LDH-A Knockdown: Changes in The LDH Isoenzyme Pro le and Variability in Glioma Response

Masahiro Shindo Memorial Sloan Kettering Cancer Center Masatomo Maeda Memorial Sloan-Kettering Cancer Center Ko Myat Memorial Sloan Kettering Cancer Center Mayuresh Mane Memorial Sloan Kettering Cancer Center Ivan J. Cohen Memorial Sloan Kettering Cancer Center Kiranmayi Vemuri Memorial Sloan Kettering Cancer Center Avi S. Albeg Memorial Sloan Kettering Cancer Center Inna Serganova Memorial Sloan Kettering Cancer Center Ronald G Blasberg (  blasberr@mskcc.org ) Memorial Sloan Kettering Cancer Center


Background
Lactate metabolism in tumors has been intensively studied recently: 1) lactate is now considered a major energy source for many tumors; 2) lactate is the major gluconeogenic precursor and 3) lactate exhibits signaling function properties (1). In this study, three murine gliomas (GL261, CT2A ,and ALTS1C1) were explored with regards to the impact of LDH-A downregulation on tumor biology, since there has been limited information on the role of the LDH-A/lactate axis in tumors of brain origin (1-4). Lactate formation in gliomas is associated with poor survival and contributes to the suppression of local immunity (5). The relationship between LDH-A expression levels and GBM malignancy, using human glioma cells and its impact on proliferation and apoptosis has been explored (6) , (7). It has long been known that many human cancers have higher LDH-A levels compared to normal tissues (8)(9)(10). It has also been shown that LDH-A plays an important role in the development, invasion ,and metastasis of malignancies (10)(11)(12). The LDH enzyme (EC 1.1.1.27, LDH) is composed of two proteins, LDH-A (predominantly found in skeletal muscle and many solid tumors) and LDH-B (predominantly found in heart muscle and brain). The LDH enzyme is a tetramer and exists in several different electrophoretic forms known as isoenzymes. They catalyze the same biochemical reaction but differ in their kinetic characteristics, physicochemical properties (different net charge) ,and response to the inhibition by pyruvate (13). The LDH tetrameric enzyme exists in two basic homo-tetrameric forms: i) LDH5 (A4 or M4) contains 4 LDH-A subunits, and ii) LDH1 (B4 or H4) contains 4 LDH-B subunits. In addition to homotetramers, LDH also exists in three hybrid forms, resulting in ve structural entities that vary in expression level in different tissues (10). The LDH-A and LDH-B isoforms occupy the mitochondrial compartment, plasma membrane and cytosol (14). Traditionally LDH-A participates in converting pyruvate to lactate, whereas LDH-B has a higher a nity for lactate, converting lactate to pyruvate (15). LDH-B has been considered to facilitate the use of lactate as a carbon energy source. More recently, it has become clear that lactate is both created and consumed in aerobic conditions, and serves as a link between glycolytic and oxidative metabolism (16). Analyses of LDH-A and LDH-B expression levels in tumors have shown that LDH-A is highly expressed in most neoplastic tissues (17)(18)(19)(20). However, the role of LDH-B and its regulation is less explored (21,22).
The role of LDH-A and LDH-B in tumor biology is complex. The relationship between tumor cell growth, the balance between LDH-A and LDH-B, the effects on tumor cell metabolism, and the tumor microenvironment (TME) is variable across different types of gliomas.

Aim
In this report, we study and compare three murine glioma cell lines and tumors following LDH-A shRNA knockdown (KD). The objective was to explore and compare the effect of LDH-A knockdown (KD) on the expression levels of LDH-A and LDH-B mRNA, protein, LDH enzymatic activity, the effect on the LDH isoenzyme pro les and the impact of these changes on the growth of cancer cells in vitro and in vivo.

Cells and culture conditions
The GL261 murine glioblastoma cell line was obtained from NCI depository (23,24). The ALTS1C1 (ALT) murine glioblastoma cell line derived from SV40 large T antigen-transfected astrocytes was kindly provided by Dr. Chiang (Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Taiwan) (25) and the CT2A high-grade murine astrocytoma cell line was kindly provided by Dr. Seyfried (Biology Department, Boston College, Boston) (26). These cell lines were cultured in DMEM media supplemented with 25 mM glucose, 10% FCS, 4 mM glutamine, and penicillin/streptomycin. LDH-A KD (knock-down) and NC (negative control) cells, derived from each cell line, were grown in the same media and 2.5 mg/L of puromycin.
Generation of LDH-A knockdown and control cell lines GL261, CT2A and ALTS1C1 cells were transfected with Sure Silencing shRNA plasmids (QIAGEN, Frederick, MD, USA) to speci cally knock-down expression of the mouse LDH-A gene as described previously (27). Stably transduced clones (KD cell lines) were developed, along with a control (NC) cell line bearing a scrambled shRNA. Based on the previous experience (12,28), we decided to use the most effective shRNAs (shRNA-2) to develop LDH-A KD cells in murine glioma cells. The transfection of GL261 cancer cells with shRNA-2 resulted in a signi cant knock-down effect for LDH-A (approximately 10% of that in wild type cells), while bulk CT2A and ALTS1C1 cells transfected with shRNA-2 had a signi cantly less level of LDH-A knock-down (40-60%). In order to enrich the level of LDH-A knock-down we used a subcloning strategy for CT2A and ALTS1C1 cell lines (27).

Western blotting
All immunoblotting experiments were performed as described previously (11,29). Cell lines underwent

LDH enzyme activity
Total LDH enzyme activity was assessed using the Cytotoxicity Detection Kit PLUS (LDH) (Roche Diagnostics) as described before (27).
Proliferation assay in vitro 2×10 5 cells were seeded in 3 mL culture media in 6-well plates, followed by counting cells at 3 different time points, 48, 72, and 96 hours after seeding cells using Countess automated cell counter (Thermo Fisher Scienti c, Waltham, MA, USA). Each sample was triplicated, and media were changed every other day.

Animal models
The animal protocol was approved by the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center. Two strains of mice were used in animal experiments. First, 1×10 6  (Envigo) or athymic nu/nu male mice (Charles River Laboratories). Second, 1×10 6 cells in 100 µL PBS mixed with 100 µL matrigel were injected subcutaneously into the right ank of immunocompetent C57BL/6 male mice (Charles River Laboratories). The volume (V) of subcutaneous tumors was calculated from caliper measurements, where V = (π/6) × x × y × z where x, y, and z are 3 orthogonal diameters. Doubling times were calculated by the equation of trend lines using GraphPad Prism.

LDH zymography
Zymography, a common method to detect isoenzymes, was used to detect tissue-speci c differences in LDH isoenzymes. This approach can directly observe 5 isozyme bands in the active state (30). Based on their different electrophoretic motility, all LDH isoenzymes can be identi ed as LDH1 (B4 or H4), LDH2 (B3A1 or H3M1), LDH3 (B2A2 or H2M2), LDH4 (B1A3 or H1M3), and LDH5 (A4 or M4). The buffer system at pH 8.6 was chosen for the best separation of the ve LDH isoenzymes (30)(31)(32)(33). Because the B polypeptide has more acidic amino acid residues than the A polypeptide, LDH1/B has the highest migration rateand LDH5/A has the lowest migration rate. The electrophoretic mobilities of the LDH isoenzymes are: LDH 1/B > LDH 2 > LDH 3 > LDH 4 > LDH 5/A.

Immunohistochemical staining and image analyses
Dissected tumors were placed into 4% paraformaldehyde for further immunohistochemistry (IHC). The immuno uorescent (IF) staining was performed at Molecular Cytology Core Facility of MSKCC using Discovery XT processor (Ventana Medical Systems). 5 µm thick, para n-embedded sections were stained for H&E and LDH-A and LDH-B staining. The sections of tumors from nude or immunocompetent mice were stained by anti-LDH-A, anti-LDH-B. The small 5-day tumors from GL261 NC and KD were stained with immune markers: anti-CD68 antibody (Catalog No. TA1518, Boster), anti-CD4 (Catalog No. AF554, R & D Systems) and anti-CD3 antibody (Catalog No. A0452, Dako). Quanti cation of morphological characteristics was performed using trainable Weka Segmentation (Image J segmentation plugin) to assess the fraction of viable tumor cells, stroma, hemorrhage and necrosis in the H&E sections.
The same approach was used to quantify LDH-A and LDH-B staining (28,34).

Statistical analysis
Results are presented as mean ± standard error unless otherwise speci ed. Statistical signi cance was determined by a two-tailed Student t-test. A p-value of <0.05 was considered signi cant. All data presented for T cells assessment using IF staining were analyzed using GraphPad Prism (version 7.0; GraphPad Software) and are presented as mean +/-SD. Results were analyzed using the unpaired Student's t-test, and statistical signi cance was de ned as p<0.05.

Effects of LDH-A knockdown on murine glioma cells
We chose three murine brain tumor models to understand the impact of LDH-A downregulation on tumor phenotype and growth potential: GL261 (35), CT2A (26) ,and ALTS1C1 (25). The wild-type cell lines were transduced with two shRNA retroviral vectors: i) shRNA-2 was used for targeting LDH-A, it was found to provide effective knockdown, based on our previous work (28,36), and ii) one scrambled control (11). In vitro comparisons between the LDH-A knockdown (KD) and the scrambled control (NC) cell lines are shown in Figures 1 and 2, S-1. Signi cant differences were observed between the three control (NC) cell lines with respect to LDH-A and LDH-B mRNA levels ( Fig. 1 A, As expected, following LDH-A shRNA KD, LDH-A mRNA levels, protein expression and enzyme activity were all signi cantly reduced compared to the control (NC) cell lines. LDH-B was also affected by LDH-A shRNA knock-down but to a variable degree. The most notable difference was a signi cantly higher LDH-B/LDH-A ratio for mRNA, protein expression and LDH enzyme activity in GL261 KD cells (Fig. 1 F-H).

Effects of LDH-A knockdown on in vitro and in vivo growth pro les
The effect of LDH-A KD on cell proliferation (in vitro growth pro le plots) was compared to that of control NC cells ( Fig. 2A-C). The cell proliferation rate (doubling time) was calculated from an exponential t of the plots (Fig. 2D). LDH-A KD prolonged the in vitro doubling time only for GL261 cells but had little or no effect on CT2A and ALTS1C1 cells. The doubling times of wild type and NC cells were similar (data are not shown). The differences in the proliferation of cell lines may suggest that there are corresponding differences in their metabolic properties since metabolism and proliferation share common regulatory pathways in cancer cells (37)(38)(39).
To study the differences of tumor growth and phenotype in different host organ locations and in both immune competent (C57BL/6) and incompetent (Nude) animals, we initially studied the three gliomas in a subcutaneous (s.c.) location, and subsequently in an intracranial (i.c.) location (to be reported in a following manuscript). The effect of LDH-A KD on corresponding growth pro les of in vivo s.c. ank tumors in immune competent C57BL/6 mice are shown (Fig. 2E-G), and the tumor doubling times were estimated (Fig. 2H). As in the in vitro studies, LDH-A KD had a signi cant effect only on the growth of s.c. GL261 tumors in C57BL/6 mice (Fig. 2E). All s.c. GL261 LDH-A KD tumors were suppressed after developing small (<50 mm 3 ) tumors, whereas most GL261 A5NC tumors grew after a long 40-day delay period (Fig. 2E). Once GL261 A5NC tumors began to grow in C57BL/6 mice, the subsequent doubling time was similar to that of wild-type GL261 tumor growth in C57BL/6 mice (data are not shown). No signi cant difference was observed between the growth pro les of the NC and KD groups of CT2A and ALTS1C1 tumors in C57BL/6 mice (Fig. 2F, G, H). Wild-type CT2A and ALTS1C1 tumors grew slightly faster than NC tumors in C57BL/6 mice, but the difference was not statistically signi cant (data not shown). These results suggest that the LDH-A knockdown did not signi cantly alter the growth of s.c. CT2A and ALTS1C1 tumors growth in C57BL/6 mice compared with NC control tumors but had a clear effect on GL261 tumor growth in C57BL/6 mice -leading to tumor regression.
In immune compromised nude mice, the growth and doubling times of NC and LDH-A KD GL261, CT2A and ALTS1C1 tumors were similar (Fig. S2A-D). Interestingly, GL261 NC tumors grew at a faster rate (doubling time, 3.1 ± 1.5 days) in nude mice than GL261 NC tumors in C57BL/6 mice (7.1 ± 0.3 days) (p=0.002). The difference between host animals was reversed with ALTS1C1 NC and CT2A LDH-A KD tumors; tumors grew more slowly in nude compared to C57BL/6 mice. These data demonstrate the variability of the NC and LDH-A KD tumor growth in host animals with different tumor microenvironments and immune responses.
LDH isoenzyme pattern of subcutaneously located murine glioblastoma tumors LDH zymography is a common method to detect LDH isoenzymes and provides for the direct observation of 5 isozyme bands in the active state (31). The LDH isoenzyme pattern for GL261, CT2A and ALTS1C1 LDH-A KD and NC subcutaneous tumors were compared to each other and to heart and skeletal muscle tissue from the same animals (Fig. 3A, B).
All NC tumors have an isoenzyme pattern similar to skeletal muscle (LDH5 and LDH-A dominant), with some formation of LDH 4, 3, 2. CT2A and ALTS1C1 LDH-A KD tumors have a similar LDH isoenzyme pattern as the NC tumors. Only GL261 LDH-A KD tumors were strikingly different, with a LDH isoenzyme pattern similar to the heart, where LDH1, 2, 3, 4 isoenzymes are most highly expressed (40). The LDH isoenzyme ratio of the brain tissue was comparable to ratios found in the heart tissue that presents mostly isoenzymes LDH1-LDH3 and low amounts of the LDH5 isoform (40). These results are consistent with the mRNA and Western blot assays and calculated LDH-B/LDH-A ratios (Fig. 1).

LDH-A and LDH-B tumor staining patterns
First, we assessed the structure of subcutaneous GL261 NC and KD tumors growing in nude mice and for CT2A NC and KD tumors growing in C57BL/6 mice by H&E staining. There was a variable pattern of necrosis, stroma and cyst formation for both tumors (Fig. 4Aa, 4Da). Second, LDH-A and LDH-B immunohistochemistry also showed a variable pattern of staining for both types of tumors (Fig. 4Ab,c  and 4Db,c). A Weka analysis (28) showed signi cantly greater LDH-A staining in both GL261 and CT2A NC tumors compared to the LDH-A KD tumors (Fig. 4B, 4E, Table S-1), consistent with the Western blot and LDH enzyme results (Fig. 1C,E,F). Considerably greater LDH-B staining was observed in GL261 LDH-A KD compared to NC tumors (Fig 4C, Table S-1), but no differences in LDH-B staining were observed between CT2A LDH-A KD and NC tumors (Fig 4F, Table S-1), also consistent with the Western blot results (Fig.1D,E,H).

In many, but not all GL261 tumor regions, there was an inverse relationship between LDH-A and LDH-B
staining intensity (Figs. 5A and S-3A). This inverse relationship was greater for LDH-A KD than NC GL261 tumors. CT2A tumors showed a different relationship; there was a more direct relationship between LDH-A and LDH-B staining intensity (Figs. 5B and S-3B).

Early immune cell in ltration of GL261 LDH-A KD and NC s.c. tumors
To compare the extent of early (day 5) immune cell in ltration of GL261 KD and NC s.c. tumors, both GL261 KD and NC cells were implanted s.c. in the same animals (C57BL/6 mice), along with a tumor-free Matrigel plug (Fig. S-4A). The size and morphology of the NC and KD tumors were similar (Fig. S-4B).  (Fig. S-4D). Inconsistent CD4 + cell staining was observed in GL261 KD HD tumor regions. Low numbers of CD68 + cells were observed in both HD and LD tumor cell regions of both NC and KD GL261 s.c. tumors (Fig. S-4E). The results of an immuno uorescent staining analysis did not reveal any signi cant differences (Fig. S-4F) but show some trend toward an increase in immune cell in ltration in LDH-A knockdown tumors.

Discussion
Previously, we reported that downregulation of LDH-A expression in 4T1 murine breast cancer cells in vitro and in 4T1 tumors located in the mammary fat pad leads to reduced glycolytic ux and increased mitochondrial respiration, leading to slower growth and the delayed onset of (or failure to develop) distant metastases in both immune compromised mice (27) and in immune competent mice (12). We now compare three murine brain tumor models (GL261 (23), CT2A (26) and ALTS1C1 (25) to better understand the impact of LDH-A downregulation (KD) on glioma tumor phenotype and growth potential. We have studied the gliomas in two different body locations (s.c. and i.c.), and in both immune competent and immune compromised animals. In this manuscript, we compare the effects of LDH-A KD on s.c. tumors. In a subsequent manuscript, we focus on the intracranial location of the three tumors and compare the effects of genetic-shRNA LDH-A knockdown and LDH drug-targeted inhibition (41, 42) on tumor cell metabolism, tumor growth and survival time. We also nd signi cantly different effects of LDH-A knockdown on GL261 compared to CT2A and ALTS1C1 cells and tumors.
The six murine glioma cell lines (comparing LDH-A shRNA knockdown (KD) to a scrambled shRNA control (NC)) expressed different levels of LDH-A and LDH-B mRNA, protein and enzymatic activity. In all cases, the LDH-A KD cells expressed less LDH-A mRNA and protein and had less LDH enzymatic activity than the corresponding NC cell lines. Interestingly, only the GL261 LDH-A KD cell line showed a higher expression of LDH-B mRNA and protein, and the LDH-B/LDH-A mRNA and protein ratios were signi cantly higher for GL261 LDH-A KD cells compared to the other 5 cell lines (Fig. 1G, H). Also, GL261 had the lowest levels of LDH-A and LDH-B, and LDH enzyme activity of the three cell lines.
LDH is the enzyme catalyzing the nal step of glycolysis and contains two subunits A and B, encoded by two genes (43). LDH-A is predominantly found in skeletal muscle and LDH-B is predominantly expressed in the heart and brain. LDH-A and LDH-B can form homo-or hetero-tetramers forming ve LDH isoenzymes: LDH-1 (4B), LDH-2 (3B,1A), LDH-3 (2A, 2B), and LDH-5 (4A) (44). These ve isoforms catalyze the same overall reaction but differ in their a nity to the substrate, inhibition concentration (Km), isoelectric point and electrophoretic mobility. The ve isoforms can be visualized in the active state using LDH zymography (45). A zymogram analysis of the 6 glioma cell lines was performed and showed an LDH-A dominant pattern for 5 of the 6 cell lines (containing mostly LDH5, LDH4 isoenzymes) (Fig. 3).
Only GL261 LDH-A KD cells showed a LDH-B dominant pattern (containing mostly LDH1, LDH2 and some LDH3, LDH4 isoenzymes). This major shift in the LDH isoenzyme pattern in GL261 LDH-A KD tumors, (from LDH-A dominant in NC tumors to LDH-B dominant in LDH-A KD tumors), can lead to differences in the kinetics of the LDH enzyme oxidative vs reductive activity and in cell metabolism (45,46). Similar variations in LDH-A and LDH-B isoforms have been detected in human glioma cells D54MG and U-251MG (6), but this difference was not explored in detail, or related to tumor growth, metabolism, and phenotype.
A careful analysis of the mRNA and protein levels in the six experimental cell lines was performed; it also shows a difference between the levels of LDH-A and LDH-B mRNAs and demonstrates the predominance of LDH-B mRNA in GL261. This difference results in a very high LDH-B/LDH-A ratio for both mRNA and protein levels in GL261 KD cells (Fig. 1G, H). In addition, the concurrent immunoblotting of all samples (with similar amounts of loading) shows the different expression levels of LDH-A and LDH-B in GL261 cells (Fig. 1E). The dominance of LDH-B in GL261 KD cells and LDH-A in CT2A and ALTS1C1 KD and NC cells implies that different murine glioma cells can develop different isoenzyme adaptation pro les. It has been suggested that this difference may relate to the origin of the experimental cells (24).
The LDH-A and LDH-B immunohistochemistry and Weka analysis con rmed the isoenzyme patterns observed with LDH zymography and the above analyses. GL261 and CT2A KD tumors showed signi cantly less LDH-A staining than their control NC tumors. Also consistent was the signi cantly greater LDH-B staining only in GL261 LDH-A KD (compared to NC tumors); no signi cant differences in LDH-B staining was observed between CT2A LDH-A KD and NC tumors. We also noted an "inverse" LDH-A/LDH-B staining relationship (high vs low) in many, but not all GL261 tumor regions. In contrast, CT2A tumors showed a more "direct" LDH-A/LDH-B staining relationship (high vs low) in many tumor regions.
The differences in LDH isoenzyme patterns and LDH-A/LDH-B immunohistochemistry were also re ected in our other experiments. First, LDH-A KD prolonged the doubling time of GL261 cells in culture. Second, GL261 LDH-A KD cells did not establish ank tumors in immune competent C57BL/6 mice, whereas GL261 NC tumors formed after a 40-day growth delay. Third, both NC and KD GL261 tumors grew more rapidly in nude mice, compared to GL261 NC tumors growing in C57BL/6 mice. These results show the combined impact of both a metabolic alteration (LDH-A KD) and the immune system (C57BL/6 vs nude mice) on the growth of s.c. located tumors. Furthermore, the ability to grow GL261 tumors in nude mice allowed us to compare the isoenzyme pro les of LDH-A KD and NC tumors using zymogram analyses.
The association of LDH-A with cancer metabolism and tumor growth has been studied extensively, and the role of LDH-A is quite different in different tumors (9,(47)(48)(49)(50). The association of LDH-B with tumors is much more complex (15). Although LDH-B expression varied in many different tumor types (15) and is overexpressed in some tumors (51)(52)(53), LDH-A is more highly expressed in most tumors.
It has been suggested that glioma cells express one of at least two different metabolic phenotypes (6,54), as re ected in their different metabolic pro les. 251MG and U-87MG glioma cells have been shown to have metabolic characteristics akin to astrocytes (that includes the production of lactate, the storage of glycogen, and the use of lactate to support neurons). These tumors exhibit a glycolytic-dependent phenotype but retain functional oxidative phosphorylation and primarily express LDH-B. In contrast, GL261 and D-54MG glioma and SH-SY5Y neuroblastoma cells display a more oxidative phosphorylationdependent phenotype, and express both LDH-A and LDH-B isoforms. Therefore, we hypothesized that LDH-A knockdown could cause a shift in the metabolic phenotype of GL261 cells (but not CT2A and ALTS1C1 cells, which remain LDH-A dominant). In the second concurrent paper we show that GL261 tumors transition from an LDH-A dominant to an LDH-B dominant phenotype following LDH-A knockdown and treatment with a speci c LDH-A/B inhibitor (GNE-R-140), which changes the metabolic phenotype of cells. These changes in the pattern of LDH isoenzyme expression were not observed in CT2A or ALTS1C1 NC and LDH-A KD tumors. The GL261 LDH-A KD and NC cell and tumor growth pro les were also signi cantly different (as described above) and re ected the differences in their LDH isoenzyme pro les. In contrast, CT2A and ALTS1C1 LDH-A KD and NC tumors showed little or no difference in their growth pro les, consistent with the absence of a difference in their LDH isoenzyme pro les. It has been suggested that the sole expression of LDH-B might identify an important biological marker of glioma cells that is critical for their progression and that it might afford a new target for anticancer drugs (6).

Conclusions
Genetically altered murine glioma cell lines and tumors (GL261, CT2A and ALTS1C1 -with LDH-A shRNA knockdown (KD) compared to the scrambled shRNA controls -NC) showed signi cant differences in their levels of LDH-A and LDH-B mRNA, protein, and enzymatic activity. The LDH isoenzyme pro les were signi cantly different between KD and NC GL261 tumors (LDH-B vs LDH-A dominant), but not in comparable CT2A and ALTS1C1 tumors (all LDH-A dominant). These differences (or lack of differences)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformation.pdf