Citations

Abstract Background : Polycythemia Vera (PV) is associated with JAK2 mutations. In essential thrombocythemia (ET), ̴ 85% of patients harbor one of 3 "driver" mutations, with frequencies of ̴ 58%, 23% and 4%, for JAK2, CALR and MPL, respectively; ̴ 15% are wild type for all three mutations and are referred to as "triple negative". We applied next-generation sequencing (NGS) with a 27-gene panel of myeloid malignancy-relevant genes, in order to describe the prevalence of "non-driver" mutations and their prognostic relevance in PV and ET. Methods: Targeted capture assays were carried out on bone marrow or whole blood DNA for the following genes: TET2, DNMT3A, IDH1, IDH2, ASXL1, EZH2, SUZ12, SRSF2, SF3B1, ZRSR2, U2AF1, PTPN11, Tp53, SH2B3, RUNX1, CBL, NRAS, JAK2, CSF3R, FLT3, KIT, CALR, MPL, NPM1, CEBPA, IKZF, and SETBP1.Paired-end indexed libraries were prepared using the NEBNext Ultra Library prep protocol (NEB, Ipswich, MA /Agilent Technologies Inc, Santa Clara, CA). Capture libraries were assembled according to Nimblegen standard library protocol (Roche Nimblegen, Inc, Basel, Switzerland). Base-calling was performed using Illumina's RTA version 1.17.21.3. Genesifter® software (PerkinElmer, Danvers, Massachusetts) was utilized to analyze targeted sequence data. Nucleotide variants were called using the Genome Analysis Toolkit (GATK -Broad Institute, Cambridge, MA). Specific variants were deemed as mutations if they are associated with a hematologic malignancy (COSMIC database), or if they have not been associated with a dbSNP. Results: 314 patients with PV (n =133; median age 64 years, 53% females) or ET (n =181; median age 58 years, 59% females) were evaluated. Median follow-up was 9.8 years for PV and 9.2 years for ET. During this time 59 (44%) deaths, 15 (11%) fibrotic and 5 (4%) leukemic transformations were documented for PV and the corresponding percentages for ET were approximately 33%, 8% and 2%. Driver mutation distribution was 98% JAK2 for PV and 52% JAK2, 24% CALR, 20% triple-negative and 4% MPL for ET. Polycythemia Vera Mutations other than JAK2, CALR or MPL were seen in 58 (44%) patients; 29% harbored one, 14% two and one 3 mutations. None of 3 JAK2 -unmutated cases expressed non-driver mutations. Mutational frequencies were 18% TET2, 11% ASXL1, 5% SH2B3, 3% SF3B1 and ≤2% SETBP1, IDH2, DNMT3A, CEBPA, CSF3R and SUZ12, SRSF2, ZRSR2, TP53, CBL, NRAS, RUNX1, KIT, PTPN11 and FLT3. "Number of mutations" significantly affected overall (OS; Figure 1) and myelofibrosis-free (MFS) survival; respective HR (95% CI) were 2.6 (1.3-5.2) and 1.7 (0.97-3.1) and 13.7 (2.9-63.4) and 5.1 (1.5-17.6) for ≥2 mutations and one mutation, respectively. OS was also adversely affected by SRSF2 (p=0.006) and RUNX1 (p=0.04) and borderline affected by TET2 (p=0.06), IDH2 (p=0.08), ASXL1 (p=0.19) and SF3B1 (p=0.19) mutations. In multivariable analysis, SRSF2 and RUNX1 retained significance whereas the others remained borderline significant. In both univariate and multivariable analyses, leukemia-free survival (LFS) was adversely affected by IDH2 and RUNX1 mutations. In univariate analysis, ASXL1, IDH2, RUNX1 and KIT mutations predicted fibrotic progression, whereas SETBP1 was of borderline significance (p=0.07); all, including SETBP1, were significant during multivariable analysis. Essential thrombocythemia Mutations other than JAK2, CALR or MPL were seen in 83 (46%) patients; 35% harbored one, 7% two and 4% three mutations; prevalence in JAK2, CALR, MPL mutated and triple-negative cases was 52%, 43%, 43% and 35%, respectively (p=0.52). Mutational frequencies were: 13% TET2, 11% ASXL1, 6% DNMT3A, 5% SF3B1, 4% CEBPA, 2% TP53, SH2B3, EZH2 and CSF3R and <2% for SETBP1, IDH2, SRSF2, ZRSR2, CBL, NRAS, RUNX1, U2AF1, KIT, PTPN11 and FLT3. "Number of mutations" significantly affected OS (Figure 2) but not MFS or LFS; HR (95% CI) for OS were 6.6 (2.5-17.8) for 3 mutations and 2.2 (1.3-3.9) for one or two mutations. In univariate analysis, survival was adversely affected by CBL, EZH2, SF3B1, SRSF2 and IDH2 mutations. In multivariable analysis, EZH2 and SF3B1 remained significant. Univariate analysis identified SETBP1 and SF3B1 mutations as risk factors for fibrotic progression and EZH2, TP53 and CSF3R mutations for leukemic transformation. Conclusions: "Non-driver" mutations occur in more than 40% of patients with PV or ET; the number of such mutations and presence of certain specific ones likely predict OS, MFS or LFS. Figure 1. Figure 1. Figure 2. Figure 2. Disclosures Pardanani: Stemline: Research Funding.

Abstract Background : In primary myelofibrosis (PMF), ̴ 88% of patients harbor one of three "driver" mutations, with mutational frequencies of approximately 60%, 22% and 6% for JAK2, CALR and MPL, respectively. Other "non-driver" mutations have also been described in PMF and some of them and their number have been associated with inferior survival (Leukemia. 2014;28:1804). We applied next generation sequencing (NGS) with a broader panel of MPN-relevant genes, in order to identify additional mutations of prognostic relevance as well as obtain additional information regarding the prognostic value of 'number of mutations'. Methods: Targeted capture assays were carried out on bone marrow or whole blood DNA specimens obtained at time of referral for the following genes: TET2, DNMT3A, IDH1, IDH2, ASXL1, EZH2, SUZ12, SRSF2, SF3B1, ZRSR2, U2AF1, PTPN11, Tp53, SH2B3, RUNX1, CBL, NRAS, JAK2, CSF3R, FLT3, KIT, CALR, MPL, NPM1, CEBPA, IKZF, and SETBP1. Paired-end indexed libraries were prepared from individual patient DNA using the NEB Next Ultra Library prep protocol on the Agilent Bravo liquid handler (NEB, Ipswich, MA/Agilent Technologies Inc, Santa Clara, CA). Capture libraries were assembled according to Nimblegen standard library protocol (Roche Nimblegen, Inc, Basel, Switzerland). Base-calling was performed using Illumina's RTA version 1.17.21.3. Genesifter® software was utilized (PerkinElmer, Danvers, Massachusetts) to analyze targeted sequence data. Nucleotide variants were called using the Genome Analysis Toolkit (GATK-Broad Institute, Cambridge, MA). Specific variants were deemed as mutations if they are associated with a hematologic malignancy (as identified by COSMIC database), or if they have not been associated with a dbSNP. Results: 180 PMF patients were evaluated (median age 63 years; 65% males). DIPSS-plus risk distribution was 32% high, 38% intermediate-2, 17% intermediate-1 and 13% low. Driver mutation distribution was 62% JAK2, 22% CALR, 9% triple-negative and 7% MPL. Karyotype was abnormal in 41% of patients and unfavorable in 12%. Mutations other than JAK2, CALR or MPL (i.e. "non-driver" mutations) were seen in 150 (83%) patients including 88% of "triple-negative" cases. 62 (34%) patients harbored one, 55 (31%) two, 16 (9%) three and 17 (10%) four or more. Mutational frequencies were: ASXL1 36%, TET2 18%, SRSF2 17%, U2AF1 17%, ZRSR2 11%, SF3B1 10%, DNMT3A 9%, CEBPA (9%), Tp53 7%, SETBP1 6%, CBL 5%, IDH1/2 5%, SH2B3 4%, CSF3R 4%, NRAS 4%, RUNX1 3% and ≤2% for SUZ12, KIT, PTPN11, NPM1 and EZH2. DIPSS-plus high/intermediate-2 risk patients displayed higher number of mutations (p=0.0004) and higher mutational frequencies for ASXL1 (p=0.02), SRSF2 (p=0.004) and CBL (p=0.02). Associations noted included JAK2 with U2AF1 (p=0.03), unfavorable karyotype with CBL (p=0.01) and normal karyotype with ZRSR2 mutations (p=0.04). At a median follow-up of 4 years, 111 (62%) deaths were documented. For examination of impact on survival, we considered 'number of mutations' and specific mutations with >2% frequency. Accordingly, in univariate analysis, survival was adversely affected by 'number of mutations' (Figure 1) and presence of ASXL1, SRSF2, IDH1/2, U2AF1, RUNX1 and CBL mutations. For multivariable analysis, we considered three categories (zero, 1-3 and ≥4) for number of mutations based on the results from univariate analysis (Figure 1); the results showed ≥4 mutations, 1-3 mutations, RUNX1, CBL, ASXL1 and SRSF2 mutations were independently associated with shortened survival; the respective HR (95% CI) were 4 (1.4-11.1), 3 (1.3-6.8), 2.9 (1.1-8.1), 2.8 (1.3-6.3), 1.8 (1.2-2.7) AND 1.7 (1.03-2.7). When the multivariable analysis was repeated including only the 150 patients with at least one non-driver mutation, the 'number of mutations' was no longer significant (p=0.35) but ASXL1, CBL, RUNX1 and SRSF2 mutations retained their significance. The prognostic relevance of ASXL1 and CBL continued to be apparent even after the addition of DIPSS-plus and driver mutation profile to the multivariable model. Conclusions: Mutations other than JAK2, CALR or MPL occur in more than 80% of patients with PMF, including those with "triple-negative" driver mutational status. The absence of such mutations is independently favorable for survival while the prognostic effect of their presence is influenced by ASXL1, CBL, RUNX1 and SRSF2 mutations. Figure 1. Figure 1. Disclosures Pardanani: Stemline: Research Funding.

Abstract Background:Morphologic complete remission (CR) is the gold standard for assessing response in acute myelogenous leukemia (AML). CR with incomplete count recovery (CRi) and CR with incomplete platelet recovery (CRp) are associated with inferior overall survival post induction therapy (Walter et al, JCO 2010 and Chen et al, JCO 2015). Furthermore, residual cytogenetic disease is associated with a higher relapse compared with patients with cytogenetic CR (CCR; Chen et al, JCO 2011). There is a paucity of literature regarding the impact of CRi, CRp and CCR prior to allogeneic transplantation. Aim:To study the impact of CRi, CRp and CCR and their impact on relapse incidence (RI), non-relapse mortality (NRM) and overall survival (OS) in reduced intensity allogeneic stem cell transplantation (RIC-SCT). Methods:After due IRB approval, patients (pts) with acute myelogenous leukemia (AML) or myelodysplastic syndrome (MDS) who received RIC-SCT at our institution between 2008-2014 were identified. All clinical and pathologic data were retrospectively extracted. CR, CRi, CRp and CCR were defined in accordance to the international working group definition. Categorical and continuous variables were analyzed using Pearson's chi-squared and Wilcoxon/Kruskal-Wallis, respectively. Survival estimates were calculated using the kaplan-meier and log-rank test. RI was estimated on a competing event analysis using Grey's model. Cox proportional regression was used to compute hazard ratios (HR). Results: A. Baseline characteristics: A total of 134 patients were identified. A total of 70 (52%) were in CR, 40 (30%) in CRi and 24 (18%) in CRp. 129 patients had evaluable cytogenetic data pre-SCT; 81 (63%) were in CCR and 48 (37%) had persistent cytogenetic disease. Baseline patient, disease and transplant characteristics stratified by peripheral count recovery pre-SCT are shown (Table 1). Significantly more patients in CR had higher KPS, higher proportion AML diagnosis, more likely to be in first complete remission (CR1), higher rate of cytogenetic CR and a higher proportion of minor ABO mismatched donor. No other significant variables were identified between the three cohorts. B. Transplant outcomes: Median follow up of the entire cohort was 41.9 months and the 2-yr OS, RI and NRM were 61.7%, 25.7% and 21.9%. Patients in CR, CRi and CRp had a similar 2-year rates of RI (26.3%, 18.4% and 28.9%, respectively; p = 0.66), 2-year NRM (20.5%, 38.8% and 22.1%, respectively; p = 0.18) and 2-year OS (61.8%, 55.1% and 58.6%, respectively; p = 0.82). RI was significantly lower for patients in CCR pre-SCT with HR 0.25 (95% CI 0.11-0.51 p = 0.0001) without a difference in overall survival HR 0.68 (0.41-1.15; p = 0.14). However, we observed a similar RI in CR, CRi and CRp patients who had persistent cytogenetic disease pre-SCT (p = 0.55). The outcome remained similar between strata after excluding MDS patients (data not shown). Conclusion: We observed equivalent post-transplant outcomes between CR, CRi and CRp despite the presence of persistent cytogenetic disease. This suggests that RIC SCT had a role in mitigating the negative prognosis associated incomplete hematologic recovery. These findings have implications as physicians may not need to postpone SCT to await count recovery. This important question should be further validated in a larger cohort of patients. Figure 1. Figure 1. Figure 2. Figure 2. Disclosures Al-Kali: Novartis: Research Funding.

Abstract Background: Severe myelosuppression (MS) and bone marrow failure (BMF) are rare but well characterized adverse effect of temozolomide (TMZ). The existing literature is limited to case reports and data reported to regulatory agencies, which usually describe very severe cases and lack long term and detailed follow-up. Methods: We identified cases of myelosuppression attributed to TMZ seen at the Mayo clinic, from 1998-2014. We also performed a literature (PubMed) search to identify published cases of TMZ induced MS and BMF. TMZ induced MS was defined as a decrease in at least 2 cell lines below normal that was clinically attributable to TMZ and persisted for > 28 days after the last TMZ dose. Severe MS was defined as bone marrow cellularity of <25% or at least 2 of: a) reticulocyte count <40,000/µL b) absolute neutrophil count (ANC)<500/ µL c) platelets <20,000/µL. Very severe MS was defined by an ANC <200/ µL. Results: A total of 15 Mayo patients and 10 published cases were identified. The baseline characteristics of patients are shown in table 1. The median interval between initiation of TMZ and diagnosis of MS was 48 days (range 26-121). The hematologic outcomes of the group are shown in table 2. Seventeen patients developed neutropenic fever, 4 had severe sepsis, and 6 experienced bleeding complications. Kaplan-Meier recovery curves for each cell line are shown in the figure. Myelosuppression was fatal in 3 cases. Most patients recovered their ANC (median-62 days, range 28-180). Persistent thrombocytopenia was seen in 19 patients, without evidence for clonal evolution or myelodysplasia. Four patients remain dependent on growth factor support (G-CSF and thrombopoietin agonists). Conclusion: MS is usually a non-fatal complication of TMZ. Most cell lines recover spontaneously and adequately with the exception of platelets, which can remain significantly suppressed for prolonged periods. Infectious complications are very common. Most patients respond to supportive care, including the use of growth factors. Figure 3 Figure 3. Disclosures No relevant conflicts of interest to declare.

Abstract Abstract 430 Background: The Dynamic International Prognostic Scoring System (DIPSS-plus) uses eight risk factors to predict overall survival (OS) in primary myelofibrosis (PMF): unfavorable karyotype, peripheral blood (PB) blast count ≥1%, platelet count <100 × 109/L, white blood cell count (WBC) >25 × 109/L, hemoglobin level <10 g/dL, red blood cell transfusion need, constitutional symptoms, and age >65 years (JCO 2011;29:392). Among these risk factors, karyotype holds a dominant role in predicting both overall (OS) and leukemia-free (LFS) survival (Blood 2011;118:4595). More recently, certain PMF-associated somatic mutations (IDH, EZH2, ASXL1, SRSF2) have been reported (in print or at recent meetings) to carry prognostic relevance for OS or LFS in PMF while other mutations (JAK2, MPL, TET2, SF3B1) did not. The purpose of the current study was to examine the individual and combined prognostic relevance of the former in the context of cytogenetic risk stratification and DIPSS-plus. Methods: An updated Mayo Clinic database of 1008 karyotypically- and DIPSS-plus-annotated patients with PMF was used to identify study patients in whom DNA at time of referral was available for mutation analysis. OS and LFS were calculated from the time of referral, which was also the time of study sample collection. Results: In an ongoing project, a total of 397 PMF patients have so far been studied for IDH1/2 (n=367), SRSF2 (n=355), ASXL1 (n=260) and EZH2 (n=43) mutations. The corresponding mutational frequencies were 5.4% (20/367), 14.4% (51/355), 31.2% (81/260) and 7% (3/43). These mutations were not exclusive of either each other or JAK2 or MPL mutations and a significant clustering was noted between SRSF2 and IDH mutations; IDH mutational frequency was 21% (10/48) in SRSF2-mutated versus 3% (9/279) in unmutated cases (P<0.0001). Mutational hotspots were P95 in exon 1 for SRSF2 and exons 1, 2 and 5 for ASXL1. OS and LFS effect have so far been examined for ASXL1, SRSF2 and IDH 1/2 mutations; in univariate analysis, all three mutations significantly affected OS (ASXL1 RR 1.9, 95% CI 1.4–2.7; SRSF2 RR 1.7, 95% CI 1.2–2.5; IDH 1/2 RR 1.8, 95% CI 1.1–3.1) whereas only IDH 1/2 (RR 5.5, 95% CI 2.5–11.8) and SRSF2 (RR 3.5, 95% CI 1.7–7.0) but not ASXL1 (p=0.29) mutations affected LFS. The observed prognostic relevance for OS was sustained during multivariable analysis that included ASXL1 with either SRSF2 or IDH mutations, as co-variates, whereas both IDH and SRSF2 mutations remained significant for LFS. Based on the above observations, study patients were classified into two groups: i) unmutated for ASXL1, SRSF2 and IDH1/2 (all unmutated; n=146) and ii) mutated for at least one of the three mutations (one or more mutated; n=130); the latter group constituted 62%, 52%, 18% and 27% of DIPSS-plus high, intermediate-2, intermediate-1and low risk patients, respectively (p<0.0001). The corresponding rates for patients with favorable and unfavorable karyotype were 47% and 43% (p=0.7) and for normal vs. abnormal karyotype were 53% and 36% (p=0.008). The “all unmutated group”, compared to the “one or more mutated group” enjoyed a significantly superior OS (RR 0.5, 95% CI 0.4–0.7) and LFS (RR 0.3, 95% CI 0.2–0.7) that was DIPSS-plus-independent for OS (p=0.03) and karyotype-independent for LFS (p=0.004). The prognostic value of distinguishing “all unmutated” from “one or more mutated” patient groups was most evident for DIPSS-plus intermediate-2 or lower risk groups (p=0.002) and its added value for high-risk group was limited (p=0.23) (Figure). Conclusions: A concomitant analysis strategy reveals that SRSF2, ASXL1 and IDH mutations carry inter-independent prognostic relevance for overall or leukemia-free survival in PMF. Any one of these three mutations is more likely to occur in patients with normal karyotype and such an event predicts shortened survival that is not accounted for by DIPSS-plus and a higher risk of leukemic transformation that is independent of karyotype. The current study suggests that such information might be useful in treatment decisions involving intermediate-2 or lower risk patients. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 423 Background: Cytogenetic abnormalities at diagnosis are seen in 20–70% of patients with myelodysplastic syndromes (MDS). The International Prognostic Scoring System -revised (IPSS-R) classifies these cytogenetic abnormalities into five prognostic groups: very good (del(11q), ΠY with overall survival (OS) of 61 months); good (normal, del(20q), del(5q), del(12p) with OS 49 months); intermediate (+8,7q-, i(17q),+19,+21 with OS 26 months); poor (der(3)q21/q26, −7, complex with 3 abnormalities with OS 16 months); and very poor (complex with >3 abnormalities with OS of 6 months) (Schanz J, JCO 2012, Greenberg P, Blood 2012). Monosomal karyotype (MK) is defined as the presence of two or more autosomal monosomies or a single monosomy associated with a structural abnormality. We have previously shown that MK identifies a prognostically worse subgroup of MDS patients with a complex karyotype (Patnaik M, Leukemia 2010). Objectives: Methods: The Mayo Clinic database for MDS was used to identify patients in whom bone marrow histologic and cytogenetic information was obtained at the time of diagnosis. WHO criteria were used for MDS diagnosis and leukemic transformation. Results: i) Patient characteristics: A total of 1,014 patients met the above-stipulated criteria. Median age of the cohort was 71 years (range: 18–98 years) with 684 (67%) males. 819 patients (81%) had primary MDS with 195 cases (19%) of therapy-related MDS. 759 patients (75%) are dead at a median follow-up of 18 months (range: 0–223 months) with leukemic transformation documented in 125 cases (12%). IPSS-R risk distribution was as follows: very low (n=149, 15%), good (n=261, 26%), intermediate (n=221, 22%), high (n=131, 13%), and very high (n=252, 25%). ii) IPSS-R karyotype (Entire cohort; n=1,014): The five cytogenetic groups by IPSS-R were as follows: very good (n=43, 4%), good (n=577, 57%), intermediate (n=187, 18%), poor (n=83, 8%), and very poor (n=124, 12%). We evaluated the prognostic impact of the above groups by Kaplan Meier and Cox proportional hazards analysis. OS for the five cytogenetic groups was as follows: good (36 months), very good (28 months), intermediate (20 months), poor (13 months), and very poor (6 months) (Figure 1). OS of patients with very good karyotype was significantly better than patients with very poor and poor karyotype (p<0.0001, RR 3.2 and p=0.04, RR 1.6 respectively) but similar to patients with intermediate and good karyotype (p=0.22 and 0.28 respectively). iii) IPSS-R karyotype (Primary MDS; n=819): Cytogenetic risk groups by IPSS-R were as follows: very good (n=38, 5%), good (n=512, 63%), intermediate (n=153, 19%), poor (n=46, 6%), and very poor (n=70, 9%). OS of the five groups was as follows: good (39 months), very good (23 months), intermediate (21.5 months), poor (14.5 months), and very poor (6.5 months). OS of patients with very good karyotype was significantly better than patients with very poor karyotype (p=0.0002, RR 2.5) but similar to patients with poor, intermediate and good karyotype (p=0.23, 0.26 and 0.18 respectively). iv) Monosomal karyotype: MK was identified in 153 (15%) cases of which 4 cases were intermediate, 42 poor, and 107 very poor by IPSS-R. OS of patients with MK was similar to those with very poor karyotype (6 months). In addition, OS of patients with very poor without MK was akin to patients with poor without MK (p=0.21). Furthermore, among primary MDS we identified 83 (8%) cases of MK of which 4 were intermediate, 21 were poor, and 58 very poor by IPSS-R. We confirmed that OS of patients with MK was similar to patients with very poor karyotype (6.5 months). Moreover, survival of patients with very poor without MK was similar to patients with poor without MK (p=0.51). Conclusions: In this independent analysis of 1,014 patients with MDS, IPSS-R cytogenetic risk characterization failed to discriminate between very good, good and intermediate prognostic groups. In addition, patients with very poor without MK have a similar prognosis to patients with poor without MK and should be included in the poor category whereas the presence of MK should warrant inclusion into the very poor risk category. Disclosures: No relevant conflicts of interest to declare.

Abstract Purpose: Myelofibrosis is characterized by bone marrow fibrosis, atypical megakaryocytes, splenomegaly, constitutional symptoms, thrombotic and hemorrhagic complications, and a risk of evolution to acute leukemia. The JAK kinase inhibitor ruxolitinib provides therapeutic benefit, but the effects are limited. The purpose of this study was to determine whether targeting AURKA, which has been shown to increase maturation of atypical megakaryocytes, has potential benefit for patients with myelofibrosis. Patients and Methods: Twenty-four patients with myelofibrosis were enrolled in a phase I study at three centers. The objective of the study was to evaluate the safety and preliminary efficacy of alisertib. Correlative studies involved assessment of the effect of alisertib on the megakaryocyte lineage, allele burden, and fibrosis. Results: In addition to being well tolerated, alisertib reduced splenomegaly and symptom burden in 29% and 32% of patients, respectively, despite not consistently reducing the degree of inflammatory cytokines. Moreover, alisertib normalized megakaryocytes and reduced fibrosis in 5 of 7 patients for whom sequential marrows were available. Alisertib also decreased the mutant allele burden in a subset of patients. Conclusions: Given the limitations of ruxolitinib, novel therapies are needed for myelofibrosis. In this study, alisertib provided clinical benefit and exhibited the expected on-target effect on the megakaryocyte lineage, resulting in normalization of these cells and reduced fibrosis in the majority of patients for which sequential marrows were available. Thus, AURKA inhibition should be further developed as a therapeutic option in myelofibrosis. See related commentary by Piszczatowski and Steidl, p. 4868