Tumor microenviromental reaction would diminish incomparable accuracy and swiftness of ow cytometric diagnosis of primary central nervous system lymphoma

Primary central nervous system lymphoma (PCNSL), a relatively rare brain tumor, bears a dire prognosis. On occasion, rapid progression of the tumor makes immediate diagnosis and initiation of therapy imperative. To achieve swift diagnosis, we adopt ow cytometry (FCM) in addition to conventional histopathology. The aim of this study is to reveal utility and drawbacks of FCM diagnosis for PCNSL. Patients with suspected PCNSL on neuroradiological ndings and received both FCM and histological diagnosis between August 2015 and April 2020 were retrospectively enrolled into the study. Tumor samples were collected by craniotomy with either of endoscope or microscope. The patients’ electronic medical records were investigated, and histological ndings, results of FCM, and other clinical data were evaluated. Twenty T-cell analysis in the other 3 4 non-lymphomatous FCM results did not show pattern Analysis FCM-D cases showed inltration T or into indicating tumor microenvironmental reaction, that those reactions deceived FCM diagnosis.

Introduction Primary central nervous system lymphoma (PCNSL) accounts for 1.9% among all brain tumors in the U.S.
[2] The prognosis of PCNSL is still poor, with 5-year survival rate is less than 50%. [3,4] In some cases, acute aggravation of the symptom due to rapid enlargement of tumor is observed, and swift diagnosis for early initiation of treatment is necessary. At this moment, histopathology of para n-embedded specimen is the diagnostic modality of choice, and this method usually takes several days for formalin-xation, preparation, hematoxylin and eosin (HE) staining and immunohistochemistry for diagnosis. Flow cytometry (FCM) plays a well-accepted role in diagnosis and evaluation of hematopoietic neoplasms including extracranial lymphoid malignancies. [5,6] FCM takes only a few hours for analysis. However, FCM in the diagnosis of PCNSL have not gained widespread popularity and only a few reports concerning this modality have been published so far. [7,8,9] We have experienced cases of PCNSLs in which surgical specimen were provided both to histopathology and FCM, and in most cases early commencement of therapy was succeeded owing to FCM diagnosis by the next day of operation. Yet, in small number of the cases, FCM failed to detect PCNSL. We have investigated the mechanism of FCM de ciency, and revealed effects of tumor microenviromental reaction such as in ltration of astrocytes or T-lymphocytes. We present the result with review of relevant literatures.

Patients
This retrospective study protocol was completed following approval from the institutional review board at Kyoto Prefectural University of Medicine. We collected data from consecutive patients with PCNSL and other brain tumors who received FCM analysis of tumor between August 2015 and April 2020. Patients were included if clinical and neuroloradiological ndings suggested lymphoma. Patients with previous history of malignancy was excluded. Clinical data were collected such as age, sex, previous use of medication including corticosteroids, blood examination including soluble interleukin-2 receptor (sIL-2R), neuroradiological ndings, and histological classi cation of the tumor.

Surgical procedure
All the patients underwent craniotomy for removal of the tumor. Three patients (Cases 5, 7, 8) received small craniotomy for endoscopic biopsy by a 4mm steerable berscope (Olympus Optical Company, Tokyo, Japan), and the rest of the patients received craniotomy and tumor removal with operative microscope (M530 OH6, Leica Microsystems GmBH, Wetzlar, Germany) equipped with Leica FL 400 (blue light uorescence module). Microscopic surgery was assisted by neuronavigation system (Stealth Station S7, Medtronic, Minneapolis, MN), and enhanced lesions were the target. After craniotomy, small corticotomy was made. Then, either of Neuroport (Olympus, Tokyo, Japan) or Viewsite (Vicor Medical, Boca Raton, FL) was inserted. Tumor site was con rmed by both neuronavigation system and direct observation by the surgeon. A few pieces of the lesion were collected with biopsy forceps and were served for intraoperative frozen section. If the results suggested neoplastic features (e.g., increased cellularity, nuclear atypia), then additional tissue samples were taken. In some cases, when brain tissue bulges, bulk removal of the tumor was added to lessen intracranial pressure. Those tumor samples are divided into half, and provided for ow cytometry and formalin-xation.
When malignant glioma was also suspected prior to surgery, 5-aminolevulinic acid (5-ALA; 20 mg/kg body weight) was administered orally to the patient 3 hours prior to the induction of anesthesia. During the operation, violet-red uorescence visualization of malignant tumor tissue after excitation with 405 nm wavelength blue light was inspected. Strength of uorescence was checked by the surgeon's objective observation. According to the ndings "No", "Vague", or "Strong" uorescent was recorded.

Flow Cytometry and Histopathological Diagnosis
Un xed tumor tissue samples were disaggregated gently, ltered with a 70-µm nylon cell strainer (Corning, Corning, NY), then washed and resuspended with phosphate-buffered saline (PBS; Takara, Shiga, Japan) containing 2% bovine albumin (BSA; Nacalai Tesque, Kyoto, Japan) and 0.1% sodium azide (Nacalai Tesque) at a nal cell concentration between 0.5-10 x 10 3 /µL. According to the manufacturer's recommendation, 100µL of this cell suspension was stained with the combination of the following monoclonal antibodies conjugated with uorescence isothiocyanate ( and HLA-DR (Immu-357); all of them were purchased from Beckmann-Coulter (Marseille, France). On the other hand, anti-surface immunoglobulin (Sm) kappa light chain-FITC and Sm lambda light chain-PE were purchased from Dako/Agilient (Santa Clara, CA). FITC or PE-conjugated mouse IgG1 (Beckmann-Coulter) were used as negative control. After incubation for 15 minutes at room temperature in complete darkness, samples were washed again and resuspended at a nal volume of 0.5 mL. Then, samples were analyzed using NAVIOS ow cytometer and Kaluza software (Beckmann-Coulter). Maximally 5,000 events were gated on forward scatter (FS) and side scatter (SS) plots. When at least 20% of gated cells expressed CD19 and/or CD20 together with restricted Sm light chain expression (kappa/lambda ratio is > 3.0 or < 0.5), samples were diagnosed as B-cell lymphoma. Concerning the diagnosis of T-NHL, aberrant expression of T-cell antigens such as CD2, CD3, CD5 or CD7 was necessary in addition to deviation of CD4/CD8 ratio (> 3.5 or < 1.0).
The pathologist (Y.S.) and hematologist (T.I.) who evaluated the analysis were blinded for each other's conclusion. Subtype classi cation of DLBCL, the germinal center B-like (GCB) and activated B-cell (ABC) types were done according to the immunostaining of CD10, BCL6, and MUM-1. [10] Statistical Analysis The data were expressed as mean ± SD. Statistical analysis was performed by using Student's t test (twotailed), or Fisher's exact test. The criterion for statistical signi cance was taken as P-values below 0.05.
Statistical analysis was performed using EZR software which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria). [11] Results We identi ed 27 consecutive patients who received both FCM analysis and histopathological diagnosis for brain tumor during the observation period. Patient demographics were summarized in Table 1. No patient was positive for human immunode ciency virus (HIV), Epstein-Barr virus (EBV), or human T-cell leukemia virus type 1 (HTLV-1). None of the patients except for two (case11, 21) received corticosteroids prior to operation. One (case 11) was given 2 days of betamethasone (8 mg/day) on the day of admission to the referring hospital. The other patient (case 21) was given repeated dose of prednisolone as much as 50 mg/day for 2 months because her disease was at rst assumed as neuro-Behçet's disease based on neuroradiological ndings. In both cases, operation was done more than 1 week later from the last administration of steroid.  Fig. 1A. The 4 patients who were not diagnosed with PCNSL, 3 were glioblastoma, 1 was diffuse midline glioma H3K27M mutated, and for all of them FCM also did not show lymphoma pattern (speci city: 100%, Table 2).
After con rmation of the diagnosis as PCNSL, patients were administered chemotherapeutic medications mainly consisted of high-dose methotrexate, procarbazine, and vincristine. In addition, rituximab was added in all cases when the tumor cells were positive for CD20. The clinical courses of all of those patients are compatible with PCNSL, hence diagnostic accuracy of histopathology was con rmed clinically also. In FCM-C group, median days from surgery to start chemotherapy was 5 days, which were shorter than FCM-D group (24 days), but not reached statistical signi cance (p = 0.057).
Next, to investigate tumor characteristics that would in uence FCM concordance, we performed univariate analysis. Univariate analysis revealed only patient age was signi cantly elder in FCM-D, but there was no correlation between FCM-D and patient sex, previous use of steroid, sIL-2R value, subtype (GCB/ non-GCB) or 5-ALA uorescence (Table 3). Then we inquired each 3 cases histologically. In case 14, FCM revealed that more than 70% of analyzed cells were T-cells without de nite aberrant antigen expression, and B cells were less about 20% of analyzed cells without de nite deviation of Sm kappa/lambda ratio (16.8/7.5 = 2.24) (Fig. 1B). By histopathological analysis, uniformly proliferating large cells with round-to-oval nuclei indicating lymphomatic change were observed (Fig. 2F). In addition, CD20 positive cells (Fig. 2G) and bcl-6 positive cells (Fig. 2I) made the diagnosis as diffuse large cell lymphoma. In that specimen, T lymphocytes (Fig. 2H) did were observed, but by integrating the ndings together, the diagnosis as diffuse large B-cell lymphoma was made. In case 17, histopathological analysis indicated co-existence of high cellularity portion and normo-cellularity portion (Fig. 2K-O). In the high cellularity portion, typical morphology of lymphoma was observed, but in the normocellularity portion in ltration of T cells were detected. The pathological diagnosis was according to the high cellularity portion, so that this case was determined as DLBCL. However, by FCM analysis, mixture of CD3 and CD5 positive T lymphocytes has lead the diagnosis indeterminate. In case 18, FCM was inconclusive because about 15% of analyzed cells were B-lymphocytes without de nite deviation of Sm kappa/lambda ratio. Histological ndings showed that the same specimen revealed mixture of lymphoma cells and reactive astrocytes (Fig. 2P-T). Compare to those, in FCM-C cases, in ltration of Tlymphocytes or astrocytes were minimal ( Fig. 2A-E). In summary, we have determined that invasion of reactive cells such as astrocytes or T lymphocytes, and low volume of tumor cells among tissue were the cause of FCM-D.

Discussion
In this study, we have demonstrated the effectiveness of FCM in the diagnosis of PCNSL. It not only is swift but also has high sensitivity and speci city. Early diagnosis of CNS lymphoma has already reported to prolongs survival. [12] The time interval between operation and beginning of chemotherapy of FCM-C group was shorter than that of FCM-D group, although this difference did not reach statistical signi cance. By a through literature search, we have found 3 reports mentioning FCM for the diagnosis of PCNSL. [5,6,7] According to them, sensitivity and speci city of FCM are 87.8-100%, and 100%, respectively. Our results also support that fact.
In our series, 3 out of 22 cases were FCM-D, but no patient background factors except for age were signi cantly different between FCM-C and FCM-D groups. We think this was because of small sample size. Instead, by histological analysis, we have disclosed one of the important reasons of discordance. As shown in case 14 and 18, tumor microenvironment (TME) interaction such as in ltration of reactive Tlymphocytes and/or astrocytes were observed. In previously reported cases of FCM diagnosis, 10 out of 77 PCNSL cases were FCM-D. Among those 10 cases, authors guessed the reason as steroid use in 2, patchy in ltrate caused by DLBCL in 1, intravascular DLBCL in 1, and unknown in 6 cases. The "patchy in ltrate caused by DLBCL" seems corresponding to the TME reaction. By recent investigations, role of TME became clear. A certain portion of PCNSL has high amount of tumor in ltrating lymphocytes (TILs), astrocytes, and microglia, [13] and such TME reaction may be related to immune checkpoints. [14] Because those TILs, astrocytes or microglias do not have monoclonality, FCM that analyze sample cells as a whole, cannot detect neoplastic nature in such situation. Histopathological diagnosis is made based on the high-cellularity portion among the specimen by manual search, so that it is not much di cult to distinguish neoplastic portion and in ltrating T-lymphocytes or astrocytes. In summary, stronger reaction of TME would be the noticeable reason of discordance.
As for steroid use, in our cases, it did not cause FCM-D. One reason of this, we presume, is the time interval between nal steroid use and surgery was long enough for the tumor cells to re-activate, and the other is employment of open biopsy. Effect of steroid has been shown to necrotize lymphoma cells, [15] while in a recent report, even under the in uence of steroid correct lymphoma diagnosis could be made.
[16] Histopathologically, in our steroid pretreated cases, some necrotic areas were observed, but area of viable cells were dominant. Open biopsy can collect tumor tissue by direct visualization. Weaknesses of stereotactic biopsy are mis-sampling and hemorrhagic complications. In one report, 6.2% of the stereotactic biopsy cases were non diagnostic, instead symptomatic hemorrhage was observed in 2.6%, and asymptomatic hemorrhage occurred in as many as 28.8%. [17] In comparison, with open biopsy, surgeon can con rm color and hardness of the tissue. Another advantage is that volume reduction is achievable. Though no bene t of volume reduction in PCNSL was believed until recently, usefulness of craniotomy was reported. [18,19] Another possible method to diagnose PCNSL is to analyze cerebrospinal uid (CSF) or using FCM for distinguishing PCNSL from GBM by propidium iodide for staining DNA. The former, in spite of relatively wide acceptance, CSF cytology suffers from its low sensitivity (2-32%). [20] Recently, diagnostic utility of cell free circulating tumor DNA in CSF has been reported. [21] Still, this method is not yet established and time consuming. Besides, collecting CSF in the situation of elevated intracranial pressure is quite dangerous and not recommended. The latter, as reported by Koriyama and colleagues in 2018, distinguishing between PCNSL and GBM by DNA aneuploidy or S phase is also useful, but that method, not checking surface markers, cannot analyze lymphoma subtype. [22] Compare to those, FCM analysis of surface marker can provide useful information to treat lymphoma.
A potential limitation of this study lies in its retrospective nature. However, we believe there does not exist selection bias because we executed FCM in all eligible patients. Another potential limitation of our study is the small sample size. We encountered 23 PCNSL patients, and FCM-D cases were only 3. That made much of statistical analysis no signi cance, therefore larger study to con rm this result is needed. In addition, we recognize that the equipment performance of ow cytometer seems also important to improve the diagnostic utility for PCNSL. We analyzed the samples using the traditional 2-3 color FCM with 1 laser throughout the course of this study, but we are now in the process of updating to 10-color FCM using 3 lasers.
In conclusion, FCM is a powerful tool for the diagnosis of PCNSL, but there are very few false negative cases: in ltration of reactive immune cells or glia, or necrosis are the possible mechanisms.

Declarations
Funding: None Con ict of interest: None of the authors have any con ict of interest to declare.
Availability of data and material: Not applicable Code availability: Not applicable Ethical approval: This research study was conducted retrospectively from data obtained for clinical purpose. This study was granted IRB approval from Kyoto Prefectural University of Medicine. Figure 1 Examples of ow cytometric analysis which resulted in positive (A: Case 12) and negative (B: Case 14) lymphoma pattern, both of which afterwards were histopathologically diagnosed as diffuse large B cell type lymphoma. In A, FCM results showed positive for CD20, but negative for CD3 and CD10. In addition, IgM-kappa, and -lambda was clearly divided. Those were indicating lymphoma. In B, CD45 (leukocyte common antigen) was positive, but B-cell surface markers (CD19, CD20) were negative. Therefore, FCM analysis was inconclusive.