Increased plasma lipids in triple-negative breast cancer and impairment in HDL functionality in advanced stages of tumors

The association between plasma lipids and breast cancer (BC) has been extensively explored although results are still conflicting especially regarding the relationship with high-density lipoprotein cholesterol (HDLc) levels. HDL mediates cholesterol and oxysterol removal from cells limiting sterols necessary for tumor growth, inflammation, and metastasis and this may not be reflected by measuring HDLc. We addressed recently diagnosed, treatment-naïve BC women (n = 163), classified according to molecular types of tumors and clinical stages of the disease, in comparison to control women (CTR; n = 150) regarding plasma lipids and lipoproteins, HDL functionality and composition in lipids, oxysterols, and apo A-I. HDL was isolated by plasma discontinuous density gradient ultracentrifugation. Lipids (total cholesterol, TC; triglycerides, TG; and phospholipids, PL) were determined by enzymatic assays, apo A-I by immunoturbidimetry, and oxysterols (27, 25, and 24-hydroxycholesterol), by gas chromatography coupled with mass spectrometry. HDL-mediated cell cholesterol removal was determined in macrophages previously overloaded with cholesterol and 14C-cholesterol. Lipid profile was similar between CTR and BC groups after adjustment per age. In the BC group, lower concentrations of TC (84%), TG (93%), PL (89%), and 27-hydroxicholesterol (61%) were observed in HDL, although the lipoprotein ability in removing cell cholesterol was similar to HDL from CRT. Triple-negative (TN) BC cases presented higher levels of TC, TG, apoB, and non-HDLc when compared to other molecular types. Impaired HDL functionality was observed in more advanced BC cases (stages III and IV), as cholesterol efflux was around 28% lower as compared to stages I and II. The altered lipid profile in TN cases may contribute to channeling lipids to tumor development in a hystotype with a more aggressive clinical history. Moreover, findings reinforce the dissociation between plasma levels of HDLc and HDL functionality in determining BC outcomes.


Background
Female breast cancer (BC) is the most commonly diagnosed cancer worldwide (11.7%) and is the fth leading cause of cancer death (6.9%) [1]. Being considered a heterogeneous disease, the molecular classi cation of breast tumors is widely utilized as a tool for therapeutic choice and prognosis. It is based on the expression of estrogen and progesterone receptors [(luminal A; LA); (luminal B; LB)], human epidermal growth factor receptor 2 (HER2), or the absence of them (triple-negative; TN) [2,3].
Over the last decades, evidence has emerged linking plasma lipid levels to development and worse BC outcomes [4][5][6][7]. This relates to tumor cell reprogramming enabling more lipids for cell division and metastasis [8]. Nonetheless, clinical and epidemiological data demonstrated either positive, negative, or no impact of hyperlipidemia on BC incidence [9][10][11][12][13]. In the same manner, the association between statins and the risk of BC is controversial [14,15].
Particularly, much attention has been paid to the role of cholesterol in the high-density lipoproteins (HDLc), with most of the results pointing to a protective role of plasma HDLc in BC risk prevention [4,10,11]. These results may relate to the bene cial HDL activities in removing excess cholesterol and oxysterols -necessary for replication and metastasis -from tumor cells [16]. Besides, HDL has antioxidant and anti-in ammatory activities and acts as a cargo lipoprotein for bioactive lipids, proteins, and microRNAs that may modulate tumor growth and evolution [17,18]. Nonetheless, there is still controversy and some studies only show a weak association between HDLc and BC, and others, a positive relation or no association at all [9,12,13,19].
In resemblance to HDL's role in atherosclerosis and other chronic non-communicable diseases, it is possible that the HDLc metric is not the best choice as a predictor variable for determining BC incidence and evolution [20]. In this sense, alterations in HDL functionality are considered, especially by taking into account the modulation of this lipoprotein composition and function by the tumor microenvironment [21].
HDL is well recognized for its ability in removing cell cholesterol allowing its uptake by the liver, secretion into bile, and excretion in feces, namely by the reverse cholesterol transport. Together with other HDL actions, reverse cholesterol transport allows the mechanistic association of HDL with the prevention of atherosclerosis. In analogy, in cell tumors, reverse cholesterol transport can be considered as a defensive mechanism that prevents cholesterol accumulation that supplies cell proliferation and metastasis. In this regard, the expression of HDL receptors, such as ABCA-1 and SR-BI are changed in tumors and related to epithelial-mesenchymal transition, tumor growth, and spreading [22][23][24][25].
HDL also mediates the transport of oxysterols produced intracellularly by the oxidation of the cholesterol molecule. Considering its higher hydrophobicity when compared to other oxysterols and cholesterol, 27HC diffuses across the plasma membrane [26]. Besides, its transference to HDL is facilitated by the ATP binding cassette transporters A-1 and G-1 (respectively, ABCA-1 and ABCG-1) located in the cell membrane of tumor cells and macrophages that in ltrate the tumor area [27]. Then, its ux outside cells is considered an additional route for reverse cholesterol transport, limiting sterol content that relates to cell in ammation, oxidative stress, and proliferation [18,28]. On the other side, as a selective estrogen receptor modulator (SERM), 27HC potentiates tumor growth and metastasis in BC. In human BC samples, the expression of the enzymes involved in the production of 27HC (CYP27A1) and metabolism (CYP7B1) are, respectively, increased and diminished and associated with a poor tumor prognosis [29]. In addition, tumor metastasis is related to 27HC-dependent activation of the liver X receptor (LXR) [29][30][31][32].
HDL removes cholesterol and oxysterols from cells, making acceptable the idea that it limits the intracellular accumulation of sterols and their negative impact on BC. In this way, it was addressed in newly diagnosed women with BC categorized according to the clinical stage of the disease and molecular classi cation of the tumor, without pharmacological or surgical intervention, compared to control women (CTR): 1) plasma lipid pro le [cholesterol total (TC), HDLc, apolipoprotein B (apoB) and triglycerides (TG)]; 2) the composition of isolated HDL in lipids, main species of oxysterols and apolipoprotein A-I (apoA-I); and 3) the ability of HDL to mediate the removal of cholesterol from macrophages. Plasma lipids were similar between CTR and all women with BC, but when categorized according to tumor molecular type, triple-negative (TN) BC cases had higher plasma TC, TG, and apoB values compared to other molecular types. Despite having lower levels of CT, phospholipids (PL), and 27HC, HDL from women with BC maintained its ability to remove cellular cholesterol when compared to HDL from CTR cases. In advanced BC (covering clinical stages III and IV), despite similar composition in apoA-I and lipids, HDL had an impaired ability to remove cholesterol from macrophages.

Methods
Two hundred and one women newly diagnosed with BC between 18 and 80 years old in any clinical stage and with the molecular classi cation of the tumor were recruited at Hospital Pérola Byington. A hundred and fty-seven women without any type of cancer were recruited at Universidade de São Paulo and at Unidade Básica de Saúde Dra. Ilza Weltman Hutzler (control group; CTR). Exclusion criteria were diabetes mellitus, chronic kidney disease (estimated glomerular ltration rate < 60mL/min/1.73m2), autoimmune diseases, smokers, alcoholics, use of contraceptives, and hormone replacement therapy, pregnancy, previous history of any cancer, and in situ breast disease. After excluding those who did not meet the eligibility criteria, 150 CTR and 163 women with BC remained in the study. All participants were informed about the study and signed an informed written consent previously approved by institutional Ethics Committees, in accordance with the Declaration of Helsinki, including approval for publication ( The molecular classi cation of tumors was obtained from medical records being accessed immunohistochemically after percutaneous biopsy or surgery, according to the American College of Pathologists [33]. Samples that were positive for hormone receptors (estrogen and progesterone) were those in which > 1% of the tumor cells showed positive nuclear staining of moderate to strong intensity on immunohistochemistry and were classi ed as Luminal A and B neoplasia if they presented the Ki67 index below or above 14%, respectively. Samples that exhibited > 10% of invasive tumor cells with strong staining in the plasma membrane were considered HER2 positive. In case of moderate staining in > 10% of the cells or strong in < 10% of the cells, the sample was re-evaluated by in situ hybridization and was considered positive if a HER2/centromere ratio > 2.0; or if a HER2/centromere ratio < 2.0 with mean HER2 > 6 signals per cell (greater than 120 signals in 20 nuclei). Tumor samples that did not express either hormonal or HER2 receptors were classi ed as triple-negative (TN) breast cancers [34].
The current investigation followed the STROBE Statement. The checklist is in the Supporting Information (S1 le). For sample calculation, the number of new cases of BC included for treatment at Hospital Pérola Byington during 2018 (2,985 cases) was taken into account; the study design, with the comparison of outcome variables between two main groups (CTR vs. BC); the main outcome variables; the effect size of the variables according to the main studies published in the area; and the probability of committing a type 1 error (0.05) and type 2 error (0.20), with 80% power, resulted in 144 patients in each group (pairing 1:1).
Isolation of plasma lipoproteins HDL (D = 1.063-1.21 g/mL) was isolated from BC and control women´s plasma by discontinuous density ultracentrifugation and immediately frozen at -80°C in a 5% saccharose solution. HDL composition in lipids (TC, TG, and PL) was determined by enzymatic techniques. ApoA-I was determined by the immunoturbidimetric method (Randox Lab. Ltd. Crumlin, UK).
Low-density lipoprotein isolation and acetylation LDL (D = 1.019-1.063 g/mL) was obtained by sequential ultracentrifugation of plasma from healthy volunteers and was puri ed by discontinuous density ultracentrifugation. After protein quanti cation by the Lowry technique [36], LDL was incubated with acetic anhydride as previously described [37]. Acetylated LDL was dialyzed and utilized to load macrophages with cholesterol.

HDL-mediated cholesterol e ux from bone marrow-derived macrophages
The Institutional Animal Care and Research Advisory Committee (Universidade Nove de Julho # 7070120821, 08/23/2021) approved the study according to the U.S. National Institutes of Health Guide for the Care and Use of Laboratory Animals. C57BL/6 J mice were housed with free access to commercial chow (Nuvilab CR1, São Paulo, Brazil) and drinking water in a conventional animal facility at 22 ± 2°C under a 12h light/dark cycle. Undifferentiated bone marrows cells were obtained from C57BL/6 wild-type mouse´s tibia and femora, as previously described [38]. Cells were differentiated into macrophages by incubating with L929 cells-conditioned medium (ATCC, American Tissue Culture Collection) and plated in culture dishes for 5 days at 37°C, under 5% (v/v) CO 2 . The medium was changed for a new one and after 6 days replaced by DMEM (low glucose, containing 1% penicillin/streptomycin and 10% heat-inactivated fetal calf serum). Macrophages were overloaded with acetylated LDL (50 µg/mL) and 14 C-cholesterol (0.3 µCi/mL) for 48 h, and after washing incubated with DMEM containing fatty acid-free albumin for equilibration of intracellular cholesterol pools. Cells were incubated for 6 h with HDL (50 µg/mL) from BC or control women; control incubations were performed in the absence of HDL (basal e ux). After incubation, the radioactivity in the medium was determined as well as in the cell lipid extract. The cholesterol e ux was determined as 14 C-cholesterol in the medium / 14 C-cholesterol in the medium + 14 C-cholesterol in cells x 100. Values from basal e ux were subtracted from those obtained after incubation with HDL, in order to express the speci c ability of HDL in removing cell cholesterol.

Oxysterols quanti cation in isolated HDL
Oxysterols (24-hydroxycholesterol, 25-hydroxycholesterol, and 27-hydroxycholesterol) were measured from 1 mL of extracted HDL as previously described [40]. Brie y, a mixture of 100 ng of oxysterol deuterium-labeled (7α-hydroxycholesterol-d7, 7β-hydroxycholesterol-d7, 25-hydroxycholesterol-d7, 27hydroxycholesterol-d7) (Avanti Polar Lipids, Alabaster, USA) was added as internal standard. Alkaline saponi cation was done by adding 10 mL of ethanolic KOH (0.4 M) solution for 2h, at room temperature with adjustment to pH 7 with phosphoric acid, followed by the addition of 20 mL of chloroform and 6 mL of water. After vigorous shaking and centrifugation at 4°C, the aqueous phase was removed and the organic phase evaporated. The lipid extract was dissolved in toluene (1 mL). Oxysterols were isolated from cholesterol by solid-phase extraction (Sigma-Aldrich Supelclean LC-Si SPE Tubes SUPELCO, Bellefonte, USA). The sample (1mL in toluene) was applied to the column previously conditioned with 2 mL of hexane, following washing with 1 mL of hexane. Sterols were eluted with 1.5% isopropanol in hexane (8mL), and oxysterols were further eluted with 30% isopropanol in hexane (6mL). The solvent was evaporated and samples were derivatized (100 µL of pyridine and 100 µL of N, O-bis (trimethylsilyl) tri uoroacetamide with trimethylchlorosilane (BSTFA; Sigma-Aldrich, St. Louis, USA), for 1 h at 60•C). One microliter of the derivatized sample (1µL) was injected into a gas chromatograph coupled to a mass spectrometer (Shimadzu GCMS-QP2010, Kyoto, Japan) by an automatic injector and analyzed in selected ion monitoring. The separation was performed on a Restek capillary column (100% dimethyl polysiloxane -RxiR − 1 ms. Cat. #13323), 30 m, internal diameter 0.25 mm, for 30 min, using helium as mobile phase, with a constant linear velocity of 44.1 cm/sec. The oven started at 240•C with an increment of 5•C/min, for 7 min up to 290•C. The mass spectrometer operated in impact electron mode at an ionization voltage of 70 eV with the temperature of the ion source at 300•C. The quanti cation was performed by comparing the peak areas of the standard curve and correcting for internal standards.

Statistical analysis
The Shapiro-Wilk test was used to analyze normality; parametric data were represented by the mean and standard deviation of the sample and compared by Student's t-test, with or without Welch correction, depending on the performance of Levene's Test regarding sample sphericity. When evaluating more than two samples, analysis of variance of non-repeated measures was used with Tukey or Games-Howell posttest, according to the analysis of variance (homo or heteroscedasticity, respectively) and with bootstrap. Covariate analyzes were performed to adjust the outcome variables with Sidak's post-test. Nonparametric data were represented by the median of lower and upper quartiles and compared with each other using the Mann-Whitney test for two samples, and when comparing more than two samples, the Kruskal-Wallis test was used. When normalization was necessary, the variables were log-transformed. Frequencies were compared using the Chi-square test. A value of P < 0.05 was considered statistically signi cant. IBM® SPSS Statistics (version 27.0), GraphPad Prisma (version 5.04) for Windows, and Microsoft® Excel for Mac (version 16.52) software were used for data tabulation and analysis.

Results
In the BC group, the median age and the frequency of postmenopausal status, dyslipidemia and hypertension were higher in comparison to the CTR, although BMI, overweight (BMI 25 kg/m 2 ) and reported statin use were similar between groups. Among tumors, the frequency (%) of histological types was: ductal (87.7), lobular (7.4), mucinous (4.3), and metaplastic (0.6). As expected, a higher frequency of LA and LB tumors was observed. 70.8% of BC women were categorized in clinical stages I and II, and 29.2% in advanced stages of disease (stages III and IV) ( Table 1   Comparisons were made using the Mann-Whitney test; data shown in median and interquartile ranges (25%-75%). P values were age-adjusted. When comparing the molecular types of BC, age and BMI were similar, but TC was higher in TN as compared to LA, LB, and HER2 tumors. Besides, plasma levels of TG, apoB, and non-HDLc were greater in TN as compared to LB and HER2 (Table 3). Although plasma lipids were different in TN tumors, the ratios TC/apoB (P = 0.065) and TG/HDLc (P = 0.091) did not reach statistical difference. Oxysterols in HDL were similar among LA, LB, HER2, and TN as well as the ability of cell cholesterol removal (Table 3).   Table 3.
When subjects were categorized according to the stage of the disease re ected by increased levels of Ki67, it was observed that plasma lipids and their ratios were similar among groups. The composition of HDL in TC, PL, apoA-I, and oxysterols was similar among stages. Only HDL-TG was increased on stage III as compared to I, II, and IV. Reduced cholesterol e ux was observed with HDL isolated from BC women in more advanced stages of the disease (III and IV) as compared to stages I and II. (Table 4).

Discussion
In the present investigation, it was addressed plasma lipid levels and HDL composition and functionality in newly diagnosed women with BC in comparison to CTR women, and with respect to the molecular classi cation of the tumor and the clinical stage of the disease. In comparison to CRT, it was demonstrated that BC cases had 1) similar concentrations of plasma lipids adjusted per age, and 2) HDL particles less enriched in TC, PL, and TG. Divergent results have been published concerning the association of plasma lipids and lipoproteins in the development and prognosis of BC. Total cholesterol is considered a contributor factor for many types of cancer including BC, although the presence of confounding factors such as heterogeneity of BC, disease duration and staging, disparity of ethnic population, age, menopausal state, lifestyle, and treatments may bias the interpretation and comparison among studies. Moreover, diabetes mellitus, insulin resistance, and obesity that change lipoprotein pro les are also recognized as potential contributors to BC.
Particularly, the majority of studies show increased levels of TC, LDL, and TG and reduced HDLc as strong predictors for BC [41][42][43][44][45]. In our case-control study, HDL plasma levels did not differ between CTR and BC cases. Body mass index was similar between groups and results adjusted per age showed similar levels of plasma TG. Conceivably, triglycerides metabolism by the lipoprotein lipase dictates HDL formation in circulation and ultimately HDLc plasma levels. Noteworthy to mention that diabetes mellitus and smoking which are conditions that reduce HDL were not included in the study and there was no interference of therapeutics since only newly diagnosed subjects were included.
Although the ability of HDL in removing cell cholesterol was preserved in all women with BC included in our casuistic as compared to CTR, the diminished content of TC, PL, and 27HC in the HDL particle from BC cases may be a consequence of alterations in cell tumors that mitigate sterol exportation to HDL. In addition to the intrinsic ability of HDL to receive cholesterol, it is important to consider that lipid e ux is modulated by cellular components, such as the bioavailability of free or unesteri ed cholesterol and the content and functionality of ABCA-1, ABCG-1, and SR-BI [46]. Characteristically, ABCA-1 is expressed in the mammary gland and its content is indicated as reduced in BC and associated with positive lymph nodes [47]. Its expression negatively affects the therapeutic e ciency of chemotherapy [48] and is referred to by some authors as a marker of TN tumors [49]. On the other hand, its de ciency contributes to an increase in cellular and mitochondrial cholesterol content, mitigating cell death processes mediated by this organelle, which favors tumor cell survival [24].
Solid tumors accumulate large amounts of cholesterol [50,51] through increased synthesis and uptake of lipoproteins [51,52]. SR-B1 mediates the e ux of free cholesterol down a concentration gradient for HDL. However, it can also promote the uptake of modi ed lipoproteins, favoring, on the contrary, the supply of cholesterol to cells and tumor progression [53]. Higher expression of SR-B1 is linked to greater aggressiveness and worse tumor prognosis [23,54,55], while mutations in the Scarb1 are related to the inhibition of tumor proliferation [56].
Furthermore, the lower content of lipids in HDL could be linked to the lower detachment of surface components of TG-rich lipoproteins during lipolysis mediated by lipoprotein lipase, in a process recently referred to as reverse transport of remnant cholesterol [57]. However, it is not possible to say which mechanisms act on the composition of HDL in view of the present results but these events cannot be evidenced by the simplistic determination of HDLc in plasma, explaining some of the controversies regarding the association of HDLc with BC.
As expected, most tumors in our casuistic were of the luminal type (68%), with a predominance of stage I and II disease (70.8%). The frequency of advanced disease (stages III and IV) was highest in women with TN tumors (66.5%) and lowest in those with LA tumors (16%). Taking into account the molecular classi cation of the tumor it was observed that: 1) TN cases had higher levels of TC, apo B, nonHDLc, and TG in comparison to LA, LB, and HER-2, but 2) HDLc and composition and functionality of the HDL particle were similar among all types. Changes observed were irrespective of BMI and may represent a distinct feature of TN tumors, channeling lipids to tumor development in a hystotype with a more aggressive clinical history, with a relatively low-survival rate and a high rate of metastasis.
Findings agree with previous studies that evidenced increased levels of VLDL-TG in TN BC, although changes in CT/apoB and TG/HDLc did not reach statistical signi cance. In this regard, it was demonstrated in a retrospective cohort that a high pretreatment TG/HDLc ratio was an independent predictor of the overall survival rate in TN tumors [58]. The enhanced expression of the LDL receptor and LDLR-related proteins 5 and 6 were found in TN tumors and associated with a higher ability of tumor growth and invasion; while the knockdown of LDLR-5 and 6 decreased tumorigenesis [51,[59][60][61].
A model of syngeneic tumor graft evidenced that tumor growth is accompanied by changes in the host lipid metabolism by stimulating and inhibiting, respectively, the synthesis and metabolism of very-low- According to the clinical stage of the disease, a reduced intrinsic ability of the HDL particle in removing cell cholesterol was observed in stages III and IV as compared to stages I and II. This was independent of changes in HDL composition and plasma lipids reinforcing the dissociation between plasma levels of HDLc and HDL functionality as a whole particle. Increasing levels of intracellular cholesterol favor the formation of oxysterols allowing us to infer an increase in the concentration of 27HC in the tumor microenvironment that negatively drives BC evolution. The pathophysiological basis of reduced HDL functionality in advanced stages of breast cancer was not investigated in the present study. The in ammatory, oxidative, and immunogenic insults prevalent in advanced disease may favor the modi cation of HDL in the tumor microenvironment, compromising its function. In this sense, HDL may even facilitate tumor propagation as previously reported in experimental studies where modi ed HDL was oncogenic [41,67]. Again, this supports the idea that the tumor energy demand adapts systemic metabolism in its favor.
The concept of HDL modulation by the tumor by reverse causation may unlink HDL, per se, as a predictor of tumor risk. In this sense, HDL would be applied more as a marker of tumor evolution than as a protector or inducer of its genesis.
To the best of our knowledge, this is the rst demonstration of loss of HDL function according to disease burden, independently of plasma lipids. Study limitations include the absence of more detailed clinical data, including components of metabolic syndrome and visceral adiposity, and lifestyle information. However, BMI was similar between the groups, and comorbidities frankly associated with changes in lipid metabolism were excluded. It was not possible to determine which components of the HDL structure, independently of lipids and apo A-I, were responsible for the impairment of cholesterol e ux. A more detailed analysis of HDL proteomics, lipidomics, and microRNAs may add new information.

Conclusions
The results of the present investigation con rmed an increased level of plasma lipids in TN tumors probably driving the more aggressive type of BC. Moreover, ndings contributed to a better understanding of the role played by HDL in BC, especially related to patients with a worse prognosis where the functionality of HDL in the removal of excess cellular cholesterol is impaired. As compared to controls only the composition of HDL was discriminative in BC subjects. A prospective cohort will help determine the prognostic value of HDL composition and functionality in predicting BC outcomes.