Sarcosine promotes trafficking of dendritic cells and improves efficacy of anti-tumor dendritic cell vaccines via CXC chemokine family signaling
Dendritic cell (DC) vaccine efficacy is directly related to the efficiency of DC migration to the lymph node after delivery to the patient. We discovered that a naturally occurring metabolite, sarcosine, increases DC migration in human and murine cells resulting in significantly improved anti-tumor efficacy. We hypothesized that sarcosine induced cell migration was due to chemokine signaling.
DCs were harvested from the bone marrow of wild type C57BL/6 mice and electroporated with tumor messenger RNA (mRNA). Human DCs were isolated from peripheral blood mononuclear cells (PBMCs). DCs were treated with 20 mM of sarcosine. Antigen specific T cells were isolated from transgenic mice and injected intravenously into tumor bearing mice. DC vaccines were delivered via intradermal injection. In vivo migration was evaluated by flow cytometry and immunofluorescence microscopy. Gene expression in RNA was investigated in DCs via RT-PCR and Nanostring.
Sarcosine significantly increased human and murine DC migration in vitro. In vivo sarcosine-treated DCs had significantly increased migration to both the lymph nodes and spleens after intradermal delivery in mice. Sarcosine-treated DC vaccines resulted in significantly improved tumor control in a B16F10-OVA tumor flank model and improved survival in an intracranial GL261-gp100 glioma model. Gene expression demonstrated an upregulation of CXCR2, CXCL3 and CXCL1 in sarcosine- treated DCs. Further metabolic analysis demonstrated the up-regulation of cyclooxygenase-1 and Pik3cg. Sarcosine induced migration was abrogated by adding the CXCR2 neutralizing antibody in both human and murine DCs. CXCR2 neutralizing antibody also removed the survival benefit of sarcosine-treated DCs in the tumor models.
Sarcosine increases the migration of murine and human DCs via the CXC chemokine pathway. This platform can be utilized to improve existing DC vaccine strategies.
KeywordsDendritic cell vaccine Immunotherapy Glioblastoma Cell migration Sarcosine CXCR2
Antigen presenting cells
- Cox 1
C-X-C chemokine receptor type 2
Glycine transporter type-1
Granulocyte macrophage colony stimulating factor
hepatitis C virus
Human immunodeficiency virus
Institutional Animal Care and Use Committee
Peripheral blood mononuclear cells
Reverse transcription polymerase chain reaction
Regulatory T cells
Antigen presenting cells (APCs) such as dendritic cells (DCs) play a critical role in activating the adaptive immune response against pathogens. DCs can effectively stimulate T cells via specialized pathways and activate them against specific antigens including antigen relevant in patients with cancer. This mechanism leads to potent immune responses that can be leveraged for the treatment of traditionally resistant tumors such as glioblastoma. DC vaccines are a novel and versatile treatment approach, and this strategy is already FDA approved for the treatment of prostate cancer [1, 2].
Due to the versatility of DC vaccines, studies are ongoing for their use in resistant malignancies like glioblastoma (GBM). Phase I/II trials in GBM patients demonstrate feasibility and safety of creating and delivering DC vaccines as well as robust anti-tumor immune responses in select patients [1, 3, 4, 5]. Importantly, DC vaccine efficacy has been shown to be tightly related to the efficiency of DC migration to the lymph node after delivery to the patient . Our group has demonstrated that patients with GBM who had increased DC migration by co-vaccinating with tetanus toxoid had significantly larger anti-tumor immune responses and improved survival (18.5 versus 36.6 months respectively in the control and treatment arms) . Therefore, methods to improve immune cell migration have the potential to make cellular immunotherapeutic strategies more potent.
In the search for a metabolite to track DCs in vivo, we began experimenting with sarcosine (N-methyl glycine) to label DCs for tracking the cells in vivo using MRI. We chose sarcosine because it is naturally occurring and non-toxic, has low expression in brain and lymph node tissue, and it is commercially available. During experimentation, we observed that sarcosine treated DCs seemed to have improved migration. We therefore began exploring the use of sarcosine as an adjuvant during DC vaccination and its impact on anti-tumor efficacy. The objective of these studies was to characterize the impact of sarcosine on DC function and migration in the context of an intradermal DC vaccine for the treatment of intracranial tumor models.
Sarcosine 98% (Synonym: N-Methylglycine) was purchased from Sigma-Aldrich. Sarcosine was dissolved in T cell or DC culture medium at 20 mM and sterile solution was prepared via passing through 0.22 um low protein binding Durapore membrane (Life Science). Intra cellular sarcosine concentration was evaluated with sarcosine assay kit from Sigma-Aldrich that detects the concentration of sarcosine via colorimetric observation at 570 nm. DCs were collected from culture media in different groups and were mixed with sarcosine assay buffer, sarcosine probe and sarcosine enzyme mix that were provided in the kit. Sarcosine standard samples were prepared. The manufacturer instructions were followed and samples were incubated in plates in a dark environment at 37 °C for 60 min. Concentrations of sarcosine in different groups were measured via colorimetric method.
DC vaccine generation
Bone marrow (BM) was harvested from the long bones and sternum of C57BL/6 mice which were purchased from Jackson Laboratory (Bar Harbor, ME). All animal projects and protocols were initiated after achieving the approval from the Institutional Animal Care and Use Committee (IACUC) of the University of Florida. Myeloid-derived cells were cultured in DC complete media including granulocyte macrophage colony stimulating factor (GM-CSF) and interleukin 4 (IL-4). Cells were cultured in six well plates for 6 days. At day 7, cells were re-plated in 60 mm dishes and DCs were electroporated with OVA-mRNA at day 8. At day 9, DCs were collected in phosphate buffered saline (PBS) for administration. DC vaccines were delivered via intradermal injection in the inguinal area.
Human DC vaccine generation
Whole blood from five healthy donors was purchased from Life South blood bank. Human DCs were generated from blood monocytes by incubating them with IL-4, GM-CSF, TNF-α, IL-1β and IL-6. This method has been previously described . pp65 RNA was produced and transfected from the full-length cDNA that was donated by Dr. Bill Britt (University of Alabama-Birmingham, Birmingham, Alabama).
Adoptive cell transfer
Spleens were harvested from PMEL or OT-I T transgenic mice. They were sliced and passed through cell strainer. Lysing solution was used to get rid of red blood cells. Supernatant was discarded and splenocytes pellet was re-suspended at 3 × 107 cells/50 ul of PBS for Intravenous (IV) infusion.
In vitro migration assay
DCs from different groups were transferred to a Corning Costar Transwell plate (Pore Size: 5.0 μm; 6.5 mm Diameter; 0.33cm2 Growth Area) that included an upper and lower chamber. The upper chamber had cells in 100 ul media without cytokines and serum. The lower chamber contained 500 ul of media with serum, CCL19 (250 ng/ml), CCL21 (250 ng/ml) and CXCL3 (250 ng/ml). The upper chamber was cast off after 5 h and the migrated cells into the lower chamber were counted on hemocytometer. In the in vitro migration experiments, CXCR2 activity was inhibited by treating cells for one hour before the migration assay with mouse anti-CXCR2 (at 1 to 50 dilution, Clone242216, R&D Systems) . For in vivo application, SB225002 (selective non-peptide antagonist of CXCR2, Sigma-Aldrich) was dissolved in vehicle (NaCl 0.9% solution plus Tween-80 0.33%) according to manufacturer’s instructions. SB225002 was injected into the mice through intraperitoneal (IP) at 50 μg (1.4 × 10–7 mol) in 200 μl per animal one hour prior each DC vaccine injection [9, 10]. In the human DC in vitro migration experiments, cells were treated with SB225002 at a concentration of 10 μM for one hour prior to the migration studies .
For the intracranial tumor model, GL261-Gp100 tumor cell suspension was prepared at 2 × 105 cells in 1:1 mixture of PBS and methylcellulose in total volume of 2.5 ul per brain. Tumor cells were implanted into the intracranial space stereotactically. The injection needle was positioned 2 mm to right and 1–3 mm above the sagittal-bregma intersection. Once the needle was inserted 3–4 mm into the brain, 2.5 ul of prepared cell mixture was injected over one minute. Bone was used to cover the injection site. One day after tumor implantation, the first GP100-RNA pulsed DC vaccine (1 × 106 cells/mouse) was injected intradermally and 48 h later 3 × 107 PMEL splenocytes were given IV. Then treatment groups received second and third GP100-RNA pulsed DC vaccines every 5 days intradermally. Animals were followed for survival analysis and were euthanized when they reached endpoint.
For the flank model, B16F10-OVA cells were grown in Dulbeccos modified eagles medium (DMEM) and the cells were inoculated into the flank of C57BL/6 mice subcutaneously at a concentration of 1 × 106 cells per 100 ml PBS. The tumor bearing mice were randomized before the first DC vaccine injection. At day 8 of post tumor implantation, 3 × 107 OT-I splenocytes were given IV and the first OVA-RNA pulsed DC vaccine was injected intradermal (1 × 106 cells/mouse). At day 10 post tumor implantation tumor size was measured at flank sites every 2 days. Mice received a second and third vaccine of OVA-RNA pulsed DC vaccines on day 12 and 16. Tumor volume was calculated in millimeters cubed (mm3) by the formula (length x width2 × 0.52) in a rectangular fashion. The animals were euthanized when tumor growth passed two centimeter in any dimensions or ulceration occurred in tumor side. A linear model was accounted to analyze the correlation of tumor volume and time in each animal.
DCs were stained with PKH26 (Red Fluorescent Cell membrane stain, Sigma-Aldrich) immediately before intradermal injection in the inguinal area. At 48 h post DC vaccine administration, spleens or inguinal lymph nodes were collected and immediately embedded inside optimal cutting temperature (OCT) compound then frozen inside the chamber that contains liquid nitrogen. The samples were cut via HM 505E cryostat machine at 6 um thickness and transferred on glass slides for immunofluorescence microscopy analysis. The slides washed with PBS at room temperature and non- specific sites were blocked by incubating sections with 2% rabbit serum was used to block the nonspecific sites in the tissue sections. Fluorophore conjugated antibodies (Anti-CD45R/B220 and Streptavidin, eBioscience)were added to the sections and incubated for overnight at 4 °C. The final stained sections were observed via EVOS inverted immunofluorescence microscope at different objectives.
To evaluate the in vivo migration, bone marrow-derived DCs from green fluorescent protein (GFP) expressing transgenic mice were injected into the wild type C57BL/6 mice intradermal. Draining inguinal lymph nodes (LN) were harvested, digested and single cell suspensions were prepared to count GFP expressing cells, and absolute cell counts were quantified in each sample.
For immune response analysis, spleens were harvested from vaccinated mice after 7 days post vaccination. RBCs were lysed, then splenocytes were transferred into in RPMI-1640 medium (Invitrogen) containing with 10% FCS, 1% l-glutamine, 1% penicillin/streptomycin. 2 × 106 cells were transferred to each well in round-bottomed 96-well plates. Cells were centrifuged at 500 g for 5 min. The cells were resuspended in staining buffer (Thermo Fisher Scientific) and stained with different conjugated antibodies at room temperature for 15 min in dark environments. The cells were stained with fixable-yellow Live/dead Then anti-CD3 (APC-Cy7), anti-CD4 (BV421), anti-CD8 (FITC), anti-CD25 (APC) from BD Bioscience and OT-1 tetramer (PE) from MGL and followed by fixation with 2% paraformaldehyde for 5 min at room temperature.
To evaluate cell phenotype, sarcosine treated and untreated DC cells were prepared by using the above method. Then murine DC cells were stained in separate wells with anti-CD11c (PE), anti-MHC class II (PE), anti-CD80 (PE) and anti-CD86 (PE) from BD Bioscience. Human monocyte derived DCs were stained with anti-CD11c (PE), anti-CD86 (PE) and anti-HLA-DR2 (PE) from BD Bioscience. The BD LSR flow cytometer and FlowJo software were used for all flow cytometry analyses.
RNA was extracted from murine myeloid DCs by RNeasy Mini Kit (QIAGEN). After reverse transcripting the extracted RNA, cDNA was implemented for gene expression analysis. Genomic expression of mouse cytokines, chemokines and chemokine receptors was evaluated using PrimePCR Selections 96 wells array plates (Bio-Rad). PrimePCR Selections 96 well array plates were analyzed using Bio-Rad Reverse Transcription PCR (RT-PCR) machine (CFX96 Touch).
An nCounter muse myeloid immune cells profiling panel was implicated to analyze the gene expression via NanoString™ Technology (XT_PGX_MmV2_Myeloid_CSO). Total RNA was extracted from sarcosine treated murine myeloid-derived DCs and the control DCs using RNeasy kit (Qiagen). nSolver software was used for data analyzing, and data was normalized by implication of positive, negative control probes and housekeeping genes.
Detection of intracellular oxidative stress level
To evaluate the intracellular oxidative stress level, sarcosine treated and untreated murine BM-DCs were prepared by using the above method. DCs were collected at day 7 without any electroporation and washed with warm PBS. Cells were reconstituted at 1.5 × 105 cells/well. Then CM-H2DCFDA (General Oxidative Stress Indicator, Invitrogen) dye was added to cells with 10 μM and the cells were incubated at 37 °C for 60 min based on manufacturer protocol. Then cells washed three times with warm PBS and returned to pre-warmed growth medium and incubated at 37°c for 10 min. Cell plates were measured on fluorescence plate reader at Excitation: 495/Emission: 527 nm. For negative control, unstained cells were examined to deduct the autofluorescence. For positive control, cells were treated with tret-butyl hydroperoxide (TBHP) with 100 μM. Fluorescence intensity was evaluated based on Z score. (z = (expression in treatment sample - mean expression in negative control sample) / standard deviation of expression in negative control sample).
Measurement of antigen uptake
Sarcosine treated and untreated murine BM-DCs generated as described above. Cells were incubated with 2 mg/ml FITC-OVA at 37 °C for 90 min. Then cells were washed three times with PBS and stained with PE-anti-CD11c. FITC-OVA uptake was evaluated as mean fluorescence intensity (MFI) in CD11c + population. Nonspecific signal of FITC was quantified by incubating cells in 2 mg/ml OVA-FITC at 0 °C for 90 min .
Spleens were collected from OT-I mice (from Jackson Laboratories) and CD8+ T cells were isolated through magnetic isolation kit (Miltenyi Biotec). CD8+ T cells were reconstituted in 96 well plate at 1 × 105/200 μl/well and labeled with carboxy fluoroscein succinimidyl ester (CFSE; Molecular Probes). Murine BM-DCs were generated and prepared as described above and electroporated with OVA-mRNA. Labeled CD8+ T cells were co-cultured with sarcosine treated and untreated BM-DCs at 2.5 × 104 cells/200 μl/well. Cells were cultured in RPMI 1640 medium with 10% Fetal Bovine Serum (FBS). Non-electroporated BM-DCs were co-cultured with CD8+ T cells as a negative control. After three days cells harvested and T cells proliferation was quantified through flow cytometry by analyzing CFSE dilution in CD3+/CD8+ population.
GraphPad Prism 7 software was used for all statistical analysis. Statistics were analyzed using one-way ANOVA for studies with multiple groups and Mann Whitney or t-tests for comparisons of two groups. Two way ANOVA was used for tumor volume and body weight analysis. Median survival was analyzed using Log-rank (Mantel-Cox) test. Unpaired t-test method was accounted for flow cytometry data. The data was considered statistically significant when p value was < 0.05. The level of significance was indicated via asterisks including p > 0.05 not significant, *p < 0.05, **p ≤ 0.01 and ***p < 0.001.
Loading cells with sarcosine
The impact of sarcosine on cell phenotype was evaluated using flow cytometry. All groups were cultured in DC media that contained sarcosine for 48 h. Sarcosine treatment did not result in cell death or changes in CD11c, MHC-II, CD80 and CD86 expression compared to control DCs (Fig. 1b).
In vitro and in vivo migration
Next, bone marrow derived DCs were used to test in vivo migration. C57BL/6 mice received the DC vaccines via intradermal injection in the inguinal area. Forty-eight hours post vaccination, draining lymph nodes and spleens were harvested for flow cytometry and immunofluorescence microscopy. Sarcosine treated DCs had significantly increased migration to the draining lymph nodes measured by flow cytometry (mean 9.457% cells for control versus 25.30% for sarcosine treated DCs, p < 0.0411) (Fig. 2b). This was confirmed using immunofluorescence (Fig. 2c, e). Interestingly, these DCs were also visualized in the spleens at just 48 h after intradermal injection (328.5 ± 43.71 control versus 128 ± 2 cells sarcosine, p < 0.05) (Fig. 2d & f).
Other functions of DCs were also tested in the setting of sarcosine treatment. Antigen uptake was measured in sarcosine treated DCs by co-culturing cells with FITC-OVA. Sarcosine resulted in decreased antigen uptake (Additional file 1: Figure S1a) However, antigen presentation measured using a T cell proliferation assay was not affected by sarcosine treatment (Additional file 1: Figure S1b-e).
Systemic immune response and anti-tumor efficacy of sarcosine treated DCs
The B16F10-OVA melanoma flank model was used to test the impact of sarcosine DCs on tumor growth. After tumor implantation, animals received infusion of antigen specific OT-1 splenocytes and DC vaccines with or without sarcosine 10 days post tumor injection. The animals that treated with sarcosine treated DCs had significantly slowed tumor growth over time compared to animals receiving non-sarcosine treated DC vaccines. Twenty-six days after tumor implantation the mean tumor volume was 1491 mm3 for the controls, 905 mm3 for DC treated mice, and 338.8 mm3 for sarcosine treated DC treated mice (p = < 0.0001, 2 way ANOVA, n = 10) (Fig. 3d). Sarcosine treated DC treatment did not result in toxicity as measured by total body weight (Additional file 1: Figure S2a). Sarcosine treated B16F10-OVA tumor cells were also not associated with an increase in tumor growth or invasion (Additional file 1: Figure S2b).
Gene expression analysis of sarcosine treated DCs
CXCR2 blockade effect on sarcosine induced migration
Next, the impact of CXCR2 blockade on anti-tumor efficacy was assessed. B16F10-OVA melanoma tumors were implanted. Animals received infusion of antigen specific OT-I splenocytes and DC vaccines with or without sarcosine. Some animals also received CXCR2 neutralizing antibody one hour before each DC vaccine injection. The animals that were treated with sarcosine treated DCs had significantly slowed tumor growth over time compared to animals receiving non-sarcosine treated DCs and sarcosine treated DCs treated with CXCR2 neutralizing antibody. The mean tumor volume 34 days after tumor implantation was 2926 mm3 for controls, 1756 mm3 for DC treated mice, 930.7 mm3 for sarcosine treated DC treated mice, 1778 mm3 for sarcosine treated DCs plus anti-CXCR2 treated mice, and 3111 mm3 for anti-CXCR2 treated mice (Fig. 6c). To further evaluated the role of CXCR2 in a relevant intracranial tumor model, GL261-GP100 tumor cells were implanted in mice via an intracranial injection. Animals received antigen specific PMEL splenocytes and DC vaccines. One group received CXCR2 neutralizing antibody one hour before each DC vaccine injection. The animals that were treated with sarcosine treated DCs had significantly prolonged survival compared to animals receiving non-sarcosine treated DCs and sarcosine treated DCs treated with CXCR2 neutralizing antibody. The median survival was 24.5 for control, 27.5 for DC treated mice, 35 for sarcosine treated DC treated mice, 26 for sarcosine treated DCs plus anti-CXCR2 treated mice (Fig. 6d).
Sarcosine effects on human DC phenotype and migration
DC vaccines are a versatile and potentially potent therapy for treatment resistant tumors such as GBM. Phase I and II studies of DC vaccines for GBM have demonstrated the ability to induce potent adaptive immune responses in patients [6, 13, 14]. We currently have an ongoing phase II clinical trial testing a CMV pp65 RNA DC vaccine for newly diagnosed GBM in which select patients have demonstrated robust immunologic and radiographic responses to treatment (ATTAC II, NCT 02465268). Our prior data has demonstrated that DC vaccine efficacy is predicted by efficient DC migration . Therefore, sarcosine-induced migration has the potential to greatly impact the translation of DC vaccines into an efficacious treatment platform for patients.
Our current data demonstrate a survival benefit of DC vaccines for an intracranial tumor model when sarcosine is added to the DCs. Prior murine studies have only shown a survival benefit when DCs are given prior to tumor implantation or given as an IP injection [15, 16]. The increased DC migration achieved with sarcosine in our studies converted an otherwise non-efficacious platform into a therapy with a survival benefit. Our study is the first description of leveraging sarcosine to increase the migration of immune cells to enhance immunotherapy. Importantly, the doses of sarcosine that used to increase DC migration do not induce tumor invasiveness or growth by itself. In addition, our data demonstrate that sarcosine treated DCs preserve the ability to present antigen and induce T cell proliferation.
These data show that the mechanism of sarcosine enhanced migration is dependent on the upregulation of CXCR2. The findings of CXCR2 upregulation in DCs is a novel finding, although CXCR2 is a known regulator of migration in human immune cells . Human dendritic cells express IL-8 receptors including CXCR1 and CXCR2 and IL-8 can attract dendritic cells through its receptors . CXCR2 expression levels in immature DCs are typically higher than mature DCs . In addition, DCs can secrete IL-8 [19, 20] and CCL5 (RANTES), MIP-la, and MCP-3  chemokines for which CXCR2 is receptor, pointing to possible autocrine function of CXCR2 for DC migration . We have shown that blocking CXCR2 nullifies sarcosine induced DC migration. Sarcosine is known to compete with glycine for the glycine transporter type-1 (Gly-T1) receptor on the cell membrane , therefore decreasing intracellular glycine. This reduction results in oxidative stress [23, 24]. This stress likely leads to an increase in arachidonic acid and Cox 1 . Cox 1 upregulates the CXC chemokine family  leading to an increase in cell migration [21, 27, 28, 29, 30].
Overall, sarcosine is a non-toxic compound that increases DC migration leading to improved outcomes in a tumor model treated with sarcosine treated DC vaccines. Sarcosine has a similar effect on human DCs. Therefore, this strategy could be easily translated into clinical protocols being utilized to treat cancer with DC based immunotherapy. Translational studies are necessary to further evaluate the efficacy of a sarcosine treated DC vaccine strategy in the treatment of brain tumors.
Sarcosine increases the migration of murine and human DCs via the CXC chemokine pathway. This increase in DC migration also resulted in a more robust anti-tumor immune response and better tumor control and prolonged survival in the intracranial and flank murine tumors models. Sarcosine is non-toxic to murine and human DCs. Further studies in human subjects are necessary to determine the utility of this treatment platform.
We acknowledge Georges Abboud and Elizabeth Ogando-Rivas for their expert technical assistance.
Provisional US Patent Pending US2017/023623 (Methods, kit and compositions for enhancing cellular therapy).
FD, MR, HA and RL conceived and designed the experiments. FD conducted the experiments. FD, AK, AD, AV and MS performed experiments. FD and MR analyzed the data and interpreted the results of the experiments. CY analyzed the PCR data. ES contributed toward editing and finalizing the manuscript. MR and DM supervised the study. All authors have read and approved the final version of the manuscript.
This study was funded by Ocala Royal Dames, ABC2/FCBTR and NINDS K08 grants held by Dr. Maryam Rahman.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
- 1.Farhad D, Aida K, Duane M, Maryam R. Dendritic cell therapy. Chichester: eLS john Wiley & sons, Ltd; 2018.Google Scholar
- 5.Ardon H, Van Gool SW, Verschuere T, Maes W, Fieuws S, Sciot R, et al. Integration of autologous dendritic cell-based immunotherapy in the standard of care treatment for patients with newly diagnosed glioblastoma: results of the HGG-2006 phase I/II trial. Cancer Immunol Immunother. 2012;61(11):2033–44.CrossRefGoogle Scholar
- 7.Nair S, Archer GE, Tedder TF. Isolation and generation of human dendritic cells. Curr Protoc Immunol. 2012;Chapter 7:Unit7 32.Google Scholar
- 13.Prins RM, Soto H, Konkankit V, Odesa SK, Eskin A, Yong WH, et al. Gene expression profile correlates with T cell infiltration and relative survival in glioblastoma patients vaccinated with dendritic cell immunotherapy. Clinical cancer research : an official journal of the American Association for Cancer Research. 2011;17(6):1603–15.CrossRefGoogle Scholar
- 14.Liau LM, Prins RM, Kiertscher SM, Odesa SK, Kremen TJ, Giovannone AJ, et al. Dendritic cell vaccination in glioblastoma patients induces systemic and intracranial T-cell responses modulated by the local central nervous system tumor microenvironment. Clin Cancer Res. 2005;11(15):5515–25.CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.