.70)	11 (55)
Negative		0 (0)	9 (19.57)	5 (25)
Not available		0 (0)	10 (21.73)	4 (20)
HER2 status
Positive		0 (0)	10 (21.73)	7 (35)
Negative		34 (100)	28 (60.87)	7 (35)
Not available		0 (0)	8 (17.39)	6 (30)
Metastasis site
Brain			0 (0)	6 (30)
Liver			0 (0)	2 (10)
Lymph Node			9 (90)	1 (5)
Skin			0 (0)	1 (5)
Lung			0 (0)	4 (20)
Ovary			0 (0)	3 (15)
Bone			0 (0)	3 (15)
Primary Breast			1 (10)	0 (0)
Not available			36 (0)	0 (0)

Sample preparation for analytical chemistry

All patient sera were treated with 1 volume of 10% trichloroacetic acid (Sigma Aldrich, MA, USA) to precipate proteins. After that, the precipated proteins were concentrated by centrifuging at 12,000 × g at 4 °C for 10 mins and removed by filtering through a 0.22 PTFE syringe filter (Merck-Millipore, CA, USA). The filtered sera were transferred to liquid or gas chromatography vials for the analysis.

Ultra-high performance liquid chromatography (uHPLC) and HPLC

Metabolites were quantified by Agilent 1290 Infinity uHPLC (Agilent, CA, USA) and Agilent 1260 Infinity HPLC (Agilent, CA, USA). uHPLC was used to determine serum levels of TRP, KYN, 3-HK, 3-HAA and AA and were then quantified using a sequential diode array UV and fluorescence detection. The column compartment was set to a temperature of 38 °C and the injection volume was 20 µl. To prevent sample degradation, the autosampler tray temperature was set at 4 °C. The flow rate was set at 0.75 mL/min with an isocratic elution of 100% of 100 mM sodium acetate, pH 4.65.

UV detectors set at a wavelength of 365 nm were used to perform identification and quantification of KYN and 3-HK. Fluorescence intensity was used to detect TRP with an emission wavelength at 438 nm and excitation wavelength at 280 nm, while 3HAA and AA used an emission wavelength of 438 nm and an excitation wavelength of 320 nm. Results were expressed as µmol/L or nmol/L and calculated using interpolation on a 6-point calibration curve. The HPLC system and column was then reset with 100% of 15 mM potassium phosphate buffer with pH 6.4. The flow rate was decreased to 0.7 mL/min with isocratic elution. KYNA in serum samples was determined using HPLC where mobile phase consisted of 95% of 50 mM sodium acetate and 50 mM zinc acetate, pH 5.2 and 5% HPLC grade acetonitrile. Flow rate was now set to 1.00 mL/min with an isocratic elution. KYNA was identified with the use of a fluorescence detector with emission wavelength of 388 nm and an excitation wavelength of 344 nm. The results were calculated using the same methods and levels of KYNA were calculated and expressed as nmol/L.

Gas chromatography/mass spectrometry (GCMS)

PIC and QUIN were quantified using the Agilent 7890 GC/MS (Agilent, CA, USA). Deproteinized serum samples and deuterated internal standards were first dried under vacuum for 1 h at 45 °C and 1 atmospheric pressure. Samples were then derivatised with trifluoroacetic anhydride and 1,1,1,3,3,3-hexafluoroisopropanol for an hour at 60 °C. Toluene and 5% sodium bicarbonate were then added to all samples to separate the organic layer. MilliQ water was used to wash samples, and the upper organic was collected and dried using pipette tips which were packed with sodium sulphate and glass wool. Samples were then injected into an HP-5MS GC capillary column (Agilent, CA, USA) and analysis was carried out with the MS operating in negative chemical ionisation mode.

Sample Preparation for proteomics

BrCa sera protein concentrations were measured using a BCA Protein Assay Kit following the manufacturer’s protocol (Thermo Fisher Scientific, USA). Protein samples were reduced with 15 mM dithiothreitol for 30 min at 60 °C followed by alkylation with 30 mM of iodoacetamide (IAA) at room temperature for 30 min in the dark. Samples were then digested with trypsin at a 1:30 ratio for 16 h at 37 °C with gentle shaking. Prior to LC-MS analysis, digested peptide mixtures were desalted and cleaned with C18 StageTips method [31].

For peptide/protein spectral library generation, to increase the chances of detecting as many proteins as possible, all BrCa sera samples were combined and the top 14 high abundance proteins (HAPs) were depleted [32]. Depleted protein samples were reduced, alkylated and digested as described above. Peptides were desalted and cleaned with Sep-Pak® C18 Cartridges (Waters Corporation, MA, USA) following the manufacturer’s protocol. Peptide mixtures were further fractionated (up to 20 fractions) using high pH reversed phased C18 peptide fractionation method [32].

LC-MS/MS

Samples were analysed by LC-MS on an UltiMate 3000 RSLCnano System liquid chromatography system coupled to a Q-Exactive HFX mass spectrometer (ThermoFisher, MA, USA). Samples (600 ng) were injected onto a peptide trap (C18 PepMap 100, 5 μm, 100 Å, 300 μm × 5 mm (ThermoFisher, MA, USA) and washed with loading buffer at 15 µL min⁻¹ (0.1% formic acid in water) for 10 min, before the trap was switched in-line with the analytical column (in-house packed ReproSil-Pur 120 C18-AQ, 3 μm, 250 × 0.075 mm, 75 μm × 30 cm) held at 45 °C and peptides were separated by gradient elution at a flow rate of 300 nL/min. Buffer A consisted of 0.1% formic acid in water and Buffer B consisted of 0.1% formic acid in 99.9% acetonitrile. The gradient increased from 2% B to 35% B over 60 min, raised to 95% B over 5 min then held for 5 min prior to re-equilibration.

The column eluent was directed into the ionisation source of the mass spectrometer operating in positive ion mode with a spray voltage of 2.7 kV. High pH library fractions were analysed using a data dependent acquisition (DDA) method and samples for quantitation were analysed using a data independent acquisition (DIA) method as described below.

DDA Spectra library generation

Recombinant protein spectral library (rPSL) approach was employed to generate DDA spectral library [33]. Peptide precursors from 350 to 1650 m/z were scanned at 120 K resolution. The 10 most intense ions in the survey scan were fragmented by HCD using a normalised collision energy of 27.5 with a precursor isolation width of 1.4 m/z. Precursors with all charge states (no exclusion) were subjected to MS/MS analysis. The MS method had a minimum signal requirement value of 8 × 10³ for MS2 triggering, an AGC target value of 3 × 10⁶, maximum ion injection time of 50 ms. MS2 scan resolution was set at 30 K, an AGC target value of 2 × 10⁵ and a maximum injection time of 50 ms and dynamic exclusion was set to 20 s.

DIA

A MS1 scan was performed from 350 to 1650 m/z at 120 K resolution, followed by MS2 scans at 20 m/z ranges given in the Supplementary Table 1 and fragmented by HCD using a normalised collision energy of 27.5. The MS2 scans had scan resolution set at 30 K, an AGC target value of 2 × 10⁵ and a maximum injection time of 50 ms.

Data processing

An ion library was generated from High pH library fractions DIA data using MSFragger (version 3.5) in the fragpipe GUI (version 18.0) [34]. DDA files were searched against a database of human proteins (Uniprot proteome UP UP000005640, downloaded September 2022, containing 79740 protein sequences with all sequences reversed as decoy targets) in the default workflow, with trypsin digestion, peptide length set to 7–50, peptide mass range set to 500–5000 and max missed cleavages set to 2. Peak matching precursor and fragment mass tolerance were set to 20 ppm. Carbamidomethylation of Cys was set as a fixed modification, and oxidation of Met and acetylation of N-terminal were set as variable modifications, with maximum variable modifications on a peptide set to 3. The spectral library of b and y ions was generated from the search results using SpecLib, with automatic selection for retention time calibration, RT Lowess fraction set to 0.01, Unimod tolerance set to 0.02 Da and fragment tolerance set to 15 ppm.

Proteins were quantified using DIA-NN based on the library generated above using a double-pass neural network classifier [35]. Mass accuracy, MS1 accuracy and scan window were all set to 0 for automatic determination, match between runs options was on and the ‘Remove likely intereferences’ options on. Precursor FDR was set at 1%, protein inference was performed at a Gene level, with a Robust LC quantification strategy, RT-dependent cross-run normalisation.

Statistical analysis

As a first step in the data analysis of the KP measurements, outliers were identified using Prism 6, GraphPad Software Inc. This involves using the ROUT method to detect potential outliers while fitting a curve with nonlinear regression [36]. The maximum false discovery rate for this function was set to 5%, with the expectation that no more than 5% of the identified outliers be false. Data was analysed using one-way ANOVA and t-tests where possible. The differences between the compared groups of values were considered to be statistically significant at a p ≤ 0.05. Graphs were generated using GraphPad Prism 8 (GraphPad, San Diego, USA). Non-parametric statistics (i.e. Kruskal-Wallis test) were used for analysis when data were not normally distributed, and corrected with Dunn’s multiple comparisons test. The p value reported for each multiple comparison was the adjusted p value.

For quantitative MS proteomics analysis, results were filtered for gene groups that had observations in at least 50% of the patient cohort, and the data was normalised so that each sample median peak intensity was adjusted to the global median. Missing values in the data set were imputed by selection of a random value from the population of the average standard deviation of genes across the data set around the minimum intensity value observed. A relative abundance comparison was performed across sample groups and analysed by t-test in GenePattern SWATH workflow [37]. The workflow allows the determination of differentially expressed proteins by ANOVA analyses on log-transformed normalised protein peak areas, along with peptide-level t-tests. Proteins were typically deemed to be differentially expressed if the ANOVA p value is less than 5% and the protein fold change exceeds 1.5 [38, 39]. Benjamini-Hochberg false discovery rate -adjusted ANOVA p values were also generated but not used by default. The cutoffs used in our analysis (i.e., a combination of fold change 1.5 and p < 0.05) have been validated in spike-in experiments, specifically in the context of SWATH and demonstrated to control the false discovery rate [37] effectively. This approach finds corroboration in its application within other studies [38, 39].

Correlation analyses were performed using SPSS version 26 (IBM Corp. Armonk, NY, USA). Spearman’s correlation coefficient tests were then used for the analysis of correlations between the KP metabolites and differentially expressed proteins. Correlation data were presented as a heatmap in colour, where the tones of red represent positive correlations and the tones of blue represent negative correlation. Black-coloured font shows the data with a p ≤ 0.05, whereas purple-coloured italic font indicates the data with a p ≤ 0.01.

Results

KP enzyme activity is associated with BrCa progression

To understand how the KP is involved in the progression of BrCa, we measured the activity of the KP enzymes and concentrations of their products in sera of stage 1, local and distant metastatic BrCa patients compared to healthy controls.

The activity of the first rate-limiting enzymes, IDO1 and TDO, that convert TRP to KYN, was calculated as the ratio of KYN/TRP. Among the patient groups, patients with stage 1 primary BrCa had the highest activity of IDO1/TDO. Patients with local or distant metastatic BrCa had significantly lower levels of IDO1/TDO activity as compared to patients with stage 1 primary BrCa (p < 0.01) and healthy controls (p < 0.05, p < 0.01 respectively) (Fig. 2a), while there was no difference in IDO1/TDO activity between patients with local and distant metastatic BrCa.

Fig. 2. KP enzyme activity measured in stage 1 BrCa, local metastatic, distant metastatic and healthy controls.

To understand how KP is regulated during the progression of BrCa, we measured the activity of the KP enzymes in healthy controls (n = 39), stage 1 BrCa (n = 34), local metastatic BrCa (n = 46) and distant metastatic BrCa (n = 20). a IDO1/TDO activity was significantly downregulated in distant metastatic BrCa (b) KMO activity was significantly elevated in distant metastatic BrCa patients as compared to healthy control. Interestingly, (c) KYNU A activity has varied expression across the cohorts while (e) KYNU B is significantly up-regulated in healthy controls as compared to other cohorts. The concentration of (d) kynurenic acid (KYNA), (f) picolinic acid and (g) quinolinic acid were significantly downregulated in both local and distant metastatic BrCa as compared to stage 1 and healthy controls. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001, one-way ANOVA and t tests.

Next, we examined the first junction of the pathway, where KYN is catabolised to 3HK through KMO, or to AA by kynureninase (KYNU) A, alternatively, to KYNA, by kynurenine aminotransferases (KATs). The activities of KMO and KYNU A were calculated as the ratios of 3HK/KYN and AA/KYN, respectively.

The activity of KMO increased significantly in patients with local (p < 0.01) or distant metastatic BrCa (p < 0.001), compared to healthy controls (Fig. 2b). The increased activity of KMO was due to the increased production of 3HK. Although KMO activity trended higher in patients with stage 1 primary BrCa compared to healthy controls, it was not statistically significant. There was no significant difference between the cancer sub-groups.

Intriguingly, the activity of KYNU A was elevated in patients with stage 1 primary BrCa and distant metastatic BrCa compared to patients with local metastasis (p < 0.0001) and healthy controls (p < 0.05) (Fig. 2c). The increased activity of KYNU A was due to the elevated production of AA. There was no statistically significant difference in the activity of KYNU A between patients with local metastatic BrCa and healthy controls.

Conversely, BrCa patients with local (p < 0.0001) and distant (p < 0.0001) metastasis had the lowest concentration of KYNA in sera, followed by stage 1 primary BrCa (p < 0.0001) compared to healthy controls (Fig. 2d). This suggests that activity of KATs was downregulated during cancer progression.

Considering that the activity of KMO was elevated in all patient groups, we proceeded to explore whether the KP would also be activated downstream of KMO. Immediately downstream of 3HK is the enzyme KYNU B whose activity can be reflected by the ratio of 3HAA/3HK and the production of QUIN and PIC. The activity of KYNU B was significantly lower in all patient groups as compared to healthy controls (Fig. 2e). In particular, the activity of KYNU B in local metastatic BrCa was lowest (p < 0.0001), followed by distant metastasis (p < 0.01) and stage I BrCa compared to healthy controls (Fig. 2e). There was no significant difference in the activity of KYNU B between stage 1 BrCa or BrCa with local or distant metastasis. The low activity of KYNU B was due to a low concentration of 3HAA, suggesting KYN is predominantly catabolised to 3HK.

The level of PIC was significantly lower in BrCa patients with local (p < 0.0001) metastasis, followed by distant (p < 0.0001) metastasis and patients with stage 1 BrCa relative to healthy controls (Fig. 2f). Within the cancer groups, patients with local (p < 0.0001) and distant (p < 0.001) metastasis had significantly lower levels of PIC compared to patients with stage 1 BrCa. A similar trend was also observed in the level of QUIN in serum where patients with local (p < 0.001) or distant (p < 0.01) metastasis had significantly lower levels of QUIN as compared to healthy controls (Fig. 2g). When compared to patients with stage 1 primary BrCa, patients with local (p < 0.01) or distant (p < 0.05) metastasis had significantly lower levels of QUIN.

Given that age can be a confounding factor in the BrCa metastasis, we also performed age- and multiple-comparison-adjusted statistical analysis and tabulated this (Supplementary Table 1). In the activity of IDO1/TDO, the only statistical significant difference was detected between stage 1 BrCa and local metastasis (p < 0.01). The level of KYNU A activity remained statistically higher between stage 1 BrCa compared to local metastasis (p = 0.0104), and a new statistical difference was observed between healthy controls and local BrCa metastasis with healthy control showing a higher activity (p = 0.0122). The significant differences in the level of PIC between the cohorts remained, with an additional statistical difference observed between healthy controls and stage 1 BrCa. The observed statistical difference in the level of QUIN between stage 1 BrCa and distant metastasis as well as between stage 1 and local metastasis remained after the adjustments. However, there was no longer any statistical difference in the level of QUIN between healthy and distant metastasis and between healthy control and local metastasis. In KYNA, the statistical differences remained between healthy controls and local metastasis BrCa, between healthy controls and distant metastasis BrCa and between stage 1 and local metastasis after age adjustment. The level of KYNA became statistically different between stage 1 BrCa and distant metastasis. In contrast, the level of KYNA is no longer a statistically significant difference between healthy control and stage 1 BrCa. No statistically significant difference was observed in KMO and KYNU B activity between the cohorts. Additionally, we have also tabulated the comparison of individual KP metabolites between the cohorts in Supplementary Table 2.

Fourteen proteins were differentially expressed in sera of patients with local and distant metastatic BrCa

Cancer secretomes (including cancer-associated proteins) have been shown to prime secondary sites for metastasis and enhance treatment resistance [20, 40]. To enhance understanding of serum proteomic signatures associated with metastatic progression in BrCa, we investigated the proteome profiles in the same local and distant metastatic BrCa cohorts used for KP analysis.

In order to improve the DDA spectra library coverage, we combined all sera samples to increase the chances of detecting proteins that were present in both studied BrCa cohort groups, followed by removing the top 14 high abundant proteins to reduce the orders of magnitude of protein concentration to detect low abundant proteins [32]. Furthermore, cancer-associated recombinant proteins were added (‘spiked’) to the spectra library to enhance the chances of detecting low abundance cancer-associated proteins in the sera samples [33]. The resulting DDA spectra library contained a total of 650 proteins including spiked recombinant proteins. (refer to Supplementary Table 3 for lists of detected protein/peptides and amino acid sequences of each peptide.)

After completing the DDA spectra library, DIA-MS quantification was performed on individual samples to identify differentially expressed proteins between local and distant metastasis groups. A total of 273 proteins were quantifiable across all samples. Of those, 14 proteins were differentially expressed (p < 0.05 and a fold change (FC) ratio ≥ 1.5) between BrCa local and distant metastasis samples (Fig. 3a, b).

Fig. 3. Sera protein quantification in local and distant metastatic BrCa.

a Volcano plot representations on differentially expressed proteins (FC > 1.5, p < 0.05) between local metastatic BrCa and distant metastatic BrCa. Red dots indicate up-regulated proteins, and blue dots indicate down-regulated proteins in distant metastatic BrCa compared to local metastatic BrCa. b Box plots illustrate the up-regulated protein expression patterns between local and distant metastatic BrCa. c Box plots illustrate the downregulated protein expression patterns between local and distant metastatic BrCa.

Among these differentially expressed proteins, nine were upregulated in patients with distant metastatic BrCa, including: special AT-rich binding protein-2 (SATB2), neural cell adhesion molecule 1 (NCAM1), cystatin-C (CST3), phospholipid transfer protein (PLTP), platelet glycoprotein Ib alpha chain (GP1BA), vasorin (VASN), hepatocyte growth factor activator (HGFAC), apolipoprotein C IV (APOC4) and haptoglobin-related protein (HPR) (see Supplementary Table 4). SATB2 was the most significantly upregulated protein (increased 8.57-fold, with a p value of 0.0069) in distant metastatic BrCa, compared to local metastatic BrCa, followed by NCAM1 (FC = 4.67, p value 0.031). The proteins CST3, PLTP, GP1BA and VASN showed significantly increased expression with fold changes greater than 2. The proteins HGFAC, APOC and HPR showed statistically significant fold changes between 1.5 and 1.7. The higher expression of these proteins in the distant metastatic cohort potentially suggests a role in BrCa metastatic progression.

Our analysis also revealed that five proteins were down-regulated in distant BrCa metastasis compared to patients with local metastasis: bone marrow stromal cell antigen 1 (BST1), collectin subfamily member 10 (COLEC10), carbonic anhydrase 2 (CA2), cadherin-5 (CDH5) and Galectin-3-binding protein (LGALS3BP) (see Supplementary Table 1). Expression of BST1 (also known as ADP-ribosyl cyclase/cyclic ADP-ribose hydrolase 2) was significantly decreased by almost 10-fold (p value 0.0013) and COLEC10 was down-regulated by 5.29-fold (p < 0.001) in patients with distant metastases compared to local BrCa metastasis. The expression levels of CA2, CDH5 and LGALS3BP were reduced by 3, 1.9 and 1.5-fold respectively (Fig. 3a, c and Supplementary Table 4).

KP metabolites and several serum proteins showed statistically significant associations with metastasis status

Correlation analysis revealed multiple statistically significant correlations between KP metabolite levels, secretome proteins and BrCa metastatic progression (Fig. 4 and Supplementary Fig. 1).

Fig. 4. Correlation analysis between KP metabolites, serum proteins and metastatic characteristics.

Correlation analysis revealed that the activity of KYNU A (as inferred by AA/KYN ratio) had the highest number of statistically significant correlations to serum proteins in patients with metastatic BrCa followed by KMO (as inferred by 3HK/KYN ratio)and QUIN.

Levels of KYNU A and PIC showed weak positive correlations with the distant metastases (R_s = 0.497, p ≤ 0.01 and R_s = 0.294, p ≤ 0.05, respectively). In parallel, both KYNU A and PIC had weak negative correlations to lymph node metastasis (R_s = −0.486, p ≤ 0.01 and R_s = −0.318, p ≤ 0.05, respectively). Notably, KYNU A also demonstrated a weak positive correlation with metastases to the brain (R_s = 0.302, p ≤ 0.05).

Seven of the cancer secretome proteins had statistically significant correlations with distant metastases. Only COLEC10 showed a moderate-strength positive correlation with metastasis to local lymph nodes (R_s = 0.605, p ≤ 0.01), whilst also having a moderate negative correlation to distant metastasis (R_s = −0.574, p ≤ 0.01). In addition, weak negative correlations were found between distant metastases and serum levels of LGALS3BP, CDH5, CA2 and BST1. The list of serum proteins demonstrating weak positive correlations to distant metastases included SATB2, CST3, PLTP, VASN, HGFAC, APOC4 and HRP (see Fig. 4).

Our work also revealed associations between the KP and the serum proteome in the distribution of metastases between the organs. Metastases to the brain had a number of statistically weak correlations to Dysregulated secretome proteins, including negative associations with LGALS3BP (R_s = −0.29, p ≤ 0.05), COLEC10 (R_s = −0.338, p ≤ 0.01) and BST1 (R_s = −0.342, p ≤ 0.01), while only CST3 showed a weak positive correlation with brain metastases (R_s = 0.284, p ≤ 0.05).

Metastasis to lung showed weak positive correlations with NCAM1 (R_s = 0.279, p ≤ 0.05) and HPR (R_s = 0.254, p ≤ 0.05), as well as a negative correlation with CDH5 (R_s = −0.276, p ≤ 0.05).

Liver, bone and ovary metastasis sites each had one weak negative correlation to different cancer secretome proteins (p ≤ 0.05). No statistically significant correlations were detected between the serum proteins and skin metastasis.

Correlation analysis shows statistically significant associations between KP enzyme activity and dysregulated serum proteins

The activity of KYNU A (as inferred by AA/KYN ratio) had the greatest number of statistically significant correlations to up- or down-regulated serum proteins in patients with metastatic BrCa (Fig. 4). The strength of these correlations was weak, Notably, KYNU A had the strongest positive correlation with VASN (R_s = 0.354, p ≤ 0.01) and the strongest negative correlation with BST1 and CA2 (R_s = −0.398, p ≤ 0.01).

KMO had two statistically significant associations with serum proteins, including a positive correlation with HPR (R_s = 0.331, p ≤ 0.01) and a negative correlation with LGALS3BP (R_s= −0.262, p ≤ 0.05).

QUIN also had two statistically significant associations with dysregulated serum proteins, including a negative correlation with HPR (R_s = −0.249, p ≤ 0.05) and a positive correlation with CDH5 (R_s = 0.261, p ≤ 0.05). IDO1/TDO and KYNA both had statistically significant positive correlations with CST3 (R_s = 0.320, p ≤ 0.01; R_s = 0.275, p ≤ 0.05 respectively).

We observed no statistically significant correlations between KYNU B or PIC and the serum proteins that we had identified as being up- or down-regulated in patients with BrCa metastasis.

Publicly available datasets analysis revealed similar trends to experimental outcome

To assess the potential clinical significance of the KP enzymes (IDO1, TDO2, KMO, KYNU, Kynurenine aminotransferase (KYAT) 1, aminoadipate aminotransferase (AADAT, other aliases include KAT2, KATII, KYAT2) and KYAT3) and serum proteins (SATB2, NCAM1, CST3, PLTP, GP1BA, VASN, HGFAC, APOC4, HPR, LGALS3BP, CDH5, CA2, COLEC10 and BST1) that we found to be differentially regulated in the studied groups and associated with the BrCa progression, we accessed publically available matching datasets (Human Protein Altas and TCGA).

For the assessment of prognostic significance, we evaluated the aforementioned genes/proteins using the data from the Human Protein Atlas (https://www.proteinatlas.org, accessed on 28th August 2024). Among the variables, only KYAT3 showed prognostic value in BrCa and was favourable for survival (Supplementary Table 5). KYAT 1, AADAT (KAT2, KATII, KYAT2) and KYAT3 are genes/enzymes involved in the conversion of KYN to KYNA. However, we did not present the enzyme ratio in the above section as the ratio will not accurately reflect the activity of a specific KAT enzyme.

Next, we performed an analysis of the orthogonal data from TCGA (https://www.cancer.gov/tcga, accessed on 28th August 2024) using the UALCAN analytical software tools [41, 42] to examine how the gene products identified in our study are linked to BrCa progression. The selected BrCa cohort included 1097 patients, while there were 114 healthy controls. The information on metastatic status was available only for the number of involved lymphatic nodes.

The overall gene expression heatmap is shown in Supplementary Fig. 2 and the individual gene expression analyses are provided in Supplementary Fig. 3. Although the expression of KYAT3 is shown to have prognostic value in BrCa, the gene expression information is not available yet in the TCGA database. As indicated by the data provided in Supplementary Fig. 3, our findings corroborate with the results reported in the TCGA.

Discussion

This study presents the first detailed analysis of the involvement of the KP in BrCa development and its metastatic progression. Importantly, to provide additional mechanistic insights, our KP study was performed in parallel with the examination of the serum proteomic profiles of BrCa patients with proven local and distant metastases.

First, we confirmed that IDO1/TDO activity was downregulated as BrCa progressed from a primary tumour to metastatic dissemination (Fig. 1a). These findings were consistent with prior research that detected lower IDO1 expression in patients with nodal and distant metastasis [43]. Our findings also agree with a study by Sakurai et al., who showed lower IDO1 activity in BrCa patients with metastasis as compared to those without metastasis [16]. The association between the downregulation of IDO1/TDO activity and the spread of BrCa can provide a potential explanation for the limited success in clinical trials examining the inclusion of IDO1 inhibitors in combination treatment. There were six clinical trials examining a range of IDO1 inhibitors. These include epacadostat (phase I/II, NCT02178722 [44]; phase I/II, NCT03328026), indoximod (phase II, NCT01792050 [45]; phase I/II, NCT01042535 [46] and novaximod/GDC-0919 (phase I, NCT02471846 [47]; phase I, NCT02048709 [48]). Although the trials of indoximod and novaximod reported minimal toxicity dosage and were well tolerated, there was no significant difference between the combined treatment arm and single dose arm. In NCT02178722, a percentage of cancer patients included in phase I/II study evaluating toxicity and optimal dosage were triple-negative BrCa patients. While a portion of the BrCa patients had stable disease, a majority of them exhibited disease progression. Therefore, BrCa was not considered for inclusion in the later phase III trial. It is important to note that the participants from the above trials were not screened for IDO-1 expression in their tumours prior to their enrolment and the outcomes of these trials have certainly dampened the potential of the KP as an immunotherapy target. However, there is one clinical trial, NCT03328026, currently recruiting to evaluate the combination treatment of SV-BR-1-GM, INCMGA00012 and epacadostat in metastatic or locally recurrent BrCa patients. Despite these findings, other research has also indicated that the expression of IDO1 has been positively associated with BrCa progression and metastasis, which contradicts our findings [18, 49, 50]. This could be due to a difference in our cohort, as most of our participants with local and distant metastasis were hormone receptor positive. There are also a number of factors including BrCa subtype, age and other co-morbidities that can influence the expression of IDO1 [51, 52]. Additionally, IDO1 expression was detected in 70% of samples that tested positive for PD-L1 expression. Tumours that were negative for PD-L1 also did not express IDO1. This study suggests that IDO1 expression may be dependent on PD-L1 expression or that IDO1 may be a compensatory mechanism for PD-L1 mediated immune evasion. Findings from this study indicate the potential benefits of a combination treatment of anti-PD-L1 and IDO inhibitor in treating BrCa patients who test positive for PD-L1 [43].

In contrast to IDO-1 activity, our study showed that KMO (as inferred by 3HK/KYN ratio) was highly elevated in cancer cohorts compared to healthy controls and its activity was associated with the spread of cancer. As KMO is the principal enzyme of the KP facilitating the production of the immunosuppressive metabolite 3HK, this finding may suggest that KMO may be involved in the malignant transformation of mammary tissue by contribution to immune tolerance. 3HK has been shown to induce apoptosis in CD4+ T cells; this enables cancer cells to evade immune surveillance allowing spread and/or proliferation at the secondary site(s) [27, 28]. A study by Huang et al. (using publicly available databases) showed that KMO was elevated in BrCa patients and was associated with worse survival among the BrCa patients. Triple-negative BrCa samples showed the highest gene amplification in KMO, when compared to adjacent normal mammary cells, while KMO activity promoted transcription of genes contributing to cancer stemness, namely β-catenin, CD44 and Nanog, for cell migration/invasion. Importantly, a knockdown study of KMO in an in vivo model showed lower lung metastasis and prolonged survival [53]. This was further supported by a later study where inhibition of KMO in MDA-MB-231, a triple-negative BrCa cell line, led to suppression of migration and invasion capability [54].

Despite the evidence of KMO involvement in human diseases, development of KMO inhibitors is still in the pre-clinical stages of development [55, 56]. Among the KMO inhibitors, CHDI-340246 has shown the most promise. A recent in vivo study has demonstrated high rates of inhibition of KMO and its product 3HK, while skewing the KP towards producing neuroprotective KYNA [57]. A later study demonstrated that it has a favourable profile of absorption, distribution, metabolism and excretion in various animal models, including primates, thus supporting its evaluation in humans [58].

In addition to examining the KP involvement in BrCa, our study also revealed 14 differentially expressed proteins: nine were significantly up-regulated and five were down-regulated in distant metastatic BrCa, compared to local metastasis. SATB2 was the most highly up-regulated protein (increased by 8.57-fold, p > 0.01). The expression of SATB2 was previously demonstrated to have a significant association with advanced tumour grade and poorer survival of BrCa patients [59]. SATB2 was also suggested as a biomarker to detect metastatic colorectal cancer in liver biopsies [60]. Our finding of other up-regulated proteins, NCAM1 [61], CST3 [62, 63], VASN [64], HGFAC [65], HGR [66] and GP1BA [67], also agreed with previous cancer studies and thus validates our MS-based proteome screening method. Our study confirmed that the expression levels of these proteins was strongly associated with distant metastatic BrCa progression and increasing tumour grade.

In addition to the above proteins, we revealed two novel findings of overexpression of PLTP and APOC4 in sera of BrCa patients with distant metastasis. These proteins are involved in lipoprotein metabolism and have not previously been linked to cancer metastasis. PLTP had previously been shown to be differentially expressed in inflammatory and non-inflammatory BrCa [68, 69]. Computer modelling has suggested that PLTP is a determining factor in the survival of BrCa patients [68]. APOC4, on the other hand, belongs to the apolipoprotein gene family, which consists of 22 members. The APOC family comprises 4 members: APOC1, APOC2, APOC3 and APOC4, which are involved in regulating the MAPK pathway. Although APOC4 has not yet been linked to cancer, studies have shown an overexpression of the other members of the APOC family in various malignant tumours [70, 71]. APOC1 was overexpressed in BrCa patients and was also used to distinguish between triple-negative BrCa and other BrCa [71]. While our results suggest that PLTP and APOC4 play a role in the metastasis of BrCa, their specific roles in cancer need further clarification.

COLEC10 has previously been reported as an upregulated protein in BrCa [72]. Our current study indicates that this protein is positively associated with the presence of local (lymph node) metastases, while it is downregulated in BrCa patients with distant metastases. COLEC10 encodes for collectin liver 1 protein, which functions as a pattern recognition molecule involved in innate immunity [73]. The study led by Chen et al. reported an increased level of COLEC10 in BrCa compared to adjacent normal tissues. Notably, more than 75% of their cancer cohort was diagnosed with stage I and II BrCa. Our data showed that COLEC10 correlated negatively with the presence of distant metastasis, this may suggest that COLEC10 has a role in early, but not in advanced stage, BrCa. Interestingly, COLEC10 was also found to be downregulated in early-stage hepatocarcinoma cancer tissues when compared to normal controls. This study indicated that the decreased COLEC10 level in the late-stage was related to a shorter overall survival [74]. More studies are required to understand the role of COLEC10 in modulating the spread of cancer.

In contrast to our results, proteins that were found to be downregulated in BrCa with distant metastases, including LGALS3BP (Galectin-3-binding protein), CDH5 (Cadherin-5) and CA2 (Carbonic anhydrase 2), were shown as upregulated proteins in other BrCa studies [75–77]. It is important to note that these studies compared metastatic BrCa to normal healthy controls, whilst our study highlighted the protein expression differences in distant and local metastatic BrCa. Finally, our most interesting finding that has not been linked to BrCa before is BST1 (ADP-ribosyl cyclase/cyclic ADP-ribose hydrolase 2). This showed the highest downregulation in distant metastatic BrCa. BST1 is a multi-functional molecule that is expressed on stromal cells such as endothelial cells [78] and immune cells, such as the adaptive and innate immune cells [79, 80]. It has been shown to play a key role in the growth of pre-B cell [79] and in facilitating immune cell recruitment to the site of inflammation [81]. Downregulation of this protein may result in an impaired immune response towards the tumour.

It is crucial to note that although our results showed correlations between some KP enzymes and proteins, these correlations could not to be interpreted as direct causal relationships [82]. Our most significant finding was the correlation KYNU A had with 9 different serum proteins dysregulated in the distant metastases cohort. These findings are novel, as no prior research has linked KYNU A with these proteins. While these correlations were statistically significant, the majority of them were weak. This indicates that although KYNU A and these proteins may have impacts on one another, there are other mediators of such associations as well. Similarly, KMO, IDO/TDO, KYNA and QUIN all had weak correlations with the identified serum proteins of interest, with no prior studies highlighting any associations between the identified proteins and the KP enzymes [83]. Further study with a larger cohort will be required to better understand the associations between KP and proteins and the progression of the disease.

Limitation

Our study has limitations. Firstly, the relatively small sample size of each cohort may impact the robustness of our findings. While our observed associations may still provide valuable insights to the role of the KP in breast cancer metastasis, future research with larger cohorts will be required to validate these findings. Next, the stage I BrCa cohort in our study included only the luminal BrCa subtype while the metastases BrCa cohorts contained a mixed population of the major BrCa subtypes. As KP activity and tumour-immune response is highly dysregulated in non-hormonal BrCa subtypes [84], this may be a confounding factor in the comparison between cancer stages. The limited volume of stage I BrCa sera restricted our ability to compare the protein profiles between the cancer cohorts and to examine their relationship with the KP metabolites. Finally, the measurement of KP activity in sera is a reflection of a systemic response but not the direct evaluation of the metabolites in the tumour samples. Analysis of tumour KP activity and its mRNA expression in relation to tumour immune profile could strengthen the notion that KMO plays a significant role in the spread of BrCa.

Conclusion

We have demonstrated for the first time that the activity of KMO is highly elevated in late-stage BrCa as inferred by the high concentration of 3HK. Given the immunosuppressive property of 3HK, this suggests that these tumours may be suppressing immune surveillance to local or distant cancer events. Our data also identified 14 differentially expressed proteins in the sera of patients with local versus distant metastasis. Three of them have not previously been linked to cancer, suggesting that their expressions may have essential roles in metastatic BrCa progression. However, our correlation analysis showed a weak association between these proteins and changes in KP activity. This suggests that these associations could be a consequence of another factor instead of a causative link.

Author contributions

Conceptualisation, HNG, GJG, LG, AG and BH; writing—original draft preparation, review and editing, HNG, GJG, LG, SBA, AG, ABB and BH; experimental work and data analysis—HNG, SBA, SK, LC, AG, ABB and BH.

Funding

HNG is supported by a Macquarie University Research Excellence Scholarship—Master of Research scholarship; SBA is supported by Cancer Council NSW funding RG23-06; GJG is supported by a Fellowship from the National Health and Medical Research Council (NHMRC) APP1176660, and Macquarie University; AG was supported by the Macquarie University Research Fellowship.

Data availability

All relevant data related to this study are included within the article or in the supplementary materials. Further data will be provided upon reasonable request to the corresponding author.

Competing interests

The authors declare no competing interests.

Ethics approval

The study was approved by the Macquarie University Ethics Committee (Ref: 5201600401). All analyses were conducted in accordance with the Declaration of Helsinki.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-024-02889-z.