**Artesunate Inhibits Growth of Sunitinib-Resistant Renal Cell Carcinoma Cells through Cell Cycle Arrest and Induction of Ferroptosis**

**Sascha D. Markowitsch <sup>1</sup> , Patricia Schupp <sup>1</sup> , Julia Lauckner <sup>1</sup> , Olesya Vakhrusheva <sup>1</sup> , Kimberly S. Slade <sup>1</sup> , René Mager <sup>1</sup> , Thomas E**ff**erth <sup>2</sup> , Axel Haferkamp <sup>1</sup> and Eva Juengel 1,\***


Received: 30 September 2020; Accepted: 24 October 2020; Published: 27 October 2020

**Simple Summary:** Renal cell carcinoma (RCC) is the most common kidney malignancy. Due to development of therapy resistance, efficacy of conventional drugs such as sunitinib is limited. Artesunate (ART), a drug originating from Traditional Chinese Medicine, has exhibited anti-tumor effects in several non-urologic tumors. ART inhibited growth, reduced metastatic properties, and curtailed metabolism in sunitinib-sensitive and sunitinib–resistant RCC cells. In three of four tested cell lines, ART's growth inhibitory effects were accompanied by cell cycle arrest and modulation of cell cycle regulating proteins. In a fourth cell line, KTCTL-26, ART evoked ferroptosis, an iron-dependent cell death, and exhibited stronger anti-tumor effects than in the other cell lines. The regulatory protein, p53, was only detectable in the KTCTL-26 cells, possibly making p53 a predictive marker of cancer that may respond better to ART. ART, therefore, may hold promise as an additive therapy option for selected patients with advanced or therapy-resistant RCC.

**Abstract:** Although innovative therapeutic concepts have led to better treatment of advanced renal cell carcinoma (RCC), efficacy is still limited due to the tumor developing resistance to applied drugs. Artesunate (ART) has demonstrated anti-tumor effects in different tumor entities. This study was designed to investigate the impact of ART (1–100 µM) on the sunitinib-resistant RCC cell lines, Caki-1, 786-O, KTCTL26, and A-498. Therapy-sensitive (parental) and untreated cells served as controls. ART's impact on tumor cell growth, proliferation, clonogenic growth, apoptosis, necrosis, ferroptosis, and metabolic activity was evaluated. Cell cycle distribution, the expression of cell cycle regulating proteins, p53, and the occurrence of reactive oxygen species (ROS) were investigated. ART significantly increased cytotoxicity and inhibited proliferation and clonogenic growth in both parental and sunitinib-resistant RCC cells. In Caki-1, 786-O, and A-498 cell lines growth inhibition was associated with G0/G1 phase arrest and distinct modulation of cell cycle regulating proteins. KTCTL-26 cells were mainly affected by ART through ROS generation, ferroptosis, and decreased metabolism. p53 exclusively appeared in the KTCTL-26 cells, indicating that p53 might be predictive for ART-dependent ferroptosis. Thus, ART may hold promise for treating selected patients with advanced and even therapy-resistant RCC.

**Keywords:** renal cell carcinoma (RCC); sunitib resistance; artesunate (ART); Traditional Chinese Medicine (TCM); growth inhibition; ferroptosis; reactive oxygen species (ROS)

#### **1. Introduction**

Accounting for ~85% of cases, renal cell carcinoma (RCC) is the most common kidney cancer and one of the most aggressive urologic cancers [1]. At initial diagnosis, RCC patients often present at an advanced stage with an accordingly poor prognosis [2]. Better understanding of the molecular modes of action underlying RCC led to the development of targeted therapies affecting angiogenic activity and immune checkpoint inhibitors. However, due to the development of resistance, the efficacy of even these targeted treatments is limited. Since RCC is an angiogenic disease, a promising avenue of treatment is to block angiogenesis, thereby suppressing the supply of oxygen and nutrients to the tumor. Initially, the anti-angiogenic activity of the tyrosine kinase inhibitor (TKI) sunitinib extends the progression-free survival of patients [3], but resistance occurs during treatment [4]. Thus, therapy resistance is one, if not the main, problem, in treating advanced RCC. Novel treatment strategies combining targeted therapy and immunotherapy have been introduced [5–7]. However, even with combined drug application resistance occurs and adverse side effects are common [5,8,9].

Certainly, in part due to the long-term curative failure of conventional therapy, the demand for traditional and alternative medicine is growing worldwide [10,11]. Patients hope that complementary therapeutic approaches would increase effectiveness and/or reduce side effects [12,13], and 40–50% of European cancer patients indeed use complementary and alternative therapies [14–16]. However, solid and reliable studies with regard to natural substances and their derivatives are sparse and the lack of proven efficacy coupled with uncoordinated self-treatment is perilous. Contraindications as well as adverse side effects of herbal compounds combined with conventional therapy also cannot be ignored [17].

Some studies have been carried out indicating anti-tumor effects of natural compounds and their derivatives, especially if applied together with established therapies or by counteracting therapy resistance [18–21]. Artemisinin from the annual mugwort (*Artemisia annua*) has been used in Traditional Chinese Medicine for over 2000 years, particularly in treating malaria [22] and is still in use. An anti-tumor effect of the artemisinin derivative, artesunate (ART), was reported in 2001 [23]. Subsequently, anti-tumor effects of ART in vitro and in vivo were reported in different tumor entities, including therapy-resistant tumors, with fewer side effects in combination with conventional therapy [23–27]. ART, a semi-synthetic water-soluble derivative of artemisinin, exhibits hydrophilic properties in contrast to the natural substance, and has better bioavailability and anti-tumor activity than artemisinin [28,29]. In therapy-sensitive and doxorubicin-resistant T-leukemia cells, ART induced apoptosis and displayed a synergistic effect in combination with doxorubicin [30]. ART's anti-tumor effect had also been demonstrated in gastrointestinal [31] and breast cancer [32]. Here, ART specifically affected neoplastic tissue and spares healthy tissue [33].

The mode of action of ART is not yet fully understood. Both in the malaria pathogen and in tumor cells, high cellular iron content seems to play a role in the response to ART (and artemisinin) [34]. A high iron content facilitates ferroptosis, an iron-dependent cell death caused by reactive oxygen species (ROS) formation [34–38]. Iron has been identified as central player in cancer progression [39]. This could explain why artemisinin and ART specifically affect tumor cells but not normal cells with lower iron content [33]. Correspondingly, RCC cells express significantly more iron-regulated genes [40]. Furthermore, RCC tissue, particularly clear cell RCC tissue, compared to healthy tissue, exhibits significant higher transferrin receptor 1 (TfR1) expression, which is responsible for iron uptake and associated with worse survival outcomes [41]. Moreover, artemisinin and its derivatives inhibit angiogenesis [42–45]. ART reduced the expression of angiogenic proteins in hemangioendothelioma cells, and thus has been postulated to be a therapeutic option for angiogenic cancers [46], including RCC. Indeed, ART exerted potent, selective cytotoxicity in therapy-sensitive RCC [47] and inhibited invasiveness in vitro and in vivo. It induced ferroptosis in therapy-sensitive RCC cells [48] and enhanced the anti-tumor effect of the TKI sorafenib [47].

Investigations exploring the effect of ART on therapy-sensitive RCC are scant and unavailable in regard to therapy-resistant RCC. Thus, the present study was designed to evaluate ART's impact on sensitive and more importantly sunitinib-resistant RCC cells, by evaluating the effect of this drug on tumor growth and underlying molecular mechanisms. The intent of this study was to implement rationale for a founded treatment option with a compound of natural origin for patients with advanced and/or therapy-resistant RCC.

#### **2. Results**

## *2.1. Confirmation of Sunitinib Resistance in RCC Cells*

Sunitinib-sensitive and sunitinib-resistant RCC cell lines, Caki-1, 786-O, KTCTL-26, and A-498, were employed with the sunitinib-sensitive (parental) RCC sub-lines serving as controls. Cells were designated sunitinib-resistant if the IC50 under escalating sunitinib dosage (0.1–100 µM) was approximately twice as high as the IC50 of the sunitinib-sensitive counterpart. Even though the A-498 cells barely reached the designated IC50 (IC50 of 19.30 µM in the resistant cells compared to 10.43 µM in their parental counterparts), all four cell lines fulfilled this specification (Table 1). Here, with an IC50 of 10.43 µM, parental A-498 revealed the weakest sunitinib response, compared to the other three RCC cell lines. The most prominent difference in IC50 was found in Caki-1 cells with an IC50 of 2.58 µM in parental and 19.13 µM in resistant cells, indicating that initial high sensitivity may lead to stronger resistance development in RCC cells. The differences in IC50 for the parental and resistant 786-O and KTCTL-26 cell lines lay between those of the A-498 and Caki-1cells. The IC50 of parental 786-O of 3.97 µM elevated in the resistant 786-O to an IC50 of 11.16 µM. In KTCTL-26 cells, IC50 of 6.37 µM in the parental cells increased to an IC50 of 13.31 µM in the resistant counterparts.

**Table 1.** Verification of sunitinib resistance: IC50 values of parental and sunitinib-resistant renal cell carcinoma (RCC) cells following 72 h application of 0.1–100 µM sunitinib *n* = 5.


#### *2.2. Artesunate Inhibits Cell Growth of Parental and Sunitinib-Resistant RCC Cells*

ART induced a dose- and time-dependent growth inhibition in all parental and resistant RCC cell lines, compared to the untreated controls (Figure 1), with comparable IC50 values for corresponding parental and resistant sub-lines. A significant growth reduction of parental Caki-1 cells with an IC50 of 10.41 µM ART after 72 h was apparent (Figure 1a). The sunitinib-resistant Caki-1 cells were similarly inhibited with an IC50 of 11.69 µM ART after 72 h treatment (Figure 1b). In both parental and sunitinib-resistant Caki-1 cells, significant growth inhibition was first reached with 5 µM ART (Figure 1a,b). Parental and sunitinib-resistant A-498 cells also first showed significant growth inhibition at a concentration of 5 µM ART, with an IC50 of 12.51 µM ART for the parental and 12.08 µM ART for the sunitinib-resistant cells (Figure 1g,h). The most prominent growth inhibition was found in the 786-O cell lines, with an IC50 of 1.62 µM ART in the parental and an IC50 of 1.99 µM ART in the resistant 786-O cells (Figure 1c,d). In 786-O cells, exposure to 1 µM ART already resulted in significant growth inhibition that further increased with ascending concentration (Figure 1c,d). In all three of the abovementioned RCC cell lines, tumor cell growth was arrested but not reduced below the initially seeded basal cell count, even at the highest concentration of 100 µM ART (Figure 1a–d,g,h). In KTCTL-26 a significant decrease in parental and sunitinib-resistant cells, below the initial seeding cell count (Figure 1e,f), did take place. ART's IC50 values for the KTCTL-26 cell lines were higher (parental: IC50 = 17.02 µM, sunitinib-resistant: IC50 = 17.79 µM) than those of the other cell lines. Comparable to Caki-1 and A-498 cells, the first significant inhibitory response in regard to KTCTL-26 growth was detected after exposure to 5 µM ART (Figure 1e,f). In summary, ART treatment significantly

suppressed tumor cell growth in all four RCC cell lines, but only in the KTCTL-26 cell line did growth reduction result in a decrease of cell counts below the number of initially seeded cells.

≤ ≤ ≤ **Figure 1.** Tumor cell growth after exposure to artesunate (ART): Tumor cell growth of parental (par) and sunitinib-resistant (res) Caki-1 (**a**,**b**), 786-O (**c**,**d**), KTCTL-26 (**e**,**f**), A-498 (**g**,**h**) cells after 24, 48, and 72 h treatment with ascending ART concentrations (1–100 µM). Cell number set to 100% after 24 h incubation. The IC50 of ART after 72 h treatment is specified. Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 5.

#### *2.3. Artesunate Impairs RCC Cell Proliferation*

Exposure to ART for 72 h contributed to significant dose-dependent inhibition of RCC cell proliferation (Figure 2). The proliferation of parental and sunitinib-resistant Caki-1 and 786-O cells was already significantly reduced after treatment with 10 µM ART, compared to the untreated controls (Figure 2a,b). Parental KTCTL-26 cells revealed a significant proliferation inhibition after exposure to 20 µM ART, while resistant KTCTL-26 cells were significantly inhibited at 30 µM ART (Figure 2c). A-498 cells behaved differently in respect to the inhibiting concentration of ART. Proliferation of the resistant A-498 cells was already significantly reduced after treatment with 20 µM ART, whereas a concentration of 30 µM ART was necessary to significantly decrease proliferation in parental A-498 cells (Figure 2d).

≤ ≤ ≤ **Figure 2.** Cell proliferation: Tumor cell proliferation of parental (par) and sunitinib-resistant Caki-1 (**a**), 786-O (**b**), KTCTL-26 (**c**), and A-498 (**d**) RCC cells incubated for 72 h with ART (10–50 µM). Untreated controls were set to 100%. Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 5.

#### *2.4. Artesunate Reduces Clonogenic Growth of the RCC Cell Lines*

In all RCC cell lines, ART induced a significant dose-dependent reduction in clone colonies after 10 days incubation (Figure 3). Ten µM ART contributed to significant inhibition of the clonogenic growth of the RCC cells, compared to the untreated controls. In parental and resistant Caki-1 cells, the administration of 50 µM ART diminished the clonogenic growth by more than 90% (Figure 3a). Microscopically, parental Caki-1 cells formed larger colonies, compared to the sunitinib-resistant Caki-1 cells (Figure 3a). Treatment of 786-O cells with 10 µM ART resulted in an approximately 50% decrease in clone colonies (Figure 3b). 786-O cells exposed to 50 µM ART completely inhibited colony formation in the parental and resulted in only a few colonies in the resistant cell line. In parental and sunitinib-resistant KTCTL-26 and A-498 cells, 10 µM ART significantly diminished the clonogenic growth by more than 50% (Figure 3c,d). KTCTL-26 colonies were no longer formed after exposure to 30 µM ART in parental and exposure to 50 µM ART in resistant cells (Figure 3c). Neither parental nor resistant A-498 colonies were detectable after exposure to 40 and 50 µM ART (Figure 3d). Microscopic comparison showed that both parental and resistant A-498 cells exhibited a lower potential to develop colonies, compared to the other RCC cell lines (Figure 3d).

≤ ≤ **Figure 3.** Clonogenic growth of RCC cells: Clonogenic growth of parental and resistant Caki-1 (**a**), 786-O (**b**), KTCTL-26 (**c**), and A-498 (**d**) cells treated with ART (10–50 µM) for 10 days. Untreated cells served as controls (set to 100%). Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001. *n* = 5.

#### *2.5. Artesunate Induces Cell Cycle Arrest in Both Parental and Sunitinib-Resistant RCC Cells* ≤ ≤

Diminished growth behavior in the parental and sunitinib-resistant RCC cell lines Caki-1, 786-O, and A-498 was accompanied by a significant G0/G1 phase arrest after exposure to ART, compared to the untreated controls (Figure 4a,b,d). Concomitantly, the number of S and G2/M phase cells significantly decreased, except in the parental 786-O cells, exhibiting only a significant reduction in the S phase (Figure 4b). In KTCTL-26 ART induced a significant G0/G1 phase arrest in the resistant cells, but no changes were apparent in the parental counterpart (Figure 4c). Overall, the effect of ART on KTCTL-26 regarding induction of cell cycle arrest was less pronounced than in the other RCC cell lines. To explore the influence of ART on cell cycle regulating protein levels, Caki-1 and 786-O were utilized as exemplary cell lines.

≤ ≤ ≤ ≤ ≤ ≤ **Figure 4.** Distribution of cell cycle phases: Proportion of parental and sunitinib-resistant RCC cells, Caki-1 (**a**), 786-O (**b**), KTCTL-26 (**c**), and A-498 (**d**), in the G0/G1, S, and G2/M phases after 48 h treatment with ART (20 µM). Untreated cells served as controls (dotted line; set to 100%). Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 5.

#### *2.6. Artesunate-Induced Cell Cycle Arrest was Accompanied by Alterations in the Expression and Activity of Cell Cycle Regulating Proteins*

Alterations in the cell cycle phases of Caki-1 and 786-O after administration of ART were associated with distinct modulation in cell cycle regulating proteins (Figure 5, Figure 6, Figure 7). The treatment of parental and sunitinib-resistant Caki-1 and 786-O cells with 20 µM ART led to a significant reduction of the cell cycle activating proteins cyclin A, cyclin B, and CDK1, as well as to deactivation of CDK1 (pCDK1), all of which are proteins involved in S and G2/M phase progression. In addition, CDK2, which associates with cyclin A during the S phase, significantly decreased in parental Caki-1 cells after exposure to ART, compared to the untreated controls (Figures 5 and 6g, Figure S1g). The expression of the tumor suppressor p27 significantly increased in parental and resistant Caki-1 cells with ART application (Figures 5 and 6b, Figure S1b). However, in 786-O cells, p27 expression significantly decreased after ART application (Figures 5 and 7b, Figure S2b). Protein expression of p21 was not significantly altered in Caki-1 cells (Figures 5 and 6a, Figure S1a) but tended to increase in parental 786-O cells and was significantly increased in resistant 786-O cells (Figures 5 and 7a, Figure S2a). Activity of CDK2 (pCDK2) was not detectable in Caki-1 and 786-O cells.

**Figure 5.** Protein expression profile of cell cycle regulating proteins: Representative Western blot analysis of cell cycle regulating proteins in parental (par) and sunitinib-resistant (res) Caki-1 (**left** panel) and 786-O (**right** panel) cells after 48 h exposure to ART (20 µM).

≤ ≤ ≤ **Figure 6.** Protein expression profile of cell cycle regulating proteins: Pixel density analysis (Western blot) of the cell cycle regulating proteins p21 (**a**), p27 (**b**), cyclin A (**c**), cyclin B (**d**), CDK1 (**e**), pCDK1 (**f**), and CDK2 (**g**) in parental and resistant Caki-1 cells after 48 h exposure to ART (20 µM), compared to untreated controls (set to 100%). Each protein analysis was accompanied and normalized by a housekeeping protein. Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 4. For detailed information regarding the Western blots see Figure S1a–g.

≤ ≤ ≤ – **Figure 7.** Protein expression profile of cell cycle regulating proteins: Pixel density analysis (Western blot) of the cell cycle regulating proteins p21 (**a**), p27 (**b**), cyclin A (**c**), cyclin B (**d**), CDK1 (**e**), pCDK1 (**f**), and CDK2 (**g**) in parental and resistant 786-O cells after 48 h exposure to ART (20 µM), compared to untreated controls (set to 100%). Each protein analysis was accompanied and normalized by a housekeeping protein. Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 4. For detailed information regarding the Western blots see Figure S2a–g.

#### *2.7. Artesunate Only Slightly Contributes to Apoptosis*

To investigate whether the growth inhibitory effect of ART was associated with cell death events, apoptosis was assessed (Figure 8). The only significant increase in apoptotic cells in parental or sunitinib-resistant RCC cells after exposure to ART occurred in resistant Caki-1 cells (Figure 8a).

≤ ≤ **Figure 8.** Apoptotic events: Parental and resistant Caki-1 (**a**), 786-O (**b**), KTCTL-26 (**c**), and A-498 (**d**) cells treated for 48 h with ART (20 µM). Untreated cells served as controls (set to 100%). Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\*\* *p* ≤ 0.001, ns = not significant. *n* = 5.

#### *2.8. Artesunate Results in Ferroptosis Induction in KTCTL-26 Cells*

Since ART has been shown to induce the iron-dependent cell death termed ferroptosis [35,36,38], this type of cell death was investigated by utilizing ferrostatin-1, a ferroptosis inhibitor. Proliferation inhibition observed under ART exposure in parental and sunitinib-resistant KTCTL-26 cells was significantly reversed following the combined administration of ART with the ferroptosis inhibitor ferrostatin-1 (Figure 9a,b). This cancellation of ART's inhibitory effect caused proliferation rates to return to those of the untreated control cells. Application of ferrostatin-1 did not cancel ART's proliferation inhibition in parental or sunitinib-resistant Caki-1, 786-O, and A-498 cells. Thus, only the KTCTL-26 cell lines were investigated in further detail.

An essential process during ferroptosis is ROS generation. To investigate whether ART in fact generates ROS, Trolox, an antioxidant was used to intercept free radicals and thus prevent ferroptosis. ART in combination with Trolox significantly abrogated the proliferation inhibition observed with ART treatment alone, so that the proliferation rate of the KTCTL-26 cells was comparable to that of the untreated controls (Figure 9c,d). To further corroborate the results of the aforementioned experiments, glutathione (GSH) expression, a part of the anti-oxidative protective system of the cells, was evaluated. GSH significantly decreased in KTCTL-26 cells after treatment with 50 µM ART, compared to the untreated controls (Figure 9e), indicating ROS generation and GSH consumption. The GSH content was more strongly reduced in parental than in resistant KTCTL-26 cells. In both parental and sunitinib-resistant KTCTL-26 cells, the inhibitory effect significantly increased when ART was combined with iron (Figure 9e). Since GPX4 is essential for anti-oxidative protection and designated as a ferroptosis related protein, GPX4 expression was also assessed. Application of 50 µM ART resulted in a significant reduction of GPX4 in both parental and resistant KTCTL-26 cells (Figures 9f,g and S3).

p53 has been described as a ferroptosis indicator [49,50]. Thus, the expression of p53 in parental and sunitinib-resistant RCC cells was investigated. Expression of p53 was not detectable in the parental and sunitinib-resistant Caki-1, 786-O, and A-498 cells (Figures 9h and S4). Notably, in KTCTL-26 cells, the only cells where ART did induce ferroptosis, distinct p53 expression in parental and even stronger p53 expression in the resistant cells was detected (Figures 9h,i and S4).

reactive oxygen species ( ≤ ≤ ≤ – β **Figure 9.** Artesunate induced ferroptosis by reactive oxygen species (ROS) formation in p53-positive KTCTL-26 cells: Ferroptosis induction (**a**,**b**) Proliferation of parental (**a**) and sunitinib-resistant KTCTL-26 cells (**b**) treated for 48 h with ART (20, 50 µM) and ferrostatin-1 (Fer-1) (20 µM). Untreated (100%) and ART mono-treated cells served as controls. Error bars indicate standard deviation (*SD*). Significant difference compared to untreated controls, except for asterisk brackets indicating significant difference between ferrostatin-1 untreated and treated cells: \* = *p* ≤ 0.05, \*\* = *p* ≤ 0.01, \*\*\* = *p* ≤ 0.001, ns = not significant. *n* = 5. Indications of ROS generation (**c**–**g**): Proliferation of parental (**c**) and sunitinib-resistant KTCTL-26 cells (**d**) treated for 48 h with ART (20, 50 µM) and Trolox (0.5 mM). Untreated cells served as controls (100%). *n* = 5. GSH level (%) of parental (par) and resistant (res) KTCTL-26 cells after 24 h incubation with ART (50 µM) and holo-transferrin (Fe) (**e**). Untreated controls served as controls (100%). *n* = 5. GPX4 expression: Representative Western blot of GPX4 expression in parental (par) and sunitinib-resistant (res) KTCTL-26 cells after 48 h exposure to ART (50 µM) (**f**). Pixel density analysis of GPX4 level (%) after 48 h exposure to ART (50 µM) in parental (par) and resistant (res) KTCTL-26 cells (**g**). Untreated cells served as controls (100%). β-actin served as internal control. *n* = 5. Protein expression of p53 (**h**,**i**) Representative Western blot analysis of p53 in parental (par) and resistant (res) Caki-1, 786-O, KTCTL-26, and A-498 cells (**h**). Pixel density analysis of p53 expression in parental (100%) and resistant KTCTL-26 cells (**i**) p53 protein analysis was accompanied and normalized by a total protein control. *n* = 3. Error bars indicate standard deviation (*SD*). Significant difference indicated by: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, ns = not significant. For detailed information regarding the Western blots of (**f**) and (**h**) see Figures S3 and S4.

#### *2.9. Artesunate Influences the Metabolism of RCC Cells*

The ART-induced ferroptosis in KTCTL-26 cells was accompanied by a significant increase in ROS. Since ferroptosis is associated with a high iron content and accelerated metabolism, ART's impact on the oxygen consumption rate (OCR) of the KTCTL-26 cells, expressed by basal respiration, adenosine triphosphate (ATP) production-coupled respiration, maximum and reserve capacities, and non-mitochondrial respiration, was assessed (Figure 10). Exposure to ART significantly inhibited the spare respiratory capacity, representing the ability of cells to enhance respiration in response to physiological or pharmacological stress, in resistant KTCTL-26 cells (Figure 10e). Decreased spare respiration capacity in the resistant KTCTL-26 cells was accompanied by diminished ATP production (Figure 10f). Also, in parental RCC cells, ATP production significantly decreased after exposure to ART (Figure 10f). ART exerted no significant effect on basal or maximum respiration in either parental or resistant cells (Figure 10c,d). No alteration in the extracellular acidification rate connected with anaerobic glycolytic activity was observed after exposure to ART in the KTCTL-26 cells, thus indicating no shift towards compensatory glycolysis.

≤ ≤ ≤

**Figure 10.** Mitochondrial respiration: Representative mitochondrial respiration in parental (**a**) and resistant (**b**) KTCTL-26 cells after 24 h treatment with 20 µM ART (=treated). Untreated cells served as controls. Data pertaining to the oxygen consumption rate (OCR) were normalized to total basal respiration (set to 100%) consisting of mitochondrial and non-mitochondrial respiration. Extracted values for mitochondrial basal oxygen consumption rate (OCR) (**c**), maximal OCR (**d**), respiratory reserve capacity (**e**), and adenosine triphosphate (ATP) production (**f**) after 24 h ART application (ART). MFI = mean fluorescence intensity. Error bars indicate standard deviation (*SD*). Significant difference to untreated control: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, ns = not significant. *n* = 4.

#### **3. Discussion**

Although current therapeutic approaches have improved progression-free survival of advanced RCC patients, the disease at this stage ultimately remains incurable due to the inevitable development of resistance to treatment. Interestingly, in the current study RCC cells exhibiting a more sensitive initial response to sunitinib developed strong resistance in the course of chronic treatment. Sunitinib-resistant Caki-1 cells were nearly 10-fold less sensitive than their parental counterparts, which initially could be held in check by a relatively low sunitinib dose. Overcoming this resistance is therefore of primary importance. Since adding ART to conventional anti-cancer therapy has been shown to overcome resistance during treatment of other tumor entities, the impact of ART on a panel of sunitinib-sensitive and sunitinib-resistant RCC cell lines was investigated.

Exposure to ART resulted in a significant inhibition of tumor cell growth and proliferation in all tested parental and sunitinib-resistant RCC cells, indicating an anti-tumor potential in highly heterogenic types of cancer. KTCTL-26 cells displayed the highest sensitivity to ART. The IC50 for ART in the RCC cells was in the lower one- to two-digit µM range (2 to 18 µM). In good accordance with other investigators, ART has been shown to inhibit cell growth of therapy-sensitive RCC cells in the two-digit µM range, up to 50 µM [47]. Combined administration of ART with sorafenib, a first-generation TKI akin to sunitinib, even further reduced cell growth [47]. Several studies on non-urological tumor entities have also demonstrated growth inhibition after ART application. In hemangioendothelioma cells, ART time- and dose-dependently reduced tumor cell growth, concomitantly decreasing the expression of VEGF-A, VEGFR1, VEGFR2, and HIF-1α [46]. Thus, it has been postulated that ART may hold promise in treating vascular tumors, of which RCC is a member. Moreover, significant growth inhibition was observed in a mouse model of hemangioendothelioma carcinoma cells following ART treatment, with significantly reduced tumor size [46]. In different gastric cancer cell lines, ART exposure has also resulted in a significant growth reduction [51]. In bone tumor cell lines, ART also impacted tumor cell growth [52]. In the present investigation, the growth of KTCTL-26 cells, a bone tumor cell line, was even diminished to below zero after ART treatment. This was accompanied by an increased number of annexin V positive cells, indicating apoptosis induction by ART. Other investigators have reported an anti-proliferative effect of ART in ovarian carcinoma cells in vitro [53] as well as in chemotherapy-sensitive and -resistant thyroid cancer cells [54]. In patients with colorectal cancer, ART application reduced disease progression through anti-proliferative action [31]. Artemisinin, the native lead compound, also inhibited proliferation in gastric cancer cell lines by up-regulating p53 [55]. Nevertheless, ART inhibits tumor cells by both p53-dependent and also -independent mechanisms [56,57].

Clonogenic growth provides information about the growth of single tumor cells at metastatic sites and advanced RCC is characterized by its ability to spread and survive at these sites. A prolonged 10 µM ART exposure of up to 10 days contributed to a significant reduction of clone colonies in all parental and respective sunitinib-resistant RCC cell lines. KTCTL-26, but also A-498 cells displayed a high sensitivity towards ART with regard to clone colony formation, followed by 786-O and Caki-1. In good accordance with Jeong et al., a significant decrease in clonogenic growth in therapy-sensitive Caki-1 and 786-O cells has been shown [47]. Not only ART, but also artemisinin significantly diminished clonogenic growth in therapy-sensitive RCC cells by down-regulating AKT, a survival protein, and up-regulating E-cadherin, an epithelial differentiation marker [58]. E-cadherin loss is associated with poor prognosis and continued spread of disease [59]. Since AKT up-regulation and loss of E-cadherin have previously been demonstrated in therapy-resistant RCC cells [60–64], it is conceivable that these proteins are also affected by ART in the sunitinib-resistant RCC cells.

Reduced cell growth and proliferation in response to ART were associated with impaired cell cycle progression. Parental and resistant Caki-1, 786-O, and A-498 cells displayed a significant G0/G1 phase arrest after exposure to 20 µM ART. Accumulation of the cells in G0/G1 correlated with a significant reduction of the cells in the S and G2/M phase. However, parental KTCTL-26 cells were not affected, and their sunitinib-resistant counterparts were only moderately affected. Concordant with the present

investigation, ART and other derivatives of artemisinin have been shown to promote cell cycle arrest in the G0/G1 phase in several tumor entities [65,66]. Application of ART in epidermoid carcinoma cells has been shown to halt cells in the G0/G1 phase [65]. Artemisinin application has resulted in a similar effect in cell cultures from endometrial tumors [66]. In therapy-sensitive RCC cells, Caki-1, and 786-O, 50 µM ART has been shown to induce a G2/M phase arrest [47], whereas the G0/G1 cell cycle arrest observed in the present investigation was induced with 20 µM ART. These differences in cell cycle arrest could be due to the different ART concentrations. Depending on dose, other investigators have demonstrated a ROS-dependent cell cycle arrest induced by ART in both the G0/G1 and G2/M phases in breast cancer cells [67].

Cell cycle progression is controlled by alternating CDK-cyclin complexes. In good accordance with the G0/G1 cell cycle arrest induced by ART, the cell cycle activating proteins CDK 1/2 and cyclin A/B, responsible for S and G2/M phase progression, were down-regulated, whereas the cell cycle inhibiting proteins, p21 in 786-O cells, and p27 in Caki-1 cells, were elevated. The CDK2-cyclin A complex mediates DNA replication in the S phase [68]. The CDK1-cyclin A complex promotes S phase transition, and CDK1-cyclin B complex drives transition from the G2 to M phase [69]. Hence, there is strong evidence that CDK1, CDK2, cyclin A, and B down-regulation by ART evokes the G0/G1 phase arrest, inhibiting growth of the RCC cells. Indeed, blocking CDK1/2 or cyclin A/B by small interfering RNA has been shown to significantly reduce cell growth in Caki-1, KTCTL-26, and A-498 cells [18,19]. Increased p21 and p27 after ART application are also indicative of cell cycle arrest, as both proteins mediate cell cycle arrest in the G0/G1 phase [70,71]. In epidermoid cancer cells, administration of ART resulted in a G0/G1 phase arrest and concomitant p27 increase [65]. Consistent with the current investigation, cell cycle arrest of the epidermoid cancer cells after ART treatment correlated with down-regulation of cyclin A1, cyclin B, and CDK2 [65]. However, in 786-O cells the expression of p27 was already high and significantly diminished after ART exposure. Studies on bone cancer have demonstrated that p27 in addition to its anti-tumor function can play a role in oncogenesis [72]. In line with this, for some renal cell carcinomas, increased p27 expression was associated with worse prognosis [73]. This may hold true for the 786-O cell lines but remains speculative and requires further investigation. Thus, ART seems to act cell type-dependently, attributable to the initial protein content and/or stage of disease. This might also be clinically important with regard to the intra-tumor heterogeneity of RCC [74], since RCC is a tumor entity harboring varying molecular signatures with different sensitivity to treatment [75].

Evidence has been presented showing that cell death may also be responsible for growth inhibition by ART [47,51]. However, in the current study, only in sunitinib-resistant Caki-1 cells were significant apoptotic effects apparent after ART treatment. Based on the dose–response curves and the fact that KTCTL-26 cells reveal no or just slight effects on cell cycle progression, it might be assumed that ART enables cell death in the KTCTL-26 cell line. However, ART did not induce apoptosis in the KTCTL-26 cells. Furthermore, 786-O and A-498 cell lines also displayed no apoptosis induction under ART treatment. Other investigators have shown that ART induced apoptosis in tumor cells, but often only after application of higher ART concentrations than used in the current study. In stomach tumors, apoptotic events were detected in vitro with concentrations upwards of 50 µM ART [51]. Similarly, induction of apoptosis was apparent after 48 h exposure to 50 µM ART in therapy-sensitive Caki-1 and 786-O cells [47]. Since the tumor cell growth of Caki-1, 786-O, and A-498 cell lines was also not reduced below the initial cell count at seeding, even with the higher 50 µM ART concentration, apoptosis induction by ART can only play a minor role in controlling RCC. This leaves the question open as to how to explain the reduction in the KTCTL-26 cells below that of the initial seeding count after exposure to ART.

One explanation of the magnitude of this ART-induced growth inhibitory effect might be induction of ferroptosis. Indeed, parental and sunitinib-resistant KTCTL-26 demonstrated a significant reversion in ART's growth inhibitory effect after additional application of the ferroptosis inhibitor ferrostatin-1, indicating that ART does induce ferroptosis. Accordingly, ferroptotic effects have been demonstrated

in cell cultures of head and neck cancer after ART treatment [25]. In pancreatic cancer cells, ART also triggered ferroptosis [36], and sorafenib combined with ART induced ferroptosis in liver cancer cells [76].

Caki-1, 786-O, and A-498 cells did not show any response to ferrostatin-1, indicating that ferroptosis does not take place in these cell lines. The ART-induced growth inhibition in these cells must therefore mainly act through cell cycle arrest.

Since ART induced ROS generation during ferroptosis [48], Trolox, a vitamin E derivative and anti-oxidant that neutralizes ROS [32], was applied to investigate whether the inhibitory effect of ART could be canceled. In both parental and sunitinib-resistant KTCTL-26 cells, the inhibitory effect of ART was significantly reversed, showing that ART can act through ROS generation. Glutathione (GSH), a key regulator of excessive ROS levels [77], was also significantly reduced after ART administration. ART evoked a stronger GSH reduction in parental KTCTL-26 cells than in resistant cells, which might mean that these resistant RCC cells have a higher basal ROS tolerance. Support for this thesis is provided by a proteomic study, showing that glutathione metabolism in sunitinib-resistant 786-O RCC cells was increased 4–5 times compared to parental cells [78]. Hence, adding ART to sunitinib treatment might counteract ROS tolerance in therapy-resistant cells, and facilitate ferroptosis. In the current investigation, combining ART with iron further potentiated the decrease of GSH in both parental and sunitinib-resistant KTCTL-26 cell lines. ART in combination with lysosomal iron led to the development of ROS and ultimately induces apoptosis via the intrinsic pathway [32]. Increased efficacy of ART in the presence of iron has been shown in pancreatic [36] and in breast cancer cell lines [32]. A high iron content within the tumor cells therefore seems to augment ART's efficacy, and tumor cells with increased iron metabolism could be selectively targeted, including RCC [79].

Phospholipid-hydroxy peroxide-glutathione peroxidase (PHGPx, gene: GPX4) is another key protein involved in augmented ROS generation. Substances containing an endoperoxid group, such as ART, directly inhibit GPX4, first sensitizing "GPX4 tumors" to ferroptosis [80] and ultimately leading to ferroptosis [81]. GPX4 was significantly reduced in parental and resistant KTCTL-26 cells after ART treatment. Over-expression of GPX4 prevented ferroptosis in colorectal cancer in in vitro studies [82]. Consequently, this might also be the case for the KTCTL-26 cell lines.

p53 has been described as a possible ferroptosis enhancer [49,50]. Interestingly, p53 was exclusively expressed in the parental and sunitinib-resistant KTCTL-26 cell lines, the only RCC cell lines demonstrating ferroptosis induction after exposure to ART. p53 inhibited cysteine influx and thus disrupted GSH metabolism [83]. Furthermore, and consistent with our results, ferroptosis could not be induced in p53-defective cells. Hence, p53 expression may impact GSH metabolism and might be a predictor for ferroptosis induction in parental and sunitinib-resistant RCC cells. Still, this is speculative and requires further investigation. In tumor cells inducing apoptosis, cell death is regulated by both p53-dependent and -independent pathways [56,57].

Along with induction of ferroptosis, ART exposure resulted in significantly diminished ATP production and spare reserve capacity in both parental and sunitinib-resistant KTCTL-26 cells, throttling the energy supply necessary for tumor cell progression. Consistent with this, ART administration in B-cell lymphoma cells and prolactinoma cells led to reduced ATP production [84,85].

The results presented here show that ART induced significant growth inhibitory effects in parental and, more importantly, sunitinib-resistant RCC cells. Although all four RCC cell lines responded to ART, cell type-specific responses were evident. This might give an insight on how ART may act in heterogeneous tumors. In parental and resistant Caki-1, 786-O, and A-498 cells, growth inhibitory effects were accompanied by cell cycle arrest in the G0/G1 phase and respective modulation of the cell cycle regulating proteins. It may therefore be assumed that ART led to growth and proliferation inhibition, but not to tumor cell death. In contrast, parental and sunitinib-resistant KTCTL-26 cells were mainly affected by ART through ferroptosis and decreased metabolism, leading to both growth inhibition and tumor cell death. Notably, p53 was only evident in the KTCTL-26 cells, indicating that p53 might be predicative for ART-dependent ferroptosis and induce a more effective drug response, which could be clinically relevant. The in vitro data give a first insight into the anti-tumor activity of ART in RCC cells that might in its strength and the respective mechanism depend on the initial protein profile of the tumor cells and therewith aspects of the intra-tumor heterogeneity. However, since in vitro data reveal an isolated tumor cell system, further investigations are necessary to verify our postulates and in vivo studies need to clarify whether ART shows similar anti-tumor effects under physical conditions in parental and sunitinib-resistant RCC.

#### **4. Materials and Methods**

#### *4.1. Cell Cultures*

Renal cell carcinoma cell lines Caki-1, 786-O, KTCTL-26, and A-498 were kindly provided by Prof. Dr. Roman Blaheta (Department of Urology, University Hospital Frankfurt, Goethe-University, GER), initially purchased from Promocell (LGC Promochem, Wesel, Germany). Caki-1 cells were grown and sub-cultured in Iscove Basal medium (Biochrom GmbH, Berlin, Germany), 786-O, KTCTL-26, and A-498 were grown in RPMI-1640 medium (Gibco, Thermo Fisher Scientific, Darmstadt, Germany). Media were supplemented with 10% fetal calf serum (FCS) (Gibco, Thermo Fisher Scientific, Darmstadt, Germany), 1% glutamax (Gibco, Thermo Fisher Scientific, Darmstadt, Germany), and 1% Anti/Anti (Gibco, Thermo Fisher Scientific, Darmstadt, Germany). Twenty mM HEPES-buffer (Sigma-Aldrich, Darmstadt, Germany) was added to the RPMI-1640 medium. Tumor cells were cultivated in a humidified, 5% CO<sup>2</sup> incubator.

#### *4.2. Resistance Induction and Application of Sunitinib and Artesunate*

Resistance to sunitinib was induced by chronic exposure to ascending sunitinib (free Base, Massachusetts LC Laboratories, Woburn, MA, USA) concentrations from 0.1–1 µM until the cells survived and adapted to the highest dosage. Sunitinib resistance in the RCC cells occurred in average 10 weeks after starting application. Thereafter, they were maintained with 1 µM sunitinib applied three times a week. The IC50 (half-maximal inhibitory concentration) of sunitinib was investigated to verify drug resistance. After starving chronically sunitinib-treated RCC cells for 3 days 0.1–100 µM sunitinib was applied for 72 h. Therapy-sensitive (parental) RCC subcell lines served as controls. RCC cells were designated as sunitinib-resistant when the IC50 under 72 h sunitinib application was approximately doubled.

Artesunate (ART) (Sigma-Aldrich, Darmstadt, Germany) was applied at a concentration of 1–100 µM. Controls (parental and sunitinib-resistant) remained ART-untreated. The IC50 of ART in parental and sunitinib-resistant RCC cells was evaluated analog to sunitinib using the 72 h growth data at a concentration of 1–100 µM ART. To evaluate possible toxic effects of sunitinib and/or ART, cell viability was determined in parallel to experimentation by testing aliquoted cells with trypan blue (Sigma-Aldrich, Darmstadt, Germany). Only viable cells were employed.

#### *4.3. Tumor Cell Growth*

Cell growth was assessed using 3-(4,5-dimethylthiazol- 2-yl)-2,5-diphenyltetrazolium bromide (MTT) dye. RCC cells (50 µL, 1 × 10<sup>5</sup> cells/mL) were seeded onto 96-well-plates. After 24, 48, and 72 h, 10 µL MTT (0.5 mg/mL) (Sigma-Aldrich, Darmstadt, Germany) was added for 4 h. Cells were then lysed in 100 µL solubilization buffer containing 10% SDS in 0.01 M HCl. The 96-well-plates were subsequently incubated overnight at 37 ◦C, 5% CO2. Absorbance at 570 nm was determined for each well using a multi-mode microplate-reader (Tecan, Spark 10 M, Crailsheim, Germany). After subtracting background absorbance and offsetting with a standard curve, results were expressed as mean cell number. To illustrate dose-response kinetics, mean cell number after 24 h incubation was set to 100%. Each experiment was done in triplicate.

### *4.4. Proliferation*

Cell proliferation was measured using a BrdU (Bromodeoxyuridine / 5-bromo-2′ -deoxyuridine) cell proliferation enzyme-linked immunosorbent assay (ELISA) kit (Calbiochem/Merck Biosciences, Darmstadt, Germany). Tumor cells (50 µL, 1 × 10<sup>5</sup> cells/mL), seeded onto 96-well-plates, were incubated with 20 µL BrdU-labeling solution per well for 24 h, fixed and stained using anti-BrdU mAb according to the manufacturer's protocol. Absorbance was measured at 450 nm using a multi-mode microplate-reader (Tecan, Spark 10 M, Crailsheim, Germany). Values were presented as percentage compared to untreated controls set to 100%.

#### *4.5. Clonogenic Assay*

The clonogenic recovery potential gives insight into the capability of the cells to form a new tumor (metastasis). Therefore, 500 cells/well were seeded on a 6-well-plate and treated for 10 days with ART. Untreated cells served as controls. RCC cells were subsequently fixed with 85% MeOH/15% AcOH and stained with Coomassie (0.5 g Coomassie Blue G250 (Sigma-Aldrich, Darmstadt, Germany), 75 mL AcOH, 200 mL MeOH, 725 mL distilled water). Amount and size of cell clone colonies were measured with a biomolecular imager (Sapphire, Azure Biosystems, Biozym, Hess. Oldendorf, Germany). Colony forming efficiency was evaluated by ImageJ analysis. A colony was defined as consisting of at least 50 cells with an area of 50.8 µm<sup>2</sup> . Untreated controls were set to 100%.

#### *4.6. Cell Cycle Phase Distribution*

For cell cycle analysis cell cultures were grown to sub-confluency. A total of 1 × 10<sup>6</sup> cells was stained with propidium iodide (50 µg/mL) (Invitrogen, Thermo Fisher Scientific, Darmstadt, Germany) and then subjected to flow cytometry (Fortessa X20, BD Biosciences, Heidelberg, Germany). Ten thousand events were collected from each sample. Data acquisition was carried out using DIVA software (BD Biosciences, Heidelberg, Germany), and cell cycle distribution was analyzed by ModFit LT 5.0 software (Verity Software House, Topsham, ME, USA). The number of cells in the G0/G1, S, or G2/M phases was expressed as a percentage.

#### *4.7. Western Blot Analysis of Cell Cycle Regulating Proteins, GPX4 and p53*

To explore the expression and activity of cell cycle and cell death regulating proteins, western blot analysis was performed. Tumor cell lysates (50 µg) were applied to 10% or 12% polyacrylamide gel and separated for 10 min at 80 V and 1 h at 120 V. The protein was then transferred to nitrocellulose membranes (1 h, 100 V). After blocking with 10% non-fat dry milk for 1 h, the membranes were incubated overnight with the following primary antibodies directed against cell cycle proteins: p21 (Rabbit IgG, clone 12D1, Cell Signaling, Frankfurt am Main, Germany), p27 (Mouse IgG1, clone 57/Kip1, BD Biosciences, Heidelberg, Germany), Cyclin A (Mouse IgG1, clone 25, BD Biosciences, Heidelberg, Germany), Cyclin B (Mouse IgG1, clone 18, BD Biosciences, Heidelberg, Germany), CDK1 (Mouse IgG1, clone 2, BD Biosciences, Heidelberg, Germany), pCDK1 (Rabbit, clone 10A11, Cell Signaling, Frankfurt am Main, Germany), CDK2 (Mouse IgG2a, clone 55, BD Biosciences, Heidelberg, Germany).

To indicate lipid peroxidation and ferroptosis related proteins the following primary antibodies were used: GPX4 (Rabbit IgG, ab41787, Abcam, Berlin, Germany), p53 (Rabbit, clone 7F5, Cell Signaling, Frankfurt am Main, Germany). HRP-conjugated rabbit-anti-mouse IgG or goat-anti-rabbit IgG served as secondary antibodies (IgG, both: dilution 1:1000, Dako, Glosturp, Denmark). The membranes were incubated with ECL detection reagent (AC2204, Azure Biosystems, Munich, Germany) to visualize proteins with a Sapphire Imager (Azure Biosystems, Munich, Germany). β-actin (clone AC-1; Sigma Aldrich, Taufenkirchen, Germany) served as the internal control, except for p53, which was normalized to total protein. To quantify total protein all membranes were stained by Coomassie brilliant blue and measured by Sapphire Imager. AlphaView software (ProteinSimple, San Jose, CA, USA) was used for pixel density

analysis of the protein bands. The ratio of protein intensity/β-actin intensity or whole protein intensity was calculated and expressed in percentage, related to the untreated controls, set to 100%.

#### *4.8. Apoptosis and Ferroptosis*

To investigate apoptotic and necrotic events, the FITC-Annexin V Apoptosis Detection kit (BD Biosciences, Heidelberg, Germany) was used to quantify binding of Annexin V/propidium iodide (PI). After washing tumor cells twice with PBS, 1 × 10<sup>5</sup> cells were suspended in 500 µL of 1 × binding buffer and incubated with 5 µL Annexin V-FITC and (or) 5 µL PI in the dark for 15 min. Staining was measured by flow cytometer (Fortessa X20, BD Biosciences, Heidelberg, Germany). Ten thousand events were collected from each sample. The percentage of apoptotic and necrotic cells in each quadrant was calculated using DIVA software (BD Biosciences, Heidelberg, Germany). Further analysis was done by FlowJo software (BD Biosciences, Heidelberg, Germany).

A BrdU cell proliferation enzyme-linked immunosorbent assay (ELISA) kit (Calbiochem/Merck Biosciences, Darmstadt, Germany) was used to evaluate ferroptosis and ROS generation. To evaluate ferroptosis, tumor cells were treated for 48 h with 20, 50, and 100 µM ART or ART combined with 20 µM ferrostatin-1 (Sigma-Aldrich, Darmstadt, Germany), a ferroptosis inhibitor. ROS generation during ferroptosis was verified by treating the RCC cells for 48 h with 20, 50, and 100 µM ART in combination with 0.5 mM Trolox (Sigma-Aldrich, Taufkirchen, Germany), an antioxidant. For more details see "Proliferation" (4.4), as described above.

#### *4.9. GSH-Assay*

The GSH level was evaluated with the GSH-Glo™ Glutathione Assay (Promega Corporation, Madison, Wisconsin, USA). Five thousand cells/well were seeded onto a 96-well-plate and incubated for 24 h with 50 µM ART or ART combined with 20 µg/mL holo-Transferrin (Fe; Sigma-Aldrich, Taufkirchen, Germany). Experiments were performed according to the manufacturer's protocol. Luminescence was measured using a multi-mode microplate-reader (Tecan, Spark 10 M, Tecan, Grödig, Austria).

#### *4.10. Evaluation of Mitochondrial Respiration and Anaerobic Glycolytic Activity*

Mitochondrial respiration (OCR = oxygen consumption rate) and anaerobic glycolytic activity (EACR = extracellular acidification rate) were assessed in real time by the Seahorse XFp Extracellular Flux Analyzer using the Seahorse XF Cell Mito Stress Test Kit (both: Agilent Technologies, Waldbronn, Germany). The EACR indicating anaerobic glycolytic activity was used to determine compensatory glycolysis. OCR was obtained by multiple parameters, including basal respiration, ATP production-coupled respiration, maximal and reserve capacities, and non-mitochondrial respiration. Cells stained with CellTracker Green CMFDA (Thermo Fisher Scientific, Darmstadt, Germany) were plated at a density of 2 × 10<sup>4</sup> cells/well and media was replaced with XF Assay media the following day 1h prior to the assay and incubated without CO2. Five measurements of OCR and ECAR were taken at baseline and after each injection of the following mitochondrial modulators: Oligomycin (1.5µM, Inhibitor of ATP synthase), carbonylcyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) (1 µM, proton gradient uncoupler), and rotenone/actinomycin A (0.5 µM, inhibitors of complex I/Complex III). Data were normalized by using Wave 2.6.1 (Agilent Technologies, Waldbronn, Germany) desktop software to the mean fluorescent intensity of cells in the area of measurement in each well. Data pertaining to the OCR were normalized to total basal respiration (set to 100%) consisting of mitochondrial and non-mitochondrial respiration. Basal and maximal respiration were calculated by subtracting non-mitochondrial OCR. Respiratory reserve capacity was calculated as the difference between maximal and basal OCR. ATP-linked OCR was estimated as the difference between basal and rotenone/actinomycin A inhibited OCR.

## *4.11. Statistical Analysis*

All experiments were performed at least three times. The evaluation and generation of mean values, the associated standard deviation, and normalization in percent were done by Microsoft Excel. Statistical significance was calculated with GraphPad Prism 7.0 (GraphPad Software Inc., San Diego, CA, USA): Two-sided T-test (Western blot, apoptosis, cell cycle), one-way ANOVA test (BrdU), and two-way ANOVA test (MTT). Correction for multiple comparisons was done using the conservative Bonferroni method. Differences were considered statistically significant at a *p*-value ≤ 0.05.

#### **5. Conclusions**

ART induced cell-type specific anti-tumor effects in both parental and sunitinib-resistant RCC cells. In three of the four tested cell lines, Caki-1, 786-O, and A-498, ART induced a strong G0/G1 phase arrest. In the KTCTL-26 cell line, the phase arrest was not as pronounced, but ART exposure additionally induced ferroptosis. In this cell line, the anti-tumor activity of ART was much stronger than in the other three cell lines where ferroptosis was not induced by ART. p53 was only detectable in the KTCTL-26 cells, possibly making it a predictive marker for ferroptosis and a better response to ART. Since RCC exhibits intra-tumor heterogeneity, this might be a clinically relevant aspect. The results presented here suggest that ART may hold promise as a new additive therapy option for selected patients with advanced and even sunitinib-resistant RCC.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/11/3150/s1, Figure S1: Detailed information about Figure 6—Protein expression profile of cell cycle regulating proteins in parental and resistant Caki-1, Figure S2: Detailed information about Figure 7—Protein expression profile of cell cycle regulating proteins in parental and resistant 786-O, Figure S3: Detailed information about Figure 9f—GPX4 expression in parental and sunitinib-resistant KTCTL-26, Figure S4: Detailed information about Figure 9h—Protein expression of p53 in parental and resistant Caki-1, 786-O, KTCTL-26, and A-498 cells.

**Author Contributions:** Conceptualization, E.J.; methodology, S.D.M., J.L., P.S. and K.S.S.; software, S.D.M. and O.V.; validation, O.V., R.M. and T.E.; formal analysis, S.D.M. and O.V.; investigation, S.D.M., O.V., E.J. and T.E.; resources, E.J. and A.H.; data curation, R.M.; writing—original draft preparation, S.D.M.; writing—review and editing, E.J., A.H. and T.E.; visualization, S.D.M.; supervision, E.J.; project administration, E.J.; funding acquisition, E.J. All authors have read and agree to the published version of the manuscript.

**Funding:** This research was funded by the Friedrich-Spicker-Stiftung. Grant number: 5.

**Acknowledgments:** The main portion of the results presented here stem from work connected with the PhD thesis of S.D.M. Some elements stem from the bachelor thesis of J.L.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Predicting Response to Immunotherapy in Metastatic Renal Cell Carcinoma**

## **Matthew D. Tucker and Brian I. Rini \***

Department of Medicine, Division of Hematology and Oncology, Vanderbilt University Medical Center, Nashville, TN 37232, USA; matthew.tucker@vumc.org

**\*** Correspondence: brian.rini@vumc.org; Tel.: +1-615-875-4547

Received: 24 August 2020; Accepted: 16 September 2020; Published: 18 September 2020

**Simple Summary:** Immunotherapy-based treatment options have become standard of care in metastatic renal cell carcinoma. Despite significant improvement in overall survival with these therapies, the tumors of many patients will eventually progress. This review highlights the ongoing efforts to develop biomarkers to help predict which patients are most likely to benefit from treatment with immunotherapy.

**Abstract:** Immunotherapy-based combinations, driven by PD-1, PD-L1, and CTLA-4 inhibitors, has altered the treatment landscape for metastatic renal cell carcinoma (RCC). Despite significant improvements in clinical outcomes, many patients do not experience deep or lasting benefits. Recent efforts to determine which patients are most likely to benefit from immunotherapy and immunotherapy-based combinations have shown promise but have not yet affected clinical practice. PD-L1 expression via immunohistochemistry (IHC) has shown promise in a few clinical trials, although variations in the IHC assays as well as the use of different values for positivity presents unique challenges for this potential biomarker. Several other candidate biomarkers were investigated including tumor mutational burden, gene expression signatures, single gene mutations, human endogenous retroviruses, the gastrointestinal microbiome, and peripheral blood laboratory markers. While individually these biomarkers have yet to explain the heterogeneity of treatment response to immunotherapy, using aggregate information from these biomarkers may inform clinically useful predictive biomarkers.

**Keywords:** biomarkers; immunotherapy; renal cell carcinoma; PD-L1

#### **1. Introduction**

An estimated 400,000 new renal cancers are diagnosed annually world-wide leading to over 175,000 deaths [1]. Early systemic therapies designed to target the immunogenicity of metastatic renal cell carcinoma (mRCC), such as interferon-alpha and high-dose IL-2, were effective in only a small percentage of patients [2,3]. While subsequent therapies designed against angiogenesis including tyrosine kinase inhibitors (TKI) targeting vascular endothelial growth factor (VEGF) and its receptor (VEGFR) improved response rates and progression-free survival, nearly all patients developed resistance [4].

The implementation of monoclonal antibodies against the immune checkpoint proteins programmed cell death 1 (PD-1), programmed death-ligand 1 (PD-L1), and anti-cytotoxic T-lymphocyte-associated protein-4 (CTLA-4) has dramatically changed the treatment paradigm for mRCC [5]. After demonstrating improved overall survival (OS) compared to the mammalian target of rapamycin (mTOR) inhibitor everolimus in the post-VEGF-R inhibitor setting, nivolumab (anti PD-1) became the first immune checkpoint inhibitor to gain FDA approval for advanced RCC in November of

2015 [6,7]. Subsequently, in April 2018, the immunotherapy combination nivolumab plus ipilimumab (anti-CTLA-4) gained approval in the first-line setting after demonstrating improved OS versus sunitinib [8]. In April and May of 2019, two additional immunotherapy-based combinations were approved in the first-line setting: pembrolizumab (anti-PD-1) plus the anti-VEGFR agent axitinib and avelumab (anti-PD-L1) plus axitinib [9–11]. Despite these advances, only a minority of patients treated with immunotherapy will have a durable response, prompting the search for predictive biomarkers. Since the early phases of development of immunotherapy in mRCC, tremendous efforts have been made towards understanding the biology of the tumor microenvironment (TME) to help identify candidate biomarkers, such as immunohistochemistry (IHC) expression of PD-L1, tumor mutational burden (TMB), polybromo-1 gene (*PBRM1*) mutations, human endogenous retroviruses (hERVs), gastrointestinal microbiota, sarcomatoid histology, and the neutrophil to lymphocyte ratio (NLR).

#### **2. Programmed Death-Ligand 1**

Expression of PD-L1 (historically denoted as B7 homolog 1) on tumor cells and tumor-infiltrating lymphocytes was initially shown to be a poor prognostic marker for patients with renal cell carcinoma based on IHC analyses performed in 2004 [12,13]. Furthermore, in 2006 Thompson et al. performed a retrospective analysis of over 300 patients with mRCC and found that the 5-year cancer-specific survival rate was 42% for patients expressing PD-L1 versus 83% for patients who were negative [14]. Subsequently, a post-hoc analysis of the phase III trial COMPARZ, comparing efficacy of pazopanib to sunitinib, found that patients treated with either agent had significantly worse OS and progression-free survival (PFS) if they were PD-L1+ compared to those who were PD-L1− [15]. Thus, tumor PD-L1 expression is a negative prognostic factor in RCC and predicts against response to anti-VEGFR therapy.

Early phase clinical trials with anti-PD-1 monotherapy showed potential for the use of PD-L1 expression as a predictive biomarker for immunotherapy in mRCC [16,17]. The phase III clinical trial CheckMate 025 demonstrated improved efficacy of nivolumab over everolimus regardless of PD-L1 status, i.e., the marker was prognostic but not predictive [6]. Interestingly, the prior association of worse prognosis was observed in both groups, as patients who were PD-L1+ had numerically lower OS compared with PD-L1- patients. PD-L1 expression has been evaluated in several randomized clinical trials (Table 1).

In the phase III CheckMate 214 trial, evaluating nivolumab in combination with ipilimumab versus sunitinib, 91% (1002/1096) of patients in the intention-to-treat (ITT) population had quantifiable tumor tissue available for PD-L1 testing [8]. Tumors were positive if they had tumor cells (from baseline tumor samples prior to therapy) with > 1% PD-L1 expression as assessed using the Dako PD-L1 IHC 28-8 pharmDx test. Multivariate analysis of baseline factors was presented in the 32-month extended follow-up report [18] and showed that PD-L1 expression was a negative predictor for survival among patients treated with sunitinib [hazard ratio (HR) 0.70; 95% CI 0.52-0.93 for patients negative for PD-L1 expression]. However, PD-L1 was not associated with survival among patients treated with nivolumab plus ipilimumab even in univariate analysis, suggesting that combination immunotherapy was able to overcome the negative prognostic effects associated with PD-L1 expression.


**Table 1.**Clinical outcomes by PD-L1 expression status from phase III clinical trials of immunotherapy in mRCC.

\* reported HR is not statistically significant

Among patients with International Metastatic RCC Database Consortium (IMDC) favorable-risk disease, only 11% (11/115) of those treated with nivolumab plus ipilimumab and 12% (13/111) of those treated with sunitinib were PD-L1+ compared to 26% (100/384) and 29% (114/392) of patients with IMDC intermediate- and poor-risk disease [8]. Exploratory analysis according to PD-L1 expression was performed in the intermediate- and poor-risk patient population. The median PFS for PD-L1+ patients was 22.8 months with nivolumab plus ipilimumab versus 5.9 months with sunitinib (HR 0.46; 95% CI, 0.31–0.67), while the median PFS between nivolumab plus ipilimumab versus sunitinib was not significantly different among those who were PD-1L− (HR 1.00; 95% CI, 0.80–1.26). While overall survival was significantly longer with nivolumab plus ipilimumab versus sunitinib in both the PD-L1+ and negative groups, the degree of improvement in overall survival was greater in the PD-L1+ patients: HR 0.45 (95% CI, 0.29–0.71) in the PD-L1+ group and HR 0.73 (95% CI, 0.56–0.96) in the PD-L1− population. Additionally, the difference in ORR between nivolumab plus ipilimumab versus sunitinib was numerically higher in the PD-L1+ group with ORR of 58% with nivolumab plus ipilimumab versus 22% with sunitinib (*p* < 0.001), compared with 37% versus 28% (*p* = 0.03) in the PD-L1− group. Complete responses were also more frequent in the PD-L1+ group with 16% CR with nivolumab plus ipilimumab versus 1% with sunitinib, compared to 7% and 1% among respective PD-L1− patients. Thus, PD-L1 expression enriches for clinical benefit with combination nivolumab plus ipilimumab but cannot be used as a predictive biomarker given the significant benefit observed in the PD-L1− group.

The role of PD-L1 expression has also been explored in combinations of anti-VEGF therapy with immunotherapy. IMmotion150, a randomized phase II trial, investigated the clinical activity of atezolizumab with or without bevacizumab against sunitinib in patients with treatment-naive mRCC [21]. This trial included numerous ancillary biomarker investigations, including PD-L1 expression. Co-primary end points were PFS in both the ITT and in the PD-L1+ patient populations. PD-L1 was measured using the Ventana SP142 IHC assay, and PD-L1 was considered positive if >1% tumor-infiltrating immune cells (ICs) expressed PD-L1. The percentage of patients considered PD-L1+ among the three treatment groups were 59% sunitinib, 52% atezolizumab, and 50% atezolizumab + bevacizumab. The initial stratification was based on PD-L1 status of >5% (instead of >1%), which is thought to explain some of the imbalance among the treatment arms. The median PFS in the PD-L1+ population was 14.7 months with atezolizumab + bevacizumab versus 7.8 months with sunitinib (HR 0.64; 95% CI, 0.38–1.08), while the median PFS in the ITT population was 11.7 months with atezolizumab+bevacizumab versus 8.4 months with sunitinib (HR 1.00; 95% CI, 0.69–1.45). Furthermore, the hazard ratios for improvement in PFS were numerically improved with increasing levels of PD-L1 expression among patients treated with atezolizumab plus bevacizumab. Thus, PD-L1 expression enriched for response to this combination, although the overall activity of this regimen is lower compared to other immunotherapy-based doublets in mRCC.

The randomized phase III trial, IMmotion 151 further explored these findings. IMmotion151 enrolled patients with clear cell or sarcomatoid histology randomized to atezolizumab plus bevacizumab versus sunitinib [20]. Co-primary end points included investigator assessed progression-free survival in PD-L1+ patients and overall survival in the intention-to-treat population. PD-L1 was measured using the Ventana SP142 IHC assay, and PD-L1 was considered positive if >1% tumor-infiltrating immune cells (ICs) expressed PD-L1. Among patients in the PD-L1+ subset, median PFS was 11.2 months in the atezolizumab plus bevacizumab arm compared with 7.7 months in the sunitinib arm; HR 0.74 (95 CI 0.57–0.96; *p* = 0.0217). Similar to IMmotion 150, the HRs for PFS were numerically improved with increasing levels of PD-L1 expression. However, overall survival in the ITT population did not cross the prespecified significance boundary, with median overall survival HR 0.93 (0.76–1.14; *p* = 0.4751) at interim analysis. The HR for median overall survival in the PD-L1+ patients was 0.84 (0.62–1.15; *p* = 0.2857). ORR was 43% (76/178) in the PD-L1+ atezolizumab plus bevacizumab group compared to 35% (64/184) in the PD-L1+ sunitinib group, per investigator assessment. This difference was not seen among the PD-L1- group; 33% (90/276) for atezolizumab plus bevacizumab versus 32% (89/276) for

sunitinib. For investigator assessed PD-L1+ patients, CR was 9% with atezolizumab plus bevacizumab vs. 4% with sunitinib, while CR was only 3% and 1% in the respective PD-L1− groups.

In the phase III KEYNOTE-426 trial [9], in which the combination pembrolizumab plus axitinib was compared to sunitinib, PD-L1 expression was not incorporated into the primary endpoint; however, PD-L1 expression was tested and reported in the exploratory analysis. Expression was assessed using the PD-L1 IHC 22C3 pharmDx assay (Agilent Technologies) and was calculated using the combined positive score [CPS; calculated as the number of PD-L1+ cells (tumor cells, lymphocytes, and macrophages) divided by the total number of tumor cells, multiplied by 100]. Seventy-seven percent of patients (822/1062) had tumor samples evaluable for PD-L1 expression, and of these 60.5% had a combined positive score >1. The 12-month OS rates among PD-L1+ patients were 90.1% with pembrolizumab plus axitinib and 78.4% with sunitinib (HR 0.54, 95% CI, 0.34–0.84). In the PD-L1− group the 12-month OS rates were 91.5% versus 78.3% respectively (HR 0.59, 95% CI 0.34–1.02). The median PFS among PD-L1+ patients was 15.3 months with pembrolizumab plus axitinib versus 8.9 months with sunitinib (HR 0.62, 95% CI 0.47–0.80), and in the PD-L1- group median PFS was 15.0 months versus 12.5 months (HR 0.87, 95% CI 0.62–1.23). Given marked benefit in both PD-L1+ and PD-L1− patients over sunitinib, there was no signal for use of PD-L1 expression as a predictive biomarker for treatment with pembrolizumab plus axitinib. However, it is notable that the poor prognostic association of PD-L1 expression with sunitinib was not observed in this study as had been shown previously, with 12 months OS 78.4% in the PD-L1+ patients and 78.3% in those who were PD-L1−. This difference may be at least partly explained by the use of a different assay and different methodology, namely the combined positive score, for determining PD-L1 expression.

The phase III JAVELIN Renal 101 trial, evaluating the combination avelumab plus axitinib versus sunitinib, incorporated PD-L1 expression into the combined primary endpoints of PFS and OS among PD-L1+ patients [11]. Expression was considered positive if >1% of immune cells were positive within the sampled tumor area as assessed by the Ventana PD-L1 SP263 assay (Ventana Medical Systems). Similar to the KEYNOTE-426 trial, JAVELIN Renal 101 also had a large number (69%, 560/812) of patients with evaluable samples positive for PD-L1. With a median follow-up of at least 13 months, PD-L1+ patients had a median PFS of 13.8 months with avelumab plus axitinib versus 7.0 months with sunitinib (HR 0.62, 95% CI 0.49–0.77) compared with 13.3 months versus 8.0 months (HR 0.69, 95% CI, 0.57–0.83) in the ITT group [19]. The overall CR rate was 3.8% with the combination avelumab plus axitinib, and 15 of these 17 patients with CR were PD-L1+. The CR rate was only 2.0% overall in the sunitinib group. Interestingly, the majority of these patients (7/9) were also PD-L1+. A new analysis of JAVELIN Renal 101 reassessed PD-L1 expression using the percentage of tumor cell positivity and found only 27% (218/812) of patients had expression >1% [22], and by using this approach 92% (196/212) would have also been considered positive using the immune cell algorithm. While there was no difference in PFS among the avelumab plus axitinib group (HR 0.89; 95% CI 0.65–1.22), PFS was shorter among PD-L1+ patients in the sunitinib arm (HR 1.57; 95% CI 1.16–2.14). Increasing the expression cutoff to 5%, 10%, and 25% did not lead to statistical difference among the avelumab plus axitinib group. Similar to KEYNOTE-426, PD-L1 status alone does not appear to predict response to immunotherapy in combination with axitinib.

One of the biggest drawbacks for PD-L1 as a predictive biomarker, is the variety of available tests and different methodologies for determining positivity. New biomarker analysis from CheckMate 214 presented at ASCO 2020 compared the previously reported PD-L1 expression data as defined using tumor cell expression >1% to the combined positive score [23]. Most notably, the percentage of patients determined to be PD-L1+ increased from 23% (113/498) to 61% (298/487) in the nivolumab plus ipilimumab group and from 25% (125/494) to 60% (298/493) in the sunitinib group, comparable to KEYNOTE-426 and JAVELIN Renal 101. The combination nivolumab plus ipilimumab significantly improved OS compared to sunitinib in both PD-L1+ and PD-L1− patients regardless of which test was used. However, stratified overall survival within the nivolumab plus ipilimumab group by PD-L1 combined positive score was not reported. Therefore, it remains unclear at this time whether the

enrichment for response seen with PD-L1 expression (as initially reported using >1% positive tumor cells) remains using the combined positive score with this immunotherapy combination.

#### **3. Tumor Mutational Burden**

Tumor mutational burden (TMB) has been theorized to predict response to immunotherapy given increased formation of neoantigens on the tumor surface which lead to enhanced immunogenicity [24]. In June of 2020, the FDA announced a tumor-agnostic approval for pembrolizumab in patients whose tumors harbor TMB > 10 mutations per megabase (muts/Mb) [25], though the utility of TMB for predicting response to immunotherapy in RCC remains unproven.

Genomic profiling on over 1600 tumor samples from a variety of solid tumor types was performed using the MSK-IMPACT assay to examine the association between TMB and response to immunotherapy [26,27]. Given the heterogeneity of TMB between different histologies [28], TMB was analyzed using pre-specified cutoff percentages within each histology. Using a binary cutoff of 20%, a significant improvement in OS was observed across the entire cohort; HR 0.061 (*p* = 1.3 × 10−<sup>7</sup> ). Patients with RCC made up 9% (*N* = 151) of the cohort. When limiting the analysis to this subgroup, no significant difference was found in OS between the patients in the top 20% of TMB and those below (cutoff 5.9 muts/Mb), HR 0.569. Using a more stringent cutoff of 10% (7.9 muts/Mb) or a more inclusive cutoff of 30% (cutoff 4.9 muts/Mb), a difference in OS was still not found.

Numerous retrospective analyses in RCC evaluating TMB have since been conducted with little to no association found. Labriola et al. evaluated 34 patients with mRCC treated with immunotherapy (32 with nivolumab, 2 with nivolumab plus ipilimumab, and 1 with pembrolizumab) who underwent genomic profiling with the PGDx elio panel [29]. Patients were grouped as either progressive disease or disease control (defined as stable disease or partial response). There was no significant difference observed in the TMB score between the two groups (*p* = 0.7682), with a mean TMB of 3.01 muts/Mb among the progressive disease group versus 2.63 muts/Mb in the disease control group. There were three patients who had a TMB score > 10 muts/Mb; two had progressive disease and one was in the disease control group.

Wood et al. examined a cohort of 431 patients with melanoma, non-small cell lung cancer, and RCC who had publicly available whole exome sequencing [30]. They determined TMB status based on consensus calls in DNA variants. Overall survival data was available for 56 patients with RCC, 50 of whom had reported response data available (excluding combination immunotherapy). Separating patients into two binary groups: TMB high (defined as those exceeding the disease-matched 80th percentile) and TMB low, there was no significant difference in overall survival (*p* > 0.05). Using logistic regression to evaluate for response probability and TMB, they found that TMB was a partial predictor of response in melanoma and non-small cell lung cancer but they found no significant difference among patients with RCC (*p* = 0.894).

Dizman et al. evaluated 91 patients at the City of Hope Comprehensive Cancer Center with mRCC who had undergone genomic profiling with DNA whole exome sequencing and RNA next-generation sequencing using the GEM ExTra assay [31]. Only patients whose genomic profiling was performed prior to initiation of systemic treatment were included for analysis. One cohort of patients were treated with immunotherapy (*N* = 32) and the other with VEGF-TKI therapy (*N* = 43). Eleven patietns (34%) patients in the immunotherapy cohort were treated with first-line nivolumab plus ipilimumab, while the remaining patients were treated with nivolumab monotherapy in either the second- or third-line settings. Patients were defined as with clinical benefit if they achieved complete or partial response of any duration or stable disease for at least six months. Overall, the median TMB was low at 1.2 muts/Mb (range 0.03–4.0) and no significant difference was seen between patients with clinical benefit versus no clinical benefit in either the immunotherapy cohort (*p* = 0.82) or the VEGF-TKI cohort (*p* = 0.091).

Braun et al. performed extensive genomic analyses on tumor samples from patients enrolled on the randomized phase III CheckMate 025 trial, treated with nivolumab monotherapy or the mTOR inhibitor everolimus and on the phase II CheckMate 010 trial [32]. The data were combined with existing whole exome sequencing and RNA-seq data from CheckMate 009 [33]. Whole exome sequencing data was available for 261 patients treated with nivolumab and 193 patients treated with everolimus. Clinical benefit was defined as having complete or partial response or stable disease with tumor shrinkage and PFS of at least six months, and they calculated TMB as the calculated sum of all non-synonymous mutations in each sample. No differences were observed in the total mutation burden between the clinical benefit group (*N* = 78) and the no clinical benefit group (*N* = 95); *p* = 0.81.

Interestingly, Huang et al. performed analysis on available somatic mutation data and transcriptome profiles from patients with clear cell RCC in the TCGA cohort (*N* = 537) [34]. TMB was defined as the total number of variants divided by the whole length of exons, 38 million (including base substitutions, deletions, and insertions). Pearson correlation analysis was used to evaluate expression of PD-1, PD-L1, and CTLA-4 with TMB. While no significant association was found between TMB and CTLA-4 (*p* = 0.270) or PD-1 (*p* = 0.493), a significant negative correlation between TMB and PD-L1 expression was determined (R = −1.51 and *p* = 0.006). Analysis from CheckMate 214 presented at ASCO 2020, showed no difference in PFS or OS between high TMB and low TMB within either the nivolumab plus ipilimumab arm or in the sunitinib arm [23], and recently published data from JAVELIN Renal 101 showed that TMB did not differentiate PFS in either the avelumab plus axitinib group (HR 1.09; 95% CI 0.79–1.50) or in the sunitinib group (HR 0.79; 95% CI 0.60–1.05 [22]. Overall, despite recent tumor-agnostic FDA approval for immunotherapy in tumors with TMB > 10 muts/Mb [25], TMB does not appear to reliably predict response in mRCC.

#### **4. RNA Gene Expression**

Gene expression profiling using RNA sequencing was evaluated in several randomized control trials (Table 2). McDermott et al. conducted pre-specified exploratory genomic analysis of IMmotion 150, the phase 2 trial of atezolizumab + bevacizumab versus sunitinib, and atezolizumab monotherapy versus sunitinib [21]. Gene expression analysis was performed by generating whole transcriptome profiles for 263 patients using RNA sequencing, TruSeq RNA Access (Illumina). Gene expression signatures previously found to be associated with angiogenesis (*VEGFA, KDR, ESM1, PECAM1, ANGPTL4,* and *CD34*), immune activation (*CD8A, EOMES, PRF1, IFNG,* and *CD274*), and myeloid inflammation (*IL-6, CXCL1*, *CXCL2*, *CXCL3*, *CXCL8*, and *PTGS2*) were used to group patients into high and low expression categories for each signature, separated by the median expression score derived for each group [35–41]. They found that the AngioHigh subgroup had increased vascular density as determined by CD31 IHC, and that the Teff High subgroup was associated with increased expression of PD-L1 on immune cells by IHC and with increased CD8+ T-cell infiltration.




#### **Table 2.** *Cont.*


**Table 2.** *Cont.*

When evaluated within the sunitinib treatment arm, high angiogenesis gene expression was associated with improved overall response (46% in AngioHigh versus 9% in AngioLow) and PFS (HR 0.31; 95% CI, 0.18–0.55). While there was no difference in PFS among patients within the AngioHigh subgroup, whether evaluated between atezolizumab plus bevacizumab versus sunitinib or between atezolizumab monotherapy versus sunitinib, there was an improvement in PFS observed among patients in the AngioLow subgroup who were treated with atezolizumab plus bevacizumab versus sunitinib (HR 0.59; 95% CI, 0.35–0.98).

Within the atezolizumab plus bevacizumab arm, high immune gene expression was associated with improved overall response (49% in Teff High versus 16% in Teff Low) and PFS (HR 0.50; 95% CI, 0.30–0.86). Additionally, when evaluated across treatment arms, Teff High was associated with longer PFS with atezolizumab plus bevacizumab versus sunitinib (HR 0.55; 95% CI, 0.32–0.95).

High expression of genes involved in myeloid inflammation was associated with reduced PFS within the atezolizumab monotherapy arm (HR 2.98; 95% CI, 1.68–5.29) but not within the sunitinib arm. To further investigate the impact of myeloid inflammation and response to immunotherapy, the investigators examined the subgroup of patients with Teff High and MyeloidHigh tumors. Within this subgroup, patients treated with atezolizumab plus bevacizumab showed improved PFS versus those treated with atezolizumab alone (HR 0.25; 95% CI, 0.10–0.60). Interestingly, among patients in the Teff HighMyeloidLow subgroup, no significant differences were seen between the atezolizumab plus bevacizumab arm and the atezolizumab monotherapy arm. Overall, while these findings require further validation, they suggest that the addition of anti-VEGF treatment to immunotherapy, may help mediate some of the immunosuppressive effects of myeloid inflammation and may provide further insight into the efficacy of other anti-VEGF plus immunotherapy combinations, such as pembrolizumab or avelumab in combination with axitinib. Additionally, they support that expression of angiogenesis genes increase tumor susceptibility to sunitinib and that immune gene expression is associated with response to immunotherapy.

These genomic profiles were further validated in the prospective randomized phase III clinical trial, IMmotion151 and presented at ESMO 2018 [42]. RNA sequencing was performed on 823 patients. Patients with Teff High had improved PFS with atezolizumab plus bevacizumab compared to sunitinib (HR, 0.76; 95% CI 0.59–0.99). While AngioHigh was associated with improved PFS within the sunitinib arm (HR, 0.59; CI 0.47–0.75), there was no significant difference in PFS across treatment arms. Notably, AngioHigh expression was more prevalent among patients with favorable risk as compared with intermediate/poor-risk, and Teff High was more frequent among patients in the intermediate/poor-risk group, providing a biologic correlate of the differential clinical effects observed in CheckMate 214.

RNA expression profiling was also prospectively evaluated in the phase III clinical trial JAVELIN Renal 101 [11,22,43]. Researchers created whole transcriptome profiles using RNA sequencing on 720 baseline tumor samples and developed a new gene expression signature, Renal 101 Immuno, derived from 26 genes involved in T-cell proliferation, natural killer cell activation, interferon gamma signaling, and others. Using this signature, patients treated with avelumab plus axitinib who had high levels of expression (at or above the median level of expression) had significantly longer PFS compared

with patients with low levels of expression (HR, 0.60; 95% CI, 0.44–0.83). Evaluating this signature in an independent dataset from the phase 1b JAVELIN 100 trial [44], high expression was again associated with prolonged PFS (HR 0.36; 95% CI 0.16–0.81). Using the gene expression signatures from the previous IMmotion studies, they also showed that the AngioHigh signature was again associated with improved PFS within the sunitinib arm but was not significantly different between the avelumab + axitinib arm versus the sunitinib arm. Among patients with AngioLow gene expression, PFS was significantly longer in the avelumab + axitinib arm compared with sunitinib. These data reinforce data from IMmotion 150 that Angiolow patients have better outcome with an immunotherapy-based regimen versus sunitinib monotherapy.

The RNA expression profiles from IMmotion150 and JAVELIN Renal 101 were further examined in a pooled analysis of available data from CheckMate 009, 010, and 025 [32]; however, no associations between high expression of any gene signature and improved clinical benefit, PFS, or OS were observed. One potential explanation for this difference is that the majority of patients included were treated with nivolumab in the second-line setting after prior anti-VEGF therapy, whereas patients in both IMmotion studies and JAVELIN Renal 101 were treatment-naïve. A multicenter retrospective analysis of 86 patients with mRCC treated with immunotherapy evaluated both a large T-effector gene panel and a smaller 5-Gene panel (*FOXP3*, *CCR4*, *KLRK1*, *ITK*, and *TIGIT*) [45]. While there was no difference observed between high and low expression of the larger T-effector panel, there was a significant difference in the ORR between the cohort with high 5-Gene expression versus low [31% (14/45) versus 2% (1/41); *p* = 0.001].

Biomarker data from CheckMate 214 presented at ASCO 2020 also reported the breakdown of six different gene expression signatures [23]. Twenty percent (109/550) of patients in the nivolumab plus ipilimumab arm and 19% (104/546) of patients in the sunitinib arm had tumor tissue evaluable to perform RNA sequencing. While AngioHigh score (as per IMmotion150) was significantly associated with improved PFS within the sunitinib arm, no other observed significant differences were observed between the remaining gene expression signatures. Of note, this was the first reported study to evaluate these signatures with the use of an anti-CTLA-4 agent in combination with anti-PD-1; additionally, the percentage of patients with tumor evaluable for testing was low. However, when dichotomizing patients to PFS < 18 months versus > 18 months, they found differences in several HALLMARK [46] gene signatures: TNFalpha signaling via NFkB, epithelial mesenchymal transition, and inflammatory response. Gene expression signatures have successfully defined several subtypes of RCC related to varying degrees of immune involvement and angiogenesis; however, these signatures require further prospective validation prior to clinical use as predictive biomarkers.

#### **5.** *Polybromo-1* **Mutations**

In addition to mutations in the *von Hippel-Lindau* (*VHL*) gene, the pathogenesis of clear cell RCC includes a several secondary mutations including in *Polybromo-1* (*PBRM-1*), which has recently been implicated as a potential biomarker for immunotherapy [47]. Whole exome sequencing was performed to analyze pre-treatment tumor samples from 35 patients with mRCC treated with nivolumab on the prospective open-label phase I study, CA209-009 [33]. Patients were grouped into three different response categories for analysis: clinical benefit (patients with complete or partial response along with patients with stable disease if they had any objective reduction in tumor size lasting at least six months), no clinical benefit (patients with progressive disease leading to treatment discontinuation within three months), and intermediate benefit (patients who did not fit into the clinical benefit or no clinical benefit categories).

Truncating or loss of function mutations in *PBRM1* were more frequent in the clinical benefit group (9/11) compared with the no clinical benefit group (3/13, *p* = 0.012) with an odds ratio for clinical benefit of 12.93 (95% CI 1.54–190.8). OS and PFS were both significantly improved in patients with *PBRM1* loss of function (*N* = 19) compared to those with *PBRM1* intact (*N* = 16); *p* = 0.0074 and *p* = 0.29 respectively. They also evaluated *PBRM1* loss of function in an additional 63 patients who were treated with anti-PD-1(L1) therapy either alone or in combination with anti-CTLA-4 therapy. Again, tumors from patients deriving clinical benefit were more likely to harbor loss of function mutations in *PBRM1* (17/27) compared to those with *PBRM1* intact (4/19, *p* = 0.0071) with an odds ratio for clinical benefit of 6.10 (95% CI 1.42–32.64).

Braun et al. sought to validate these findings using archival tumor tissue from patients treated with nivolumab or everolimus from the randomized phase III trial, CheckMate 025 [48]. Of note, archival specimens were obtained prior to any treatment (including anti-VEGF therapy.) *PBRM1* mutations were identified in 29% (55/189) treated with nivolumab and in 23% (45/193) of patients who received everolimus. Among those treated with nivolumab, *PBRM1* mutations were present in 39% (15/38) of responding patients (either complete or partial response) compared to 22% (16/74) of non-responding patients (odds ratio 2.34, *p* = 0.04). Overall survival (HR, 0.65; *p* = 0.03) and progression-free survival (HR, 0.067; *p* = 0.03) were both associated with the presence of *PBRM1* mutations. However, among those treated with everolimus there was no significant difference between responders who harbored *PBRM1* mutations (1/5) and non-responders who harbored *PBRM1* mutations (10/56; *p* = 0.64). There was no significant association of *PBRM1* mutation with OS (HR, 0.81; *p* = 0.27) or PFS (HR 0.83; *p* = 0.32) among those treated with everolimus.

The association of *PBRM1* mutations with anti-PD-1 therapy was further investigated using pooled data from patients treated with nivolumab in either CheckMate-009, CheckMate-010, or CheckMate-025 who underwent whole exome sequencing [32]. Collectively, there was a significant benefit in OS and PFS for patients harboring *PBRM1* mutations (*p* < 0.001 and *p* = 0.0056 respectively). They also evaluated the presence of *PBRM1* mutations along with the degree of T-cell infiltration present in the tumor. Using CD8 immunofluorescence on 153 tumor samples from nivolumab treated patients, they quantified the density of CD8+ cells in the tumor center and at the tumor margin. Tumors were classified as "immune excluded" if at least five-fold more CD8+ cells were present in the tumor margin compared to the tumor center, "immune desert" if they were not "excluded" but still had below the 25th percentile of CD8+ cells in the tumor center, and "immune infiltrate" if they were not "excluded" and had greater than the 25th percentile of CD8+ cells in the tumor center. The majority (73%) of samples were classified as "immune infiltrated." While there was no significant association between the degree of tumor infiltration and clinical benefit, there was an association observed between the presence of *PBRM1* mutations and lower T-cell infiltration (*p* = 0.013). *PBRM1* mutations were detected in 47% of immune "deserts" and 29% of immune "excluded" tumors, but only 22% of immune "infiltrated' tumors (*p* = 0.01 for non-infiltrated versus infiltrated tumors).

However, there are also data to suggest that *PBRM1* mutations are associated with an immunosuppressive and pro-angiogenesis tumor microenvironment. Using a murine model, Liu et al. observed that *PBRM1* inactivation was associated with a less immunogenic tumor microenvironment, which was validated using human gene expression data from the TCGA-KIRC dataset [49]. The IMmotion150 dataset [21] was also analyzed, showing that patients with *PBRM1* mutations had a significantly lower ORR with both atezolizumab monotherapy and atezolizumab plus bevacizumab. Gene expression data from IMmotion150, TCGA, and the International Cancer Genome Consortium showed increased angiogenesis signatures among patients with *PBRM1* mutations from all three cohorts [49]. Clinical data from mRCC patients with pancreatic metastases demonstrated *PBRM1* mutations were associated with improved response to anti-VEGF therapy, supporting that *PBRM1* mutant tumors may have a more angiogenic phenotype [50]. Additionally, biomarker data from CheckMate 214 presented at ASCO 2020 showed no significant difference in PFS or OS between *PBRM1* wild type or mutant within either the nivolumab plus ipilimumab arm or in the sunitinib arm [23]. Furthermore, analysis from JAVELIN Renal 101 showed no association of *PBRM1* with PFS in either treatment arm [22]. Given conflicting results from multiple analyses, *PBRM1* mutations are not ready for clinical use as a predictive biomarker and require further investigation to understand their role in the tumor microenvironment. Some of the discrepant results may be due to the different populations studied, such as treatment-naïve versus VEGF-refractory.

#### **6. Human Endogenous Retroviruses**

Human endogenous retroviruses (hERVs) represent a group of long terminal repeat retrotransposons that collectively make up about 8% of the human genome [51,52]. Despite being normally silenced in somatic tissue, their expression has been reported in multiple cancer types, including RCC [53]. Aberrant transcriptional activation of hERVs has been theorized to induce an antitumor immune response and up-regulation of immune checkpoint pathways, increasing sensitivity to immunotherapy [54].

Using a cohort of 24 mRCC patients, Panda et al. used RNA sequencing to investigate an association of hERV expression and response to single agent PD-1(L1) therapy [54]. The RNA level of *ERV3-2* was measured using real-time quantitative PCRs (RT-qPCRs) with two independent primers. Regardless of which primer was used, they found that *ERV3-2* expression was significantly higher in responders compared with non-responders. Patients were also classified as either *ERV3-2* high or low based on an optimal cutoff derived from the receiver operating characteristic curves. *ERV3-2* high patients were significantly more likely to respond and had significantly longer PFS.

Post-hoc analysis from patients treated with nivolumab on CheckMate 010 was performed using RT-qPCR on 99 formalin-fixed paraffin-embedded tissue (FFPE) pretreatment tumors to determine expression levels of pan-*ERVE4*, pan-*ERV3.2*, *hERV4700 GAG* or *ENV*, and the reference genes *18S* and *HPRT1* [55]. Patients were dichotomized as high or low expression using the 25th percentile as the cutoff. Using this cutoff, only *hERV4700 ENV* was significantly associated with PFS and response. PFS was 7.0 months among the high-*hERV4700 ENV* group versus 2.6 months among the low expression group (*p* = 0.010), and the ORR for high-*hERV4700 ENV* was 35.6% versus 12.5% (*p* = 0.036).

Using pooled data from CheckMate 009, CheckMate 010, and CheckMate 025, Braun et al. found that hERV expression determined using RNA sequencing correlated with expression obtained by RT-qPCR; however, the authors note that *ERV3-2* expression was not reliably inferred using this technique, which highlights a potential limitation of using RNA sequencing to measure hERV expression in FFPE tissue [32]. Despite this limitation, they did find two hERVs (*ERV2282* and *EVR3382*) that were weakly associated with response, PFS, and OS when using expression as a continuous variable; however, when divided into high and low expression (based on median expression of each hERV) these were no longer significant for both improved PFS and OS. hERV expression presents a relatively new candidate biomarker that requires further prospective validation as well as improved reproducibility of technique.

#### **7. Gastrointestinal Microbiome**

Recent studies have shown a link between the intestinal microbiome and cancer immunosurveillance, including a role in response to immunotherapy [56–58]. A small Japanese study evaluated 31 patients with mRCC treated with immunotherapy and retrospectively separated them by antibiotic use within 30 days of treatment or not [59]. The median PFS for patients treated with antibiotics (*N* = 5) was 2.8 months compared with 18.4 months (*p* < 0.001) in the group without antibiotics (*N* = 26). Subsequently, Lalani et al. performed a retrospective analysis on two cohorts to explore the association of antibiotic use and response to immunotherapy in mRCC: the first, a single center cohort of patients who received anti-PD-1(L1) therapy (*N* = 146), and the second, a trial-database from patients treated with interferon, anti-VEGF therapies, or mTOR inhibitors (*N* = 4144) [60]. Antibiotic use was defined as anytime from 8 weeks before the start of therapy through 4 weeks after initiation. In the anti-PD-1(L1) cohort, patients with antibiotic exposure (*N* = 31) had a significantly lower ORR (12.9% versus 34.8%, *p* = 0.026) and shorter PFS (HR, 1.96, 95% CI 1.20–3.20, *p* = 0.007) compared to the group without antibiotic exposure (*N* = 115). In the trial-database cohort ORR was significantly lower (19.3% versus 24.2%, *p* = 0.005), PFS significantly shorter (HR 1.16, 95% CI 1.04–1.30), and OS significantly shorter (HR 1.25, 95% CI 1.10–1.41) in the antibiotic group compared with the no antibiotic group. However, in subgroup analysis the authors note that the difference in the trial-database group was driven by patients treated with interferon (*N* = 510) and prior cytokine therapy (*N* = 520), while no

difference in OS was observed between antibiotic users in either the anti-VEGF (no prior cytokines) group (*N* = 2454) or in the mTOR group (*N* = 660).

Baseline fecal samples were collected from patients with mRCC treated with nivolumab on the NIVOREN GETUG-AFU 26 phase 2 clinical trial to investigate the relationship between the microbiome and response to immunotherapy [61]. Patients were dichotomized by prior antibiotic exposure, namely those who had received antibiotics within the two months prior to treatment with nivolumab (*N* = 11) and those without prior antibiotic exposure (*N* = 58). The ORR was lower in the antibiotic group at 9% versus 28% in the group without prior antibiotics (*p* < 0.03). Median PFS and OS were also longer in the no antibiotic group (5 months and NR) compared with the antibiotic group (2 months and 25 months; *p* = 0.03 and *p* = 0.04). They subsequently performed analysis of the relative taxonomic abundance for prevalent fecal bacteria between the two groups and found overrepresentation of *Eubacterium rectale* (*p* = 0.02) in the no antibiotic group and overrepresentation of *Erysipelotrichaceae bacterium* and *Clostridium hathewayi* in the antibiotic group (*p* < 0.02).

Additionally, they separated the group without prior antibiotic exposure into responders (*N* = 30) and non-responders (*N* = 28) and compared the various taxonomical fecal bacterial profiles. Patients in the responder group were more likely to harbor overrepresentation of *Akkermansia muciniphila, Bacteroides salyersiae*, and *Eubacterium siraeum* compared with non-responders. Likewise, patients in the non-responder group were more likely to have overrepresentation of *Erysipelotrichaceae bacterium, Clostridium hathewayi*, and *Clostridium clostridioforme*.

Recently, Salgia et al. performed an observational study of 31 patients with mRCC treated with either nivolumab monotherapy (77%) or nivolumab plus ipilimumab (23%) and assessed the gut microbiota composition using metagenomic sequencing [62]. Of note, patients with antibiotic exposure within 14 days were excluded. Using the Shannon diversity index, greater microbial diversity was associated with clinical benefit (*p* = 0.001). Prospective and controlled studies are warranted to further explore and validate using the microbiome to predict response to immunotherapy, as well as to explore the role of alteration of the host microbiota to enhance therapeutic response.

#### **8. Sarcomatoid Di**ff**erentiation**

Sarcomatoid differentiation, present in around 10% of patients with RCC, is associated with more aggressive clinical features, including shorter time to relapse, worse IMDC prognostic classification, and shortened PFS and OS with anti-VEGF therapies [63–65].

A comparison of 40 tumors with sarcomatoid histology present versus clear cell RCC without sarcomatoid features found that tumors with sarcomatoid histology were significantly more likely to have co-expression of PD-L1 on tumor cells and on tumor-infiltrating-lymphocytes [66], suggesting the potential for improved response to immunotherapy. In the IMmotion151 trial, 61% (86/142) of patients with sarcomatoid histology were positive for PD-L1 expression [67], and analysis presented at ESMO 2018 showed that tumors with sarcomatoid histology were more likely to be AngioLow and Teff High compared with non-sarcomatoid tumors [42]. The HR for PFS in the sarcomatoid subgroup was 0.45 (95% CI 0.26–0.77) in the PD-L1+ population and 0.52 (0.34–0.79) in the ITT population [67]. Given that PFS was improved in both the PD-L1+ and PD-L1− populations, sarcomatoid histology may be an independent predictor of response.

Post-hoc analysis of patients from CheckMate 214 identified sarcomatoid features in 16.4% (139/847) of patients with intermediate/poor-risk RCC [68]. Overall survival was significantly improved in this subgroup with nivolumab plus ipilimumab versus sunitinib (HR 0.45; 95% CI 0.3–0.7). The ORR was 60.8% with nivolumab plus ipilimumab versus 23.1% with sunitinib (*p* < 0.0001), and the CR rate was 18.9% versus 3.1% respectively. Of note, 52% (69/133) of patients with quantifiable PD-L1 expression were positive. Iacovelli et al. performed a meta-analysis of the four published phase III randomized combination immunotherapy clinical trials that included patients with sarcomatoid histology [69]. Collectively there were 467 patients included with sarcomatoid features present, 226 randomized to immunotherapy-based combination treatment and 241 randomized to sunitinib. Overall, PFS

significantly favored treatment with immunotherapy-based combinations (HR 0.56; 95% CI, 0.43–0.71). Excluding JAVELIN Renal 101 (given OS data was not mature at time of analysis), OS also significantly favored treatment with immunotherapy-based combinations (HR 0.56; 95% CI 0.43–0.82). The ORR was 52.6% for combination therapy versus 20.7% with sunitinib, with CR rates of 11.5% and 0.8% respectively. Overall, this meta-analysis supports the use of immunotherapy-based combinations in the front-line setting for patients with mRCC with sarcomatoid features, although if a specific regimen has greater effect awaits further study. Additional work exploring the role of immunotherapy-based combinations in other distinct RCC histologies is necessary.

#### **9. Neutrophil to Lymphocyte Ratio**

The baseline neutrophil to lymphocyte ratio (NLR) has been shown to have a negative prognostic association with RCC and to predict a more aggressive phenotype [70–72]. While the poor prognostic association of PD-L1 expression appears to be mitigated by immunotherapy [8–10], this does not appear to be the case for NLR.

Zahoor et al. performed a retrospective analysis of 90 patients treated with nivolumab as second-line or later therapy for mRCC. After multivariate analysis, a higher baseline NLR was associated with an increased risk of progression (HR 1.86, 95% CI, 1.05–3.29) [73]. Additionally, Lalani et al. performed a retrospective review of 142 patients treated with anti-PD-1(L1) therapies and found that baseline NLRs were significantly higher in the poor IMDC risk group compared to those with favorable or intermediate risk (*p* < 0.001) [74]. Interestingly, the results indicate that the NLR between 4 to 8 weeks post treatment initiation was a more accurate predictor of response than at baseline, and an increase in NLR > 25% was associated with shorter PFS (HR 2.60, 95% CI 1.53–4.39).

Furthermore, exploratory analysis of the JAVELIN Renal 101 trial, presented at ASCO 2020, demonstrated significantly improved OS (HR 0.51; 95% CI 0.30–0.87) and a trend toward improved PFS (HR 0.85; 95% CI 0.63–1.15) among patients treated on the avelumab plus axitinib arm who had a baseline NLR < the median compared to patients with baseline NLR > the median [75]. While the overall differences between groups is somewhat modest, the ratio is cost-effective and easily performed in clinic without the need for additional testing. Additional prospective validation is needed for pre-treatment laboratory biomarkers such as NLR.

#### **10. Conclusions**

While immunotherapy-based combinations have become the standard of care for first-line mRCC, the majority of patients will eventually progress on these regimens. Additionally, immune-related adverse events can lead to serious toxicity and have the potential to be life-threatening [76]. Therefore, the ability to predict which patients are most likely to respond would have significant clinical impact. Despite tremendous investigation on numerous candidate biomarkers, none have yet proven ready for clinical practice.

PD-L1 expression has been shown to enrich for response, most notably in CheckMate 214; however, patients with negative expression can still respond and maintain the potential for complete response. Furthermore, the commercially available anti-PD-L1 clones currently in use are highly variable, and PD-L1 expression patterns have been shown to be heterogenous throughout different tumor regions [77]. Therefore, immunotherapy should not be withheld for patients who are known to be PD-L1−. New imaging modalities are being developed to quantify PD-1 and PD-L1, which may help reduce some of the variability in future studies [78,79]. TMB, while recently approved as a tumor-agnostic biomarker for response to pembrolizumab, has been shown to be an unreliable predictor in RCC and should not be used in clinical decision making for these patients.

Gene expression using RNA sequencing has generated new understanding into the biology of RCC and patterns of response to therapy. Improvement in PFS with atezolizumab plus bevacizumab were observed in both IMmotion150 and IMotion151 among Teff High patients; however, this improvement was not observed in updated analysis of CheckMate 214. Future prospective trials should incorporate pre-specified analyses of gene expression signatures to assess their potential for clinical utility.

Individual gene alterations, such as loss of function mutations in *PBRM1*, have demonstrated mixed results, and should not be used clinically at this time. Likewise, while antibiotic exposure has been associated with decreased response to immunotherapy, these data have yet to be prospectively validated and should neither prohibit patients who have recently required antibiotics from receiving immunotherapy nor should it limit clinicians' use of antibiotics for infected patients. Given the enhanced efficacy observed among patients with sarcomatoid differentiation, these patients should receive upfront immunotherapy-based combinations instead of VEGF-TKI alone when systemic therapy is warranted.

While data regarding the NLR as a predictive biomarker is limited and is impacted by the known poor prognostic associations in RCC. However, such peripheral blood biomarkers have the benefit of being cost-effective and readily available; continued efforts toward identifying laboratory biomarkers is warranted.

In addition to developing biomarkers associated with radiographic response, there are several other facets of clinical practice that may be improved with the addition of biomarkers. Biomarkers that predict durability may help identify patients who can safely discontinue therapy after achieving a clinical response, and biomarkers that predict immune-related adverse events may help determine which patients should be observed more closely and potentially for which immunotherapy should be avoided or limited. While some candidate biomarkers may only enrich for response, these insights also help with developing novel therapies and combinations which may lead to improved outcomes with immunotherapy.

**Author Contributions:** M.D.T. and B.I.R. performed data research, wrote, and edited this review. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** Matthew D. Tucker reports no disclosures. Brian I. Rini reports: Research Funding to Institution: Pfizer, Merck, GNE/Roche, Aveo, Astra-Zeneca, BMS, Exelixis, Consulting: BMS, Pfizer, GNE/Roche, Aveo, Synthorx, Compugen, Merck, Corvus, Surface Oncology, 3DMedicines, Arravive, Alkermes, Arrowhead, GSK Stock: PTC therapeutics.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
