ATF4 as a Prognostic Marker and Modulator of Glutamine Metabolism in Oestrogen Receptor-Positive Breast Cancer Open Access

Metabolic reprogramming is a well-documented “cancer hallmark” that allows cancer cells to produce energy, proliferate rapidly, metastasize, and survive in harsh tumour microenvironments [1]. Glutamine is the second primary metabolite, after glucose, to support cancer cell proliferation [2]. The importance of glutamine is highlighted throughout its numerous functions, namely, facilitating macromolecule synthesis of nucleotides, lipids, and proteins [3] and supporting redox balance [4]. Furthermore, the metabolism of glutamine via glutaminolysis aids the replenishment of intermediates within the tricarboxylic acid cycle [5].

Numerous studies support the role of AATs in breast cancer (BC). Solute carriers (SLC), SLC1A5, SLC3A2, and SLC7A5, which have high affinity to glutamine, are associated with the aggressive nature of ER-positive BC [6‒9]. However, the regulation of AAT expression associated with glutamine transport within BC has yet to be explored. In this respect, there is a need to further explore the specific mechanisms behind transporter expression and consequent effects in BC. Activating transcription factor 4 (ATF4/CREB-2) has previously been implicated in the control of AAT within autophagy-deficit tumour cells [10].

ATF4 is a stress-responsive gene, belonging to the ATF/cyclic adenosine monophosphate response element binding protein (ATP/CREB) family [11]. ATF4 gene, also known as cyclic AMP-responsive element-binding protein 2 (CREB-2), is located on 22q13.1, a region frequently associated with loss of heterozygosity in BC [12]. ATF4 protein is a DNA binding transcription factor and also involved in protein-protein interactions. It consists of DNA binding, dimerization, and C-terminal regulatory domains. ATF4 forms part of the integrated stress response (ISR) pathway, underlying the downregulation of protein synthesis during cellular stress and amino acid starvation [13]. Previous studies by Ye et al. [14] revealed upregulation of the ATF4 ISR pathway within tumours, implicating ISR and potential ATF4 necessity in cancer cell adaptation to tumour microenvironment.

There is increasing evidence that ATF4 upregulation is vital to long-term cell survival through promoting the expression of adaptive genes. ATF4 regulates the expression of genes associated with metabolism, oxidative stress, protein synthesis, and AAT [15]. Furthermore, ATF4 is linked to angiogenesis [16] and metastasis [17, 18]. In this respect, ATF4 expression is shown to be advantageous to cancer cells, enabling cell proliferation and preservation despite cellular stresses caused by this heightened activity.

Whilst ATF4 is associated with a poor patient survival in triple-negative (TN) BC [19], the importance of ATF4 as a prognostic biomarker as well as its specific role on AATs in ER+ BC remains undetermined. Therefore, it is hypothesized that high ATF4 expression regulates glutamine-associated AATs in aggressive BC subtypes. This study aims to assess ATF4 expression and its prognostic value within BC, in association with glutamine-associated AAT expression.

The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset was used to generate data on genomic and transcriptomic profiling in n = 1,980 BC cases using the Affymetrix SNP 6.0 and Illumina HT-12 v3 platforms. The association between ATF4 gene copy number variations (CNVs) and mRNA expression was investigated. ATF4 mRNA expression was dichotomized into high- and low-expression groups using a median cut-off log2 intensity value of 7.53, and subsequent associations between expression groups and various clinicopathological parameters, molecular BC subtypes, and patient outcome were subsequently evaluated. The online BC molecular datasets, BC Gene Expression Miner v4.4 (n = 4,712) (http://bcgenex.centregauducheau.fr) [20], and the KM plotter (http://kmplot.com) [21] were used as an external validation datasets. ATF4 gene expression was correlated with gene expression of AAT showing high affinity for glutamine: SLC1A5, SLC3A2, SLC3A2, SLC6A14, SLC6A15, SLC6A19, SLC7A5, SLC7A6, SLC7A8, SLC7A8, SLC7A9, SLC7A11, SLC28A1, SLC28A2, SLC28A3, SLC28A5, SLC28A7, and SLC28A8.

A cohort of n = 1,341 patients younger than 70 years, with early-stage operable BC, were enrolled into the Nottingham Primary Breast Carcinoma Series and presented to Nottingham City Hospital, UK between 1986 and 2006. Patients were managed in accordance with a uniform protocol. Survival data were maintained on a prospective basis, which included breast cancer-specific survival (BCSS) defined as the time in months from primary surgery to the time of BC-related death. Full patient characteristics of the cohort are summarized in Supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000539564). Protein expression for ER, PR, HER2, and AATs with high affinity for glutamine: SLC1A5, SLC3A2, SLC7A5, SLC7A8, SLC7A11, and SLC38A2 was previously determined [9, 22, 23]. BC molecular subtypes were defined, based on the tumour IHC profile and the Elston-Ellis mitotic score as ER+/HER2− low proliferation (mitotic score 1), ER+/HER2− high proliferation (mitotic scores 2 and 3), HER2-positive class: HER2+ regardless of ER status, TN: ER−, PR−, and HER2−.

Western blot was used prior to immunohistochemistry to ensure specificity of anti-ATF4 antibody (EPR18111, Ab184909, Abcam, UK) in ZR-75-1 (BC) and HeLa cell lysates (cervical adenocarcinoma) (American Type Culture Collection; Rockville, MD, USA). The primary antibody was diluted to 1:500, whilst the secondary antibody (IRDye 800CW Donkey Anti-Rabbit, LI-COR Biosciences, UK) was diluted to 1:15,000. Blocking solution, to prevent non-specific staining, was 5% milk (Marvel Original Dried Skimmed Milk, Premier Food Group Ltd., UK) in PBS-Tween 20 (0.1%). Mouse monoclonal anti-beta-actin antibody (Sigma-Aldrich, A5441) was used as a loading control. The A-Fluorescence detection method, using wavelengths at 800 nm using Odyssey Fc Imaging Studio 4.0 (LI-COR Biosciences, UK), was used to detect ATF4. A single specific band at 50 kDa was visualized at the correct predicted size for ATF4 (online suppl. Fig. 1).

Immunohistochemistry was performed on 4 μm full-face invasive BC tissue sections (n = 21), to determine pattern of tissue staining and tissue microarray (TMA) sections incorporating 0.6 mm cores of invasive BC tissue (n = 2,225) as previously described [24]. The Novolink Max Polymer Detection System (RE7280-K, Leica Biosystems, UK) was used to stain protein expression according to the manufacturer instructions. Heat-induced epitope retrieval was performed using citrate buffer (pH 6.0) for 20 min in a microwave (Whirlpool JT359 Jet Chef 1000 W). Tissues were incubated with ATF4 primary antibody, diluted at 1:300 overnight at 4°C. High-resolution digital images of TMA slides were scanned using a NanoZoomer (Hamamatsu Photonics, Welwyn Garden City, UK) and visualized using Xplore (Philips, UK) at a magnification of ×20.

Only TMA cores containing more than 15% invasive tumour tissue were assessed. To evaluate the extent of ATF4 expression, the semi-quantitative modified H-score was used, assessing both the intensity and percentage of nuclear staining. The intensity of staining was measured on a scale of 0–3, 0 indicating no staining, 1 indicating weak staining, 2 indicating moderate staining, and 3 indicating strong staining. The percentage of nuclear staining within the cores was then subjectively evaluated. The final H-score was calculated by multiplying the percentage of positive cells (0–100) by the intensity (0–3), producing a total range of 0–300.

An independent scorer (RE), blind to clinical data and scores, scored 10% of cases. An inter-scorer correlation was calculated using Pearson’s 2-tailed correlation coefficient producing a coefficient of r = 0.8 with p = 3.6 × 10⁻³⁹, thus suggesting high inter-scorer reliability. ATF4 protein expression was dichotomized into high- and low-expression groups using the median H-score value of 130.

To determine associations between ATF4 expression and various ATTs, statistical analysis was performed using the SPSS v26.0 Statistical Software (IBM SPSS Statistics, SPSS Inc., Chicago, IL, USA). ATF4 CNV/mRNA and ATF4 protein correlation within continuous variables was calculated using Pearson’s correlation coefficient. Meanwhile, categorical data were evaluated through the use of Pearson’s χ² analysis. Wherein multiple statistical analyses were run, p values were adjusted using the Bonferroni-Holm correction to account for multiple testing. Significant differences within multiple continuous variables were evaluated using one-way analysis of variance (ANOVA). Survival curves were evaluated using Kaplan-Meier and log-rank testing, in relation to BCSS. Furthermore, Cox regression was used in multivariate analyses to identify independent prognostic factors. Within all analyses, p values <0.05 were considered statistically significant. A summary of datasets/experimental cohort together with analyses is shown in Figure 1.

A total of 2.7% (54/1,980) BC showed gain of ATF4 gene CN, whilst 11.8% (234/1,980) showed ATF4 CN loss. ATF4 gene copy number gain and mRNA expression were strongly correlated in BC, with higher mRNA levels associated with gene CN gain (p < 0.001, Fig. 2).

ATF4 immunoreactivity in full-face breast tissue sections showed nuclear, homogeneous staining of invasive BC cells, indicating that TMA cores are representative for the whole tumour. Within TMA cores, there were variable levels of staining intensities in invasive BC cells, ranging from absent to strong (Fig. 3). Due to different cohorts, it was not possible to correlate ATF4 protein expression with either ATF4 CNV or mRNA.

ATF4 CN gain was significantly associated with high-grade tumours, while ATF4 CN loss was linked to low-grade tumours (p < 0.001; Table 1). High ATF4 mRNA expression was associated with younger age at diagnosis, high tumour grade, and the poor NPI prognostic group (all p < 0.001; Table 1). These associations were confirmed using Breast Cancer Gene-Expression Miner (p < 0.05; online suppl. Fig. 2). ATF4 protein expression however was not significantly associated with any of the clinicopathological parameters (Table 1).

At the genomic level, findings revealed a striking link between ATF4 gene alterations and BC molecular subtypes. ATF4 CN gain was predominant in ER−, PR−, and TN tumours, while ATF4 CN loss was prevalent in ER+ and PR+ BC (all p < 0.001, Table 2). These findings were further corroborated by significantly higher ATF4 mRNA expression in ER− and PR− BC in the METABRIC (both p < 0.001, Table 2) and Breast Cancer Gene Expression Miner (p < 0.05; online suppl. Fig. 2) datasets. High ATF4 mRNA expression was observed in the basal-like immunosuppressed compared with basal-like immune-activated, luminal androgen receptor, and mesenchymal subtypes (p < 0.001; online suppl. Fig. 2H).

High ATF4 protein expression was significantly more prevalent in TN BC, while low ATF4 protein expression was characteristic of ER+ and PR+ tumours (p = 0.0002; Table 2). Within PAM50 subtypes, ATF4 CN gain was predominantly observed in basal tumours, while ATF4 CN loss was prevalent in luminal tumours (p < 0.001; Table 2). This pattern was mirrored by high ATF4 mRNA expression in basal tumours and low expression in luminal tumours (p < 0.001; Table 2). These findings were further validated using Breast Cancer Gene Expression Miner (p < 0.05; online suppl. Fig. 2). ATF4 CN gain and high mRNA was associated with Cluster 10 (TN), whereas ATF4 CN loss and low mRNA expression were associated with METABRIC Integrative Cluster 8 (Luminal A) (p < 0.001; Table 2).

A remarkably strong correlation was observed between ATF4 and SLC3A2 mRNA in the METABRIC dataset (r = 0.81, p = 0.0003; Table 3), while moderate positive correlations were found between ATF4 and SLC1A5, SLC3A2, SLC6A15, SLC7A5, SLC7A7, SLC7A11, SLC38A2, and SLC38A8 in both METABRIC and Breast Cancer Gene Expression Miner datasets (all p ≤ 0.012; Table 3). Intriguingly, weak negative correlations were observed between ATF4 and SLC7A8 mRNA in both datasets (r > −0.23, p < 0.0001; Table 3). ATF4 and SLC1A5 and SLC7A11 mRNA were correlated similarly in all TNBC subtypes (p < 0.0001, online suppl. Fig. 2I). ATF4 protein expression exhibited a positive relationship with SLC1A5 and SLC7A11 protein (p = 0.04, p = 9.8 × 10⁻¹¹ respectively; Table 4), while no significant association was observed with SLC3A2, SLC7A5, SLC7A8, or SLC38A2.

Contrary to expectations, neither ATF4 CNV nor ATF4 mRNA expression was associated with patient overall survival in the entire cohort (Fig. 4) or within specific biological subtypes (data not shown). This lack of association is further corroborated by independent analyses using Breast Cancer Gene Expression Miner and Kaplan Meier Plotter datasets (online suppl. Fig. 3). In addition, there was no significant difference in those tumours showing ATF4 CN loss and low ATF4 mRNA (p = 0.465). Similarly, ATF4 protein expression showed no correlation with BCSS in the overall cohort (Fig. 4) or in biological subtypes (data not shown).

The interplay between ATF4 and glutamine metabolism-related AATs was further investigated. There was no association between ATF4 CNV or ATF4 mRNA expression and any of the AATs with patient outcome (data not shown). However, ATF4 protein together with SLC1A5 or SLC7A11 revealed a differential impact on patient outcome. Co-expression of ATF4 and SLC1A5 protein was significantly associated with poor BCSS in ER+ tumours only (p < 0.001, Fig. 5a–c). Conversely, high ATF4 and high SLC7A11 expression conferred a better survival in the whole cohort (p = 0.003; Fig. 5d) irrespective of ER status (data not shown). Multivariate analysis revealed that ATF4 alone or in combination with SLC1A5 was not an independent prognostic factor for BC, beyond tumour grade, size, and nodal stage (data not shown).

ATF4 has been identified in BC subtypes, but its relevance as a prognostic marker is not well understood. Amino acids are essential for cell survival, especially in tumour cells, which have high proliferation rates and increased amino acid demand [25]. AATs are therefore vital for nutrient supply to cancer cells. Additionally, glutamine metabolism is associated with cancer cell metabolic reprogramming and is closely linked to AATs [4]. In this study, we aimed to assess ATF4 expression, its prognostic value, and potential association with glutamine-associated AATs in large cohorts of BC patients.

Overall, we confirm that ATF4 CN gain, high ATF4 mRNA, and ATF4 protein are associated with aggressive BC. This is consistent with the previous literature where Fan et al. showed high ATF4 expression in metastatic BC, whilst González-González et al. associated ATF4 with increased aggressiveness in TN BC [19, 26]. In some cases, where there is no CNV, increased ATF4 expression in cancer cells could be due to activation of the ISR, which is necessary for cancer cell survival and proliferation, especially in aggressive tumours [15, 27]. ATF4 has also been shown to play a protective role in maintaining normal, healthy cell development. For example, ATF4 is a critical regulator of osteoblast differentiation and plasma cell viability [28, 29]. ATF4 overexpression also decreases proliferation and accelerates mammary gland involution in transgenic mice during pregnancy and lactation [30]. These studies demonstrate that ATF4 is essential for the differentiation and survival of rapidly proliferating cells, as well as for normal breast development.

ATF4 expression is associated with several AATs at the transcriptomic level, but only SLC1A5 and SLC7A11 at the protein level in this cohort. This suggests that the relationship between ATF4 and AATs is complex and dependent on the transporter in question. The influence of ATF4 on AATs may reflect its dual role in promoting both survival and apoptosis [31]. For example, high expression of SLC1A5, SLC3A2, and SLC7A5 is associated with poor prognosis in highly proliferative ER+ BC [9]. This is likely due to their role in regulating intracellular amino acid concentrations. ATF4 mRNA expression is positively correlated with these transporters, reflecting its pro-survival role. However, ATF4 is also strongly associated with SLC7A11, which exchanges intracellular glutamate for extracellular cystine [32]. This suggests that ATF4 may also promote cancer cell sensitivity to glucose starvation by decreasing intracellular glutamate. While this has not yet been documented in BC, it suggests that ATF4 may have pro-apoptotic effects in this disease.

Moreover, high ATF4 with low expression of SLC1A5 conferred longer survival rates within in ER+ BC. Whilst it could be suggested that the low expression of SLC1A5 is responsible for this result alone, both this study and numerous others have implicated elevated ATF4 expression with that of elevated SLC1A5 and consequent cancer cell survival [33]. Perhaps one explanation for this may be through post-translational modifications. Phosphorylation at various threonine residues has been shown by Bagheri-Yarmand et al. [34] to reduce ATF4 activity at the promoters of pro-apoptotic targets NOXA and PUMA. Whilst it is unclear if this interaction is specific to pro-apoptotic targets, it demonstrates the complex mechanisms by which ATF4 activity is modulated.

These findings suggest a complex interplay between ATF4 and AATs in BC biology and underscore the potential role for ATF4 as a prognostic marker in ER+ BC, offering a unique opportunity for risk stratification and personalized treatment strategies. Future investigations are needed to confirm this.

We thank the Nottingham Health Science Biobank and Breast Cancer Now Tissue Bank for the provision of tissue samples.

This study was approved by the Nottingham Research Ethics Committee 2 under the title “Development of a molecular genetic classification of breast cancer” and the North West – Greater Manchester Central Research Ethics Committee under the title “Nottingham Health Science Biobank (NHSB)” (reference number 15/NW/0685). This study was performed according to the REMARK guidelines for tumour prognostic studies [35]. Written informed consent was obtained from all individuals and all samples were anonymized.

The authors declare no conflict of interests.

No funding was received for this study.

R.P., L.H.A., and R.E.: writing, review, editing, methodology, formal analysis, and interpretation. B.K.M., B.E., and A.F.: methodology and writing – original draft, reviewing, and editing. I.O.E. and E.A.R.: supervision and writing – review and editing. A.R.G: conceptualization, formal analysis and interpretation, supervision, and writing – review and editing.

The data that support the findings of this study are not publicly available due to ethical reasons but are available from the corresponding author (A.R.G.) upon reasonable request.