An overview of the study cohort is presented in Fig. 1, including clinical characteristics, immunohistochemistry and summarized molecular data. Immunohistochemistry for VDAC1 (to assess mitochondrial density) and OXPHOS-subunits was successful on 43 tumors/adenomas, comprising 20 PIT1-lineage, 13 SF1 lineage, 9 TPIT-lineage, and one multilineage tumor/adenoma (SF1/PIT1-lineages). Additional mtDNA analysis was successful in 21 cases, of which nine harboured mtDNA mutations (Table 1).

Fig. 1

Overview of the study cohort. Selected clinical characteristics of the patients, PitNET/adenoma subtypes (grouped by lineage), chromosomal alteration patterns, immunohistochemistry for porin (VDAC1) and subunits of complexes I-V, and mtDNA mutation status (where available) are shown. Heterogeneous/patchy immunohistochemical staining was observed across all lineages and stainings. No mtDNA mutations were observed in tumors/adenomas of PIT1-lineage. In contrast, the majority of SF1-lineage tumors/adenomas harboured disruptive mtDNA mutations in complex I, which were associated with reduced immunohistochemical expression of NDUFB8 (complex I subunit) *in %, compared to normal adenohypophyseal tissue

Mitochondrial density

Mitochondrial density, assessed by porin (VDAC1) expression, was significantly higher in tumor/adenoma tissue (162.3 ± 29.2) compared with normal adenohypophyseal tissue (111.1 ± 22.1; p < 0.0001). Thirty-eight of 43 tumors/adenomas (88.4%) showed mean intratumoral VDAC1 staining intensity above the normal range.

Heterogenous/patchy staining

Of 43 tumors/adenomas, 24 showed heterogenous/patchy staining pattern (Fig. 2), whereas 17 showed homogenous staining. Two samples were too small to assess heterogeneity (1xSF1 and 1xTPIT-lineage).

Among 24 tumors/adenomas with heterogeneous staining, 23 were patchy in 2 or more stainings, and four tumors/adenomas demonstrated a heterogeneous pattern in all six stainings. Heterogenous staining was most frequently observed for SDHA (CII, 19/24), followed by NDUFB8 (CI, 17/24), ATP5F1A (CV, 16/24), VDAC1 (porin, 15/24), MT-CO1 (CIV, 9/24), and UQCRC2 (CIII, 5/24).

Staining heterogeneity was most prevalent in SF1-lineage tumors/adenomas (11/12). In contrast, approximately only half of TPIT-lineage tumors/adenomas (3/8) and PIT1-lineage tumors/adenomas (11/23) showed heterogeneous staining. Moreover, tumors/adenomas of PIT1-lineage did not show heterogeneity in UQCRC2 (CIII) and MT-CO1 (CIV) stainings.

Fig. 2

Examples of heterogenous/patchy intratumoral staining (a) Case 34 (b) (SF1-lineage), m.11038del (91%) detected (c) Case 41 (d) (SF1-lineage), no mtDNA mutations detected. Magnification x10

Immunohistochemical expression of CI-CV subunits

An overview of immunohistochemical staining intensities per case is provided in Fig. 1 and summarized by lineage in Table 1.

Table 1 Summary of immunohistochemical staining patterns per lineage

Decreased NDUFB8 (CI) expression was observed most frequently (n = 12), occurring either in isolation (n = 8) or in combination with decreased expression of other complexes (4). Decreased NDUFB8-expression was predominantly observed in tumors/adenomas harbouring mtDNA mutations (n = 8) but was also observed in two tumors/adenomas without detectable mtDNA mutations (e.g., Fig. 2b).

Decreased UQCRC2 (CIII) and MT-CO1 (CIV) occurred exclusively in combination with decreased NDUFB8-expression. Decreased ATP5F1A (CV) expression was observed in a single tumor/adenoma and was also a part of a multi-complex deficiency.

Increased expression of NDUFB8 or MT-CO1 was not observed.

In multiple linear regression analysis, SDHA (p = 0.004) and ATP5F1A expression (p = 0.03) were independently associated with higher VDAC1 expression, reflecting increased mitochondrial density. In contrast, NDUFB8 (p = 0.02) and MT-CO1 (p = 0.02) showed inverse associations.

Lineage-specific expression patterns and the overall distribution of immunohistochemical staining intensities are illustrated in Fig. 3.

Fig. 3

Immunohistochemical staining intensity ratios for subunits of respiratory complexes I–V and VDAC1 in pituitary adenomas/PitNETs (a)Tumor-to-normal staining intensity ratios stratified by lineage. The dashed line at 1.0 indicates parity with normal tissue. A lineage-specific trend toward reduced NDUFB8 expression (CI) is observed in SF1-lineage tumors/adenomas, whereas VDAC1 expression, reflecting mitochondrial density, is increased across all lineages (b) Intratumoral staining heterogeneity in seven tumors/adenomas harbouring mtDNA mutations affecting complex I. Ratios represent staining intensity in regions corresponding to the heteroplasmic mutation relative to other regions within the same tumor/adenoma

mtDNA sequencing

mtDNA mutations were detected in nine of 21 tumors/adenomas in which mtDNA sequencing was successful (Table 1). Two novel missense and one novel loss-of-function variant were identified.

The MT-ND1 m.3631T > C variant results in substitution of a highly conserved serine residue within a transmembrane domain (p.Ser109Pro), spanning amino acid residues 100–120. The variant was absent from gnomAD 4.1 and MitoMap. According to ACMG criteria, it was classified as a variant of uncertain significance (PM2 moderate, PP3 supporting). Multiple in silico prediction tools support a pathogenic effect (Apogee2, Pathogenic, 0.87; Hmtvar, Pathogenic, 0.84; AlphaMissense, Pathogenic, 0.90; BayesDel_addAF; Uncertain; 0.057, T; DEOGEN2, Benign, 0.23, T; LIST_S2, Uncertain, 0.95, D; MutationAssessor, Pathogenic, 4.3, H; PhyloP100, 4.8; PROVEAN, Pathogenic, -4.6, D; Sift4G, Pathogenic, 0.0010, D; GERP RS, 4.5; Varity_R, 0.95).

The MT-ND4 m.11484G > A variant was detected in one tumor/adenoma and affects a highly conserved glycine residue within a transmembrane domain (p.Gly242Asp), spanning residues 224–244. This variant has been reported at relatively low heteroplasmy (25%) in one individual in gnomAD but was absent in gnomAD 4.1 and MitoMap. In the tumor/adenoma sample, heteroplasmy reached 84%. According to ACMG criteria, this variant was also classified as of uncertain of uncertain significance (PM2 moderate, PP3 supporting), with the majority of in silico tools (7/11) predicting a damaging or pathogenic effect (Apogee2, Pathogenic, 0.74; Hmtvar, Pathogenic, 0.88; AlphaMissense, Pathogenic, 1.0; BayesDel_addAF, Benign, -0.19, T; DEOGEN2, Uncertain, 0.44, T; LIST_S2, Uncertain, 0.93, D; MutationAssessor; Pathogenic, 5.2, H; PhyloP100, 9.4; PROVEAN, Pathogenic, -6.0, D; Sift, Pathogenic, 0.0, D; Sift4G, Pathogenic, 0.0, D; GERP RS, 5.1; Varity_R, 0.97).

The novel loss-of-function variant in MT-ND2 (m.5366_5367del; c.896_897del) causes a frameshift leading to a premature stop codon (p.Ser299TyrfsTer10). The variant was absent from gnomAD and MitoMap.

The remaining pathogenic mtDNA variants identified in this cohort had been previously reported (Table 1). An (Table 2) overview of all 161 detected mtDNA variants across the 21 PitNETs/adenomas is provided in Supplemental Material.

Table 2 All detected mtDNA mutations

All tumors/adenomas carrying mtDNA mutations showed decreased regional NDUFB8-expression accompanied by a complementary increase in mitochondrial density (Supplemental Fig. 1). Intratumoral expression variability in tumors/adenomas harbouring mtDNA mutations is visualised in Fig. 3B.

mtDNA mutation status and chromosomal alterations

Among the nine tumors/adenomas harbouring mtDNA mutations, six (all SF1-lineage) showed stable genomes without chromosomal alterations. The remaining three showed disputed genomes: two (TPIT-lineage) showed near-haploid genomes with loss of heterozygosity (LOH) affecting multiple chromosomes due to whole chromosome loss with expected or detected endoreduplication, and one (SF1-lineage) showed chromosomal imbalances due to copy number (CN) gain.

No mtDNA mutations were detected in 12 tumors/adenomas. Among these, four exhibited stable genomes (1xPIT1-lineage, 3xSF1-lineage), while eight demonstrated disrupted genomes. Of the latter, three showed imbalances due to CN loss or LOH (2xPIT1-lineage, 1xTPIT-lineage), two showed imbalances due to CN gain (1xPIT1-lienage, 1xSF1-lineage), and three (3xPIT1-lineage) demonstrated mixed pattern with combined CN gains and losses.

Correlations of clinical parameters with mtDNA mutation status and mitochondrial protein expression

No significant associations were observed between mtDNA mutation status and clinical parameters. A non-significant trend toward higher mutation frequency with increasing age was noted (OR 1.04 per year, p = 0.2). mtDNA mutation status was not associated with sex (p = 0.66), invasiveness (p = 0.61), functional status (p = 0.18), or repeated surgery (p = 0.40).

In multivariable logistic regression analysis including mitochondrial density (VDAC1) and complex I (NDUFB8) expression, lower CI expression was independently associated with the presence of mtDNA mutations (p = 0.03), whereas VDAC1 expression was not (p = 0.36). Consistently, CI expression was significantly lower in mtDNA-mutated tumors in univariate analysis (p = 0.003). A trend toward increased mitochondrial density in mtDNA-mutated tumors was observed but did not reach statistical significance (p = 0.067).

Neither VDAC1 nor NDUFB-expression was significantly associated with invasiveness (p = 0.50 and p = 0.11, respectively) or repeated surgery (p = 0.44 and p = 0.47, respectively). NDUFB8-expression showed a weak association with sex (higher expression slightly reduced odds of being male, p = 0.028), and a modest positive association with functional status (p = 0.40, OR = 1.01 per unit increase), although the overall regression model for functional status did not reach statistical significance.

Oncocytic phenotype

Oncocytic morphology was assessed on H&E sections using a three-tier classification. Interobserver agreement was limited (see Methods), and discrepant cases were resolved by consensus.

No significant association was observed between oncocytic morphology and mitochondrial density as assessed by VDAC1 expression. Lineage-specific analyses likewise revealed no consistent association.

Similarly, mtDNA mutation status was not associated with oncocytic morphology or mitochondrial density.

Genome stability was also not associated with oncocytic morphology or mitochondrial density.

Representative cases are shown in Supplemental Fig. 2 to illustrate the lack of concordance between oncocytic appearance and VDAC1 staining intensity. Both morphologically non-oncocytic and oncocytic tumors demonstrated variable mitochondrial density, indicating that routine H&E-defined oncocytic features are not a reliable surrogate for mitochondrial mass in this cohort.