**Increased Intraspecies Diversity in** *Escherichia coli* **Biofilms Promotes Cellular Growth at the Expense of Matrix Production**

**Andreia S. Azevedo 1,2,3,4,\*,**† **, Gislaine P. Gerola 1,**†**, João Baptista 1,**†**, Carina Almeida 1,2,5, Joana Peres <sup>1</sup> , Filipe J. Mergulhão <sup>1</sup> and Nuno F. Azevedo <sup>1</sup>**


Received: 19 June 2020; Accepted: 14 November 2020; Published: 17 November 2020

**Abstract:** Intraspecies diversity in biofilm communities is associated with enhanced survival and growth of the individual biofilm populations. Studies on the subject are scarce, namely, when more than three strains are present. Hence, in this study, the influence of intraspecies diversity in biofilm populations composed of up to six different *Escherichia coli* strains isolated from urine was evaluated in conditions mimicking the ones observed in urinary tract infections and catheter-associated urinary tract infections. In general, with the increasing number of strains in a biofilm, an increase in cell cultivability and a decrease in matrix production were observed. For instance, single-strain biofilms produced an average of 73.1 μg·cm−<sup>2</sup> of extracellular polymeric substances (EPS), while six strains biofilms produced 19.9 μg·cm<sup>−</sup>2. Hence, it appears that increased genotypic diversity in a biofilm leads *E. coli* to direct energy towards the production of its offspring, in detriment of the production of public goods (i.e., matrix components). Apart from ecological implications, these results can be explored as another strategy to reduce the biofilm burden, as a decrease in EPS matrix production may render these intraspecies biofilms more sensitive to antimicrobial agents.

**Keywords:** biofilms; *Escherichia coli*; intraspecies community; EPS matrix; peptide nucleic acid-fluorescence in situ hybridization; urinary tract infections; catheter-associated urinary tract infections; confocal laser scanning microscopy

### **1. Introduction**

Microorganisms live in a wide variety of environments usually enclosed in communities attached to a surface. This type of community behavior can evolve towards complex multicellular structures, termed biofilms [1–3]. These are aggregates of microbial cells embedded in a matrix of extracellular polymeric substances (EPS) that provide several survival advantages, namely, nutrient capture, enzyme retention and resistance/tolerance to antimicrobial agents [4]. Within these structures, bacteria are highly sociable, communicating with each other by secreting important molecules

(e.g., cell-signaling molecules, toxins and matrix components), leading to complex interactions during biofilm development [5,6]. This social environment may involve cooperation, competition, or even both.

Microbial competition can manifest as a response to space and nutrient limitation [7,8]. During microbial competition, microorganisms can inhibit the growth of and even kill neighboring cells by secreting broad-spectrum antibiotics [9] or secreting virulent proteins and toxins [10]. However, cooperation is also very prevalent in biofilm communities [11]. For instance, microorganisms can interact to increase their resistance, increasing the tolerance of the whole consortium to antimicrobial agents [12,13], or to enhance the biofilm-forming potential of the consortium [14]. In addition, metabolic cooperation and cross-feeding have been proven to increase the overall fitness of biofilms [15,16].

Another interesting theory adapted to the microbial world and, in particular, to the biofilm field, is the "public goods" dilemma [17–19]. This theory is based on the observation that, when in society, some bacteria cooperate by secreting a product/resource that becomes available to neighboring bacteria and, hence, benefits the overall consortium. Other bacteria, named cheaters, exploit the "public goods" without contributing for its production, leaving the metabolic burden of synthesizing these molecules to the producing bacteria [20]. Hence, cheaters are likely to outcompete the "public goods" producers and decrease the overall fitness of the consortium [21].

While several examples of cooperation can be found in the biofilm literature for interspecies biofilms, there are far fewer studies with intraspecies biofilms [22–24]. Moreover, and to the authors' knowledge, the number of strains assessed in intraspecies biofilms has always been lower than three. This begs the question: is there a maximum number of strains after which a cooperative behavior in intraspecies biofilms is no longer observed? To answer this question, and as a case study, we formed intraspecies biofilms with up to six strains of *Escherichia coli* isolates from urine, and analyzed their biofilm-forming ability under conditions mimicking the urinary tract infections (UTIs) and catheter-associated urinary tract infections (CAUTIs); UTIs are the most common type of healthcare-associated infection reported. Approximately 75% of hospital-acquired UTIs are associated with CAUTIs [25].

To assess cooperative or competitive behavior, the phylogenetic relatedness between the six *E. coli* strains was investigated and correlated to their performance as biofilm producers. As the self-produced EPS matrix is the most recognizable "public good" under the microbial biofilm context [15], the EPS production was grouped according to the number of strains present in the biofilm, as an indicator of social behavior. Additionally, the spatial location of species within the biofilm architecture was determined by employing a multiplex peptide nucleic acid fluorescence in situ hybridization (PNA-FISH) and confocal laser scanning microscopy (CLSM) analysis.

### **2. Results and Discussion**

### *2.1. Single- and Multi-Strain Biofilm Growth*

This work started with the expectation that related strains would cooperate in a biofilm consortium. However, the stage of biofilm development when this behavior, if existent, would occur, was not easily predictable. Moreover, inter-experimental variability in biofilm formation could overshadow the expected changes in metabolism. To increase the chances of picking up this behavior, 63 biofilm growth experiments (Table S1) were conducted, generating an equal number of biofilm growth curves for cultivable cell counts (Figure S1) and total biomass (Figure S2). We then calculated and compared the areas below the curves of the biofilm growth from 0 to 48 h using the trapezium rule [26,27] for both cultivable cell numbers (Figure 1a) and total biomass (Figure 1b).

**Figure 1.** Biofilm formation profiles exhibited by the combination of *Escherichia coli* strains during 48 h, at 37 ◦C in AUM (artificial urine medium). Example of calculation of the area under the curve for cultivability (**a**) and total biomass (**b**). Cultivability areas showing a relationship between the number of cultivable cells and the number of strains within the biofilm (**c**). CV (crystal violet) areas show a trend of biomass reduction as the number of different strains within the biofilm increases (**d**). The results represent three independent experiments. Results are presented as the mean ± standard deviation. A—partial area; T—time; V—Log (CFUs·cm<sup>−</sup>2) value or O.D.620 nm value.

For all consortia, from 2 h to 48 h, the cultivable cell counts significantly increased over time (*p* < 0.05). In particular, for single-strain biofilms, CFUs counts averaged from 3.6 log CFUs·cm−<sup>2</sup> at 2 h to 6.5 log CFUs·cm−<sup>2</sup> at 48 h (*p* < 0.05) (Figure S1a). However, for consortia of 1, 3 and 4 strains there was not a significant difference from 24 h to 48 h in terms of cultivability (*p* > 0.05), i.e., a stationary growth phase seemed to be reached. Concerning the 6 strains biofilm, CFUs counts averaged at 2 h, 5.3 log CFUs·cm<sup>−</sup>2, significantly increased to 6.2 log CFUs·cm−<sup>2</sup> at 24 h, and 6.7 log CFUs·cm−<sup>2</sup> at 48 h, (*p* < 0.05) (Figure S1f).

When plotting the number of cultivable cells with the number of strains present in the biofilm (Figure 1c), an increase in the number of strains was accompanied with a slight increase in the number of cultivable cells. Nevertheless, there is no statistically significant difference between the cultivability of 1, 2 and 3 strains consortia (*p* > 0.05). Biofilms composed of 5 and 6 strains are statistically different from 1, 2 and 3 strains consortia (*p* < 0.05). While a slight increase in the number of cultivable cells might have been expected, since the number of cells in the initial suspension also increased with the number of strains, the total produced biomass decreased in general with an increase in the number of strains, particularly for biofilms composed of more than two strains (Figure 1d). In fact, the total biomass significantly decreased when comparing 1 and 2 strains consortia to the remaining biofilms (*p* < 0.05). This behavior is evident comparing the average optical density (O.D.620 nm) of 0.245 for single-strain and 0.048 for six-strains at 48 h (*p* < 0.05) (Figure S2). Still, an unexpected higher decrease was obtained for biofilms composed of three strains, not confirmed in later EPS quantification, which can evidence that loosely attached biofilm could have been dragged out during washing steps (i.e., EPS production still occurred but the physico-chemical forces of attachment to the surface was weaker).

For biofilms up to 5 strains, from 2 h to 48 h, the produced biomass significantly increased with time (*p* < 0.05) (Figure S2). However, when six strains were present, a significant increase in biomass production over time was not observed (*p* > 0.05) (Figure S2f). In addition, the biomass seemed to decrease for the 3 strains consortia during the incubation period (*p* < 0.05) (Figure S2c).

### *2.2. Biomass and Matrix Production as a Function of the Number of Strains in a Consortium*

A cluster analysis (Figure 2) was conducted to provide a statistical basis on the observation that an increasing number of strains in a biofilm would affect its behavior, plotting the cultivable cell counts versus total biomass (Figure 2a) and versus EPS matrix (Figure 2b). Overall, a higher number of strains led to a lower amount of biomass production (*p* < 0.05). In particular, two strains biofilms were mainly clustered as low cell numbers and high biomass producers. Biofilms composed by three strains were clustered as very low biomass producers, contradicting the overall behavior. Biofilms composed by 4, 5 and 6 strains were clustered as high cell numbers and low biomass producers (Figure 2a).

**Figure 2.** Cluster analysis for cultivability areas versus CV areas (**a**) and for cultivability areas versus EPS (extracellular polymeric substances) matrix (**b**) for 48 h-aged biofilms. Clusters are observable according to the number of strains in consortia. A trend for multi-strain biofilms to produce less EPS matrix is noticeable. The graph (**a**) is subdivided in three clusters: low cell numbers and biomass (1); high cell numbers and low biomass (2); low cell numbers and high biomass (3). The graph (**b**) is subdivided in four clusters: low cell numbers and EPS matrix (1); high cell numbers and low EPS matrix (2); high cell numbers and EPS matrix (3) and low cell numbers and high EPS matrix (4). The results represent three independent experiments. Results are presented as the mean.

To confirm these clusters, the EPS matrix was quantified for a subset of biofilms. Analyzing the clusters (Figure 2b), there is a tendency for multi-strain biofilms to produce less EPS matrix than single-strain biofilms. Biofilms composed by one and two strains were clustered as high EPS producers while biofilms with 4, 5 and 6 strains grouped as high cell numbers and low EPS producers. In fact, the produced EPS matrix decreased almost linearly with the increase in the number of strains in the consortia (Figure S3). Interestingly, for biofilms composed of three strains, the above-mentioned low biomass production was not confirmed by EPS matrix quantification by dry weight after lyophilization (Figure 2b).

Overall, this work shows that an increase in the number of *E. coli* strains in an intraspecies biofilm redirects the metabolism of the microorganisms towards offspring production at the expense of EPS matrix. However, more work is required in the future with other relevant bacterial species and other biofilm growth conditions (e.g., media culture, surfaces, hydrodynamic conditions) to investigate if the observed intraspecies phenomenon presented in this study is maintained.

### *2.3. How Long Does It Take for Microorganisms to Adapt Their Metabolism?*

To better understand when the metabolism of the microorganisms is directed towards cell growth rather than EPS matrix production during biofilm growth, a cluster analysis for each of the time intervals of the experiment was performed (Figure 3). It is clear that, for 2 h-old biofilms, the trend for multi-strain biofilms to produce less biomass can already be observed (Figure 3a).

**Figure 3.** Cluster analysis for cultivability areas versus CV areas for each time point of the experiments: 0–2 h (**a**), 2–6 h (**b**), 6–24 h (**c**) and 24–48 h (**d**). The clustering phenomenon is more evident during the first hours of biofilm development, namely up to 24 h of growth. The graph (**a**) is subdivided in three clusters: low cell numbers and biomass (1); high cell numbers and low biomass (2); low cell numbers and high biomass (3). The results represent the mean of three independent experiments.

In fact, as the biofilm matures, the biofilm clustering according to the number of strains starts to be less evident (Figure 3c,d). After two hours of biofilm growth, two strains biofilms were clustered as low cell numbers and high biomass producers, while biofilms composed by 4, 5 and 6 strains and most of the biofilms composed by 1 strain were grouped as high cell numbers and low biomass producers. This behavior is maintained for the subsequent time points, with the exception of most of the biofilms composed by 4 strains which present low cell numbers after 48 h of growth. Interestingly, after 2 h, some of the three strains biofilms are grouped together with 4 strains biofilms. It is only afterwards that they develop the less expected behavior of very low biomass production. These results indicate that the differentiated behavior occurs predominantly within the first hours of biofilm development, which is expected. In fact, *E. coli* is known to rapidly direct its metabolism when adapting to diverse and sudden stress conditions [28,29]. In a study of Drazic et al. [30], major changes in the metabolic profile of *E. coli* occurred as early as after 5 min of hypochlorite-induced stress, in terms of relative concentrations of fatty acids, amino acids, acetic and formic acid, which were readily regenerated after 40 to 60 min.

### *2.4. Impact of Phylogenetic Closeness in Biomass Production for Two-Strain Biofilms*

In theory, the microbial cooperation is promoted by high relatedness of microbial cells. To infer on this, seven housekeeping genes (*adk*, *fumC*, *gyrB*, *icd*, *mdh*, *purA* and *recA*) were sequenced in order to infer whether genetic similarity among the strains would impact the offspring and total biomass production of the overall consortia, according to the - CbI. According to this analysis, the percentage of different nucleotides among these isolates is below 1.8%. In a study of Lukjancenko et al. [31], the highest phylogenetic difference obtained, when comparing several *E. coli* strains using the same housekeeping genes, was around 2%, which seems to be in agreement with our findings.

Due to the lack of adequate methods to statistically assess the phylogenetic closeness impact of several strains in biofilm formation, only two-strain biofilms were analyzed. Urinary isolates (UI) 1, UI2 and UI6 present very low genetic variability between each other. UI4 and UI5 are also genetically similar. UI3 is the most genetically distant within these strains of *E. coli* (Figure 4a). The impact of the phylogeny was then assessed in terms of biofilm biomass (Figure 4b) and cell production (Figure 4c).

**Figure 4.** Heat maps showing the phylogenetic differences between the *E. coli* urinary isolates in percentage (**a**) and their biofilm-forming ability when combined with each other in pairs (**b**,**c**). Phylogenetic percentages were calculated using the different nucleotides in the 7 housekeeping genes between the six *E. coli* strains, based in the neighbor-joining method and pairwise distance. The biofilm-forming ability of the combined consortia was scored in terms of total produced biomass (**b**) and cultivability (**c**) according to the - Combinatorial biofilm index (- CbI scoring: cooperative (<0.875), neutral (0.875 < - CbI < 1.125) and antagonistic (>1.125)] for 0–48 h period of biofilm formation. This analysis on total biomass suggests that both high and low phylogenetic distance between the strains can be accompanied with an increase in the produced biomass, namely, when UI3, the most distant, was paired with all the remaining, as well as when related strains (UI1 with UI2; UI4 with UI5) were combined. All the pairs in terms of cultivability were scored as neutral.

The biomass and cultivability of the biofilms were scored as cooperative, neutral or antagonistic according to the - Combinatorial biofilm index (- CbI). Interestingly, in terms of total biomass, all the combinations of the UI3 (the most distant) with the remaining were scored as cooperative (0.63 < - CbI < 0.80). Still, the combinations of related strains such as UI1 with UI2 as well as UI4 with UI5 were also scored as cooperative. This analysis suggests that both high and low phylogenetic distance between the strains can be accompanied with an increase in the produced biomass. A neutral score is obtained when analyzing the - Cbi for all the consortia in terms of cultivability (Figure 4c). From this analysis, it appears that for two-strains biofilms, the phylogenetic relatedness is inversely proportional to the biomass production ability of the two-strain consortium. An open question remains on whether this behavior will still be observed when more than two-strain biofilms are analyzed in the future.

### *2.5. Spatial Organization of Biofilms Using PNA-FISH Combined with CLSM*

It has been demonstrated that microorganisms locate and organize themselves within biofilms according to the nature of their microbial interactions. In general, microorganisms organize in three main forms: a segregation form associated with competition, a co-aggregation/intermixing structure related with cooperation and a layering arrangement found in cooperative or competitive behavior [32].

In the present study, to better understand the behavior between the strains, a multiplex PNA-FISH combined with CLSM was performed for the analysis of the spatial distribution of *E. coli* multi-strain biofilms. First, the hybridization of the PNA probes was previously optimized in terms of temperature and formamide concentration. Optimal conditions for multiplex FISH were obtained at 50 ◦C and 30% (*v*/*v*) formamide. PNA-UI126 was specific for UI2 and 6. However, probe PNA-UI5 exhibited a certain level of fluorescence for the remaining strains when alone. Nevertheless, this can be resolved by multiplex FISH analysis, making it possible to identify a consortium composed by three strains (Figure S4). The CLSM 3D images of the morphology of the biofilms, z planes and their cross-section are presented in Figure 5, Figure S5 and Figure S6 for 6 h, 24 h and 48 h of biofilm growth, respectively. Overall, the tested consortia showed, in general, a coaggregation structure. Concerning biofilms incubated for 6 h, aggregates could be found, such as the one at the center of Figure 5e, composed mainly by strains 3 and 5. In fact, the strains seem to be well mixed and aggregated to each other. Similar findings were obtained when *E. coli* is present alone in consortium with other species [33,34].

**Figure 5.** Three-dimensional organization of 6 h-aged biofilm formed in AUM and in polystyrene coupons by a consortium of three *E. coli* strains (UI3—blue, UI5—green and UI6—red). (**a**) Examples of CLSM (confocal laser scanning microscopy) images obtained of the layers within the biofilm at different heights (**a** = 0 μm; **b** = 0.7 μm; **c** = 2.1 μm; **d** = 3.5 μm). (**e**) Cross-section of the biofilm.

Intraspecies coaggregation is stated as important for the development of biofilms, enabling metabolic interactions, cell–cell communication and genetic exchange [35]. Therefore, we theorize that the strains can co-exist in the same space and are just enough far apart phylogenetically to cooperate when intraspecies diversity is increased in the surroundings. To understand this observed intraspecies phenomenon and to explain why an increase in diversity alters the EPS/cell numbers on these biofilms, we suggest that an intraspecies facultative cooperation strategy occurs, allowing them to save energy or to share resources. On the other hand, the proximity of the strains within the consortia may also facilitate the exchange of genes that confer advantages or the accumulation of mutations, helping them to adapt to the environmental conditions when the number of strains in a consortium is increased.

### **3. Materials and Methods**

### *3.1. Bacterial Strains and Growth Conditions*

In this study, six different *E. coli* strains were isolated from urine samples of patients using Cystine lactose electrolyte deficient agar (CLED) medium and MacConkey agar medium (Liofilchem, Roseto degli Abruzzi, Italy). Agar plates were incubated for 24 h at 37 ◦C. The identity of the six *E. coli* isolates was confirmed by sequencing the 16S ribosomal RNA (16S rRNA) gene, performed by StabVida, Lda (Caparica, Lisbon), and later confirmed using the Basic Local Alignment Search Tool (BLAST; https://blast.ncbi.nlm.nih.gov). The *E. coli* strains were named urinary isolates (UI) followed by a number from 1 to 6 (e.g., UI1).

For each experiment, isolates were recovered from −80 ◦C glycerol stock cultures on Tryptic Soy Agar (TSA) (Merck, Darmstadt, Germany) and grown for 24 h at 37 ◦C.

### *3.2. Single- and Multi-Strain Biofilms Assays*

For the preparation of each inoculum, isolated colonies for each strain were inoculated into artificial urine medium (AUM) [36] and incubated overnight (16–18 h), at 37 ◦C and 150 rpm. Subsequently, cell concentration was assessed by optical density at 620 nm (O.D.620 nm) and each inoculum was diluted in AUM in order to obtain a cell concentration of 106 CFUs·mL<sup>−</sup>1.

In order to evaluate the biofilm-forming ability of each strain and in consortium, single- and multi-strain biofilms were grown as previously described [37]. Briefly, 200 μL of each inoculum prepared in AUM (106 CFUs·mL<sup>−</sup>1) were added to each well of a 96-well tissue culture plate (Orange Scientific, Braine-l'Alleud, Belgium). Regarding multi-strain biofilms, an adequate volume of each strain culture was mixed, keeping a final volume of 200 μL with an initial concentration of 106 CFUs·mL−<sup>1</sup> for each strain and added to each well. The plates were incubated at 37 ◦C, under static conditions, for 48 h. At predetermined time points (2, 6, 24 and 48 h), the biofilms were washed with 200 μL of 0.85 % (*w*/*v*) sterile saline solution to remove non-adherent and loosely attached cells. Then, biofilm formation was assessed by crystal violet (CV) staining (for total biomass quantification) and colony forming units (CFUs) counts (for cultivable cell counts). Quantification of extracellular polymeric matrix of single- and multi-strain biofilm was also performed for 48 h biofilms. These assays were performed with two or more independent experiments. Tested combinations are shown in Table S1 in Supplementary Information.

### *3.3. Biomass Quantification by Crystal Violet Staining*

The produced biomass by single- and multi-strain biofilms in 96-well tissue culture plate was assessed by the CV staining [38]. The washed biofilms were fixed with 250 μL of 99% ethanol (*v*/*v*) for 15 min. Then, ethanol was removed, and the plates air-dried. Subsequently, biofilms were stained with 250 μL of CV (Merck, Germany) for 5 min. Microplates were rinsed with water, air-dried and the CV was resolubilized by adding 200 μL of 33 % (*v*/*v*) glacial acetic acid (Merck, Germany) to each well. Plates were stirred for 2 min and the content was transferred to new 96-well plates for O.D. (570 nm) measure using a microtiter plate reader (Spectra Max M2, Molecular Devices, Sunnyvale, CA, USA).

### *3.4. Cultivability Assessment*

The number of cultivable biofilm cells was determined by CFUs counting as previously described by Azevedo, Almeida, Melo and Azevedo [37]. Briefly, after the washing step, 200 μL of 0.85 % (*w*/*v*) sterile saline solution were added into each well containing the biofilms. Biofilms were sonicated for 4 min (70 W, 35 kHz, Ultrasonic Bath T420, Elma, Singen, Germany) and 100 μL of the disrupted biofilms were serially diluted (1:10) in 0.85% (*w*/*v*) sterile saline solution and plated in triplicate in TSA plates (sonication conditions were previously optimized by Azevedo et al. [34]. The TSA plates were incubated at 37 ◦C for 14 h. The number of CFUs was expressed in logarithm per microtiter plate well's bottom and side area (log CFUs·cm<sup>−</sup>2).

### *3.5. EPS Matrix Quantification by Dry Weight*

EPS quantification was performed to a subset of biofilms (P1, P2, P3, P4, P5 and P6 (1 strain); P7 and P21 (2 strains); P22 and P41 (3 strains); P42 and P56 (4 strains); P57 and P62 (5 strains); P63 (6 strains)). For that, an extraction procedure was applied to separate the exopolymeric substances from the microbial cells [39,40]. The washed biofilms were sonicated in an ultrasonic bath for 4 min (70 W, 35 kHz, Ultrasonic Bath T420, Elma, Singen, Germany). The bacterial suspensions were transferred to universal tubes (50 mL) and sonicated (10 s, 25% amplitude). The bacterial suspensions were vortexed

for 2 min and centrifuged for 10 min at 3000 rpm (at 4 ◦C). Then, the supernatants were filtered through a membrane (0.2 μm), using a syringe, to pre-weighed tubes. The tubes were frozen at −80 ◦C during 48 h for liquid extraction by lyophilization. After lyophilization, the tubes were weighed, and the total mass of the EPS matrix was determined.

### *3.6. Multilocus Sequence Typing for Phylogenetic Analysis*

Multilocus Sequence Typing (MLST), as previously described by Liu et al. [41], was used to determine the diversity and phylogenetic relationships of the six *E. coli* strains. The sequencing of seven housekeeping genes, *adk* (Adenylate kinase), *fumC* (Fumarate hydratase), *gyrB* (DNA gyrase subunit B), *icd* (Isocitrate dehydrogenase), *mdh* (Malate dehydrogenase), *purA* (Adenylosuccinate synthetase) and *recA* (recA protein, repair and maintenance of DNA), was performed by StabVida, Lda (Caparica, Lisbon) following the protocols specified at the *E. coli* MLST website (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli). For each obtained DNA sequence, the consensus sequence was generated after all gaps and single nucleotide changes were checked in the chromatograms for forward and reverse sequences using the Geneious 9.0.4 software (Biomatters Limited, New Zealand). Afterwards, sequences of all seven genes were concatenated for each isolate and aligned using Geneious 9.0.4 software. The phylogenetic analysis was inferred by the neighbor-joining algorithm using Geneious 9.0.4 software for the calculation of the pairwise distances in percentage [42].

### *3.7. Impact of the Phylogenetic Closeness in Biomass Production*

In order to evaluate the phylogenetic impact in the biofilm development when the referred strains were combined, a scoring method was applied to the CV and CFUs areas (48 h growth) obtained for all the combinations. Therefore, the biofilm-forming ability and cultivability of the *E. coli* strains were scored according to the - Combinatorial biofilm index (- CbI) adapted from Baptista et al. [43] presented in Equations (1) and (2):

$$\text{Cb}\_{I(N)} = \frac{\text{Curve Area}\_{(N)}}{\text{Curve Area}\_{(\text{Combined})}}, N \in [1, 2, 3, 4, 5, 6] \tag{1}$$

where *CbI* is the *Cb* index, which is calculated for each strain to be compared, dividing the curve area (total biomass or cultivability) of each strain for the curve area when those strains are combined. Then, the sum of the *CbI* for each strain in consortia gives the - CbI as exemplified in Equation (2):

$$\sum\_{i=1}^{n} \, ^n \mathbf{C} b\_{l(i)} = \left( \mathbf{C} b\_{l(1)} + \, ^c \mathbf{C} b\_{l(2)} + \dots + \, ^c \mathbf{C} b\_{l(n)} \right) / n \tag{2}$$

The obtained values for - CbI are then scored according to the following criteria: <0.875—Cooperative; 0.875 to 1.125—Neutral; >1.125—Antagonistic. In addition, heat maps were constructed to better infer about the impact of the phylogenetic distance between the strains.

### *3.8. Optimization of the PNA-FISH Protocol*

Two PNA-FISH probes were designed to target the *E. coli* strains (one probe for UI1, UI2 and UI6 strains; and another one for UI5 strain). Mismatches in the 16S rRNA sequences between the six *E. coli* strains were analyzed using Clustal Omega multiple sequence alignment tool (https: //www.ebi.ac.uk/Tools/msa/clustalo/). Probes were selected according to their base pair (bp) length, mismatch localization, Guanine/Cytosine (GC) content, theoretical melting temperature point (Tm) and Gibbs free energy (ΔG < −13) [27] (Table 1). The probes were attached to Alexa Fluor 488 (PNA-UI5) and Alexa Fluor 594 (PNA-UI126) signaling molecules via a double 8-amino-3, 6-dioxaoctanoic acid (AEEA) linker (Panagene, Daejeon, South Korea, HPLC purified > 90%). For UI3 and UI4 strains, 4 -6-Diamidino-2-phenylindole (DAPI; Merck, Germany) was used as a counterstain.


**Table 1.** Sequences of the PNA-FISH probes, their target positions in the 16S rRNA and theoretical values including bp (base pair), % GC, ΔG, and Tm.

UI5 and UI2 strains were selected to optimize the hybridization conditions of the two PNA-FISH probes. A range of temperature and formamide concentrations were evaluated for a better microscopic signal using a LEICA DMLB2 epifluorescence microscope (Leica Microsystems Ltd.; Wetzlar, Germany) coupled with a Leica DFC300 FX camera (Leica Microsystems Ltd.; Wetzlar, Germany), with 100× oil immersion fluorescence objective. Images were acquired using Leica IM50 Image Manager, Image processing and Archiving software. The hybridization procedure in microscope slides was performed according to Almeida et al. [44]. Briefly, smears (30 μL) of each bacterial strain (OD620nm = 0.1 ≈ 10<sup>8</sup> cells·mL<sup>−</sup>1) in sterile distillate water were applied in microscope slides and immersed in 4% paraformaldehyde (30 μL) (Sigma-Aldrich, St. Louis, MO, USA) for 15 min. Then, 30 μL of 50% ethanol were added to the smears for 15 min and air-dried. Afterwards, the smears were covered with 20 μL of hybridization solution containing 10% dextran sulphate (Sigma-Aldrich, St. Louis, MO, USA), 10 mM NaCl (Sigma-Aldrich, St. Louis, MO, USA), 0.1% (*wt*/*vol*) sodium pyrophosphate (Sigma-Aldrich, St. Louis, MO, USA), 0.2% (*wt*/*vol*) polyvinylpyrrolidone (Sigma-Aldrich, St. Louis, MO, USA), 0.2% (*wt*/*vol*) Ficol (Sigma-Aldrich, St. Louis, MO, USA), 5 mM disodium EDTA (Sigma-Aldrich, St. Louis, MO, USA), 0.1% (*vol*/*vol*) Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA), 50 mM Tris-HCl (Sigma-Aldrich, St. Louis, MO, USA), 30% or 50% (*vol*/*vol*) of formamide (Acros Organics, Belgium), and 200 nM of PNA probe. Smears with hybridization solution without PNA probe were performed as negative controls. Samples were covered with coverslips and placed in an incubator for a range of temperatures (48 ◦C to 54 ◦C) for 90 min. Then, the microscope slides were immersed in a pre-warmed washing solution containing 5 mM Tris base (Sigma-Aldrich, St. Louis, MO, USA), 15 mM NaCl (Sigma-Aldrich, St. Louis, MO, USA) and 1% (*vol*/*vol*) Triton X (pH = 10, Sigma-Aldrich, St. Louis, MO, USA), without their coverslips for 30 min at the same temperature of the hybridization step. The smears were covered with a drop of non-fluorescent immersion oil (Merck, Germany) and watched at the microscope. The microscope slides were stored for a maximum of 24 h in the dark at 4 ◦C before microscopy.

### *3.9. Study of the Spatial Organization in E. coli Multi-Strain Biofilms Using PNA-FISH Combined with Confocal Laser Scanning Microscopy Analysis*

In order to assess the biofilm spatial organization and the strains distribution, the PNA-FISH and DAPI staining were performed directly in biofilms formed in polystyrene coupons as previously described [27,33]. Briefly, the biofilm formation was conducted as previously described. Then, 3 mL of the bacterial suspensions were added to each well of 12-well microtiter plates, with previously- placed sterilized polystyrene coupons (prepared according to Azevedo et al. [45]) at the bottom. The plates were then incubated for 6 h, 24 h and 48 h, under static conditions. After the incubation period, the coupons were carefully transferred and washed in another sterile 12-well microtiter plates with 3 mL of 0.85% (*w*/*v*) sterile saline solution and dried for 15 min at 60 ◦C. Afterwards, the biofilms in the coupons were placed carefully in a microscope slide, and the hybridization was performed as previously described for 90 min at 50 ◦C. When UI3 and UI4 were present, DAPI staining was performed. For this, a drop of DAPI solution (0.1 mg·mL<sup>−</sup>1; Merck, Germany) was added to the coupons for 10 min in the dark at room temperature.

After the FISH procedure, the analysis of multi-strain biofilms structure and the location of the bacterial strains was made using a confocal laser scanning microscopy (Olympus BX61, Model FluoView 1000) and the multichannel simulated fluorescence projection of images and vertical cross-sections through the biofilm were generated by using the FluoView 1000 Software package (Olympus). During the analysis, a 60× water-immersion objective (60×/1.2 W) was used.

A laser excitation line 405 nm and emission filters BA 430-470 (blue channel) were used for DAPI observation; PNA-UI5 probe coupled to Alexa Fluor 488 was observed using a laser excitation line 488 nm and emission filters BA 505-540 (green channel); and for observation of PNA-UI126 probe coupled to Alexa Fluor 594, a laser excitation line 559 nm and emission filters BA 575-675 (red channel) was used.

### *3.10. Statistical Analysis*

The results were compared using one-way analysis of variance (ANOVA) by applying the Tukey multiple-comparisons post-hoc test, using the Statistical Package for the Social Sciences (SPSS) Statistics 25 (IBM, Armonk, NY, USA). All tests were performed with a confidence level of 95%. The results were also compared by model-based clustering, using R software. The standard methodology selects the number of clusters according to the Bayesian information criterion (BIC) [46,47].

### **4. Conclusions**

In conclusion, these results may represent another way to fight biofilm-related infections. We show that even in fully functional microorganisms, cells are induced to produce less EPS matrix when other strains are present. Nonetheless, most clinical biofilms are thought to be caused by a single strain. Hence, introducing multiple avirulent strains of the infecting microorganism at the site of infection, we might be inducing an EPS matrix-deficient biofilm which is, in theory, more susceptible to antibiotics/antimicrobial treatment. Moreover, as most of the antibiotics are most active in dividing cells [48], promoting cellular growth while increasing diversity, we might also increase susceptibility to antibiotics. For such strategy to work broadly, this approach must be applicable to intraspecies biofilms of other species and to pre-formed biofilms. Future lines of work will also include testing of these multi-strain biofilms in the presence of currently used antibiotics.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/11/818/s1, Figure S1: Number of cultivable cells in single-strain biofilms and in multi-strain biofilms combining two, three, four, five and six *E. coli* strains, during 48 h. Standard deviations of three independent replicates are displayed. Figure S2. O.D. values for total biomass quantification (CV method) for single-strain biofilms and for multistrain biofilms combining two, three, four, five and six *E. coli* strains, during 48 h. Standard deviations of three independent replicates are displayed; Figure S3. Mean EPS matrix concentration ± standard deviation of biofilms formed by consortia of 1 up to 6 different *E. coli* strains after 48 h of incubation at 37 ◦C in AUM. The produced EPS matrix decreases linearly with the addition of strains in consortia, adjusted with an R-squared of 0.94. Figure S4. Multiplex FISH using both PNA-UI5 and PNA-UI126 probes using UI5 and UI2, respectively, at 50 ◦C hybridization temperature and 30% formamide concentration. a–Green filter; b–Red filter; c–Filters overlapping. A magnification of 1500× was used. Figure S5. Three-dimensional organization of 24 h aged biofilm formed in AUM and in polystyrene coupons by a consortium of three *E. coli* strains (UI3-blue, UI5-green and UI6-red). (a) Examples of CLSM images obtained of the layers within the biofilm at different heights (a = 0 μm; b = 1 μm; c = 2 μm; d = 3 μm). (e) Cross section of the biofilm. Figure S6. Tri-dimensional organization of 48 h aged biofilm formed in AUM and in polystyrene coupons by a consortium of three *E. coli* strains (UI3-blue, UI5-green and UI6-red). (a) Examples of CLSM images obtained of the layers within the biofilm at different heights (a = 0 μm; b = 2 μm; c = 4 μm; d = 6 μm). (e) Cross section of the biofilm. Table S1: Number of possible combinations (P) of the six different *E. coli* strains used for biofilms formation assays.

**Author Contributions:** Conceptualization, A.S.A., C.A. and N.F.A.; Methodology, G.P.G. and J.B.; Formal Analysis, A.S.A., G.P.G., J.B. and J.P.; Investigation, A.S.A., G.P.G. and J.B.; Writing–Original Draft Preparation, A.S.A., G.P.G. and J.B.; Writing, A.S.A., C.A., F.J.M. and N.F.A.; Supervision, F.J.M. and N.F.A.; Project Administration, A.S.A. and N.F.A.; Funding Acquisition, A.S.A. and N.F.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was financially supported by Base Funding—UIDB/00511/2020 of the Laboratory for Process Engineering, Environment, Biotechnology and Energy—LEPABE—funded by national funds through the FCT/MCTES (PIDDAC); Project POCI-01-0145-FEDER-030431 (CLASInVivo) and project

POCI-01-0145-FEDER-029841 (POLY-PREVENTT), funded by FEDER funds through COMPETE2020—Programa Operacional Competitividade e Internacionalização (POCI) and by national funds (PIDDAC) through FCT/MCTES; Strategic funding of UIDB/04469/2020 of the Centre of Biological Engineering–CEB–funded by national funds through the FCT; Project BeMundus Brazil Europe/Erasmus Mundus scholarship granted by BM13DF0014.

**Conflicts of Interest:** The authors report 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/).

### *Article* **Pitfalls Associated with Discriminating Mixed-Species Biofilms by Flow Cytometry**

### **Tânia Grainha, Andreia P. Magalhães , Luís D. R. Melo \* and Maria O. Pereira \***

Centre of Biological Engineering, LIBRO—Laboratório de Investigação em Biofilmes Rosário Oliveira, University of Minho, Campus de Gualtar, 4710-057 Braga, Portugal; taniagrainha@ceb.uminho.pt (T.G.); amagalhaes@ceb.uminho.pt (A.P.M.)

**\*** Correspondence: lmelo@deb.uminho.pt (L.D.R.M.); mopereira@deb.uminho.pt (M.O.P.); Tel.: +351-253-601-989 (L.D.R.M.); +351-253-604-402 (M.O.P.)

Received: 10 September 2020; Accepted: 21 October 2020; Published: 27 October 2020

**Abstract:** Since biofilms are ubiquitous in different settings and act as sources of disease for humans, reliable methods to characterize and quantify these microbial communities are required. Numerous techniques have been employed, but most of them are unidirectional, labor intensive and time consuming. Although flow cytometry (FCM) can be a reliable choice to quickly provide a multiparametric analysis, there are still few applications on biofilms, and even less on the study of inter-kingdom communities. This work aimed to give insights into the application of FCM in order to more comprehensively analyze mixed-species biofilms, formed by different *Pseudomonas aeruginosa* and *Candida albicans* strains, before and after exposure to antimicrobials. For comparison purposes, biofilm culturability was also assessed determining colony-forming units. The results showed that some aspects, namely the microbial strain used, the morphological state of the cells and the biofilm matrix, make the accurate analysis of FCM data difficult. These aspects were even more challenging when double-species biofilms were being inspected, as they could engender data misinterpretations. The outcomes draw our attention towards the need to always take into consideration the characteristics of the biofilm samples to be analyzed through FCM, and undoubtedly link to the need for optimization of the processes tailored for each particular case study.

**Keywords:** *Pseudomonas aeruginosa*; *Candida albicans*; mixed-species biofilm analysis; flow cytometry

### **1. Introduction**

The ability of microorganisms to form a biofilm is an important feature in clinical, industrial and environmental settings [1]. Biofilms are well-structured microbial communities adhered to biotic or abiotic surfaces enclosed within a self-produced extracellular polymeric matrix. The matrix is the major structural component of the biofilm [2] and is denominated as an extracellular polymeric substance (EPS). The EPS is composed of polysaccharides, proteins, nucleic acids, lipids and other biopolymers such as humic substances [3]. Most natural biofilms are polymicrobial and all members of the community contribute with their own EPS components, resulting in a more complex matrix [4].

The biofilms associated with numerous infectious diseases (e.g., infections of oral cavity, otitis, cystic fibrosis) are described as being of a polymicrobial nature with bacteria coexisting with pathogenic yeasts or filamentous fungi [5]. *Pseudomonas aeruginosa* is a ubiquitous bacterium and an opportunistic pathogen frequently isolated from healthy humans as part of the human microbiota and can coexist in mixed infections with the polymorphic fungus *Candida albicans* [6,7]. *C. albicans* is a commensal yeast able to initiate invasive growth and develop health problems in compromised individuals [8,9]. Notably, *C. albicans* is one of the few fungal species causing disease in humans [10]. *P. aeruginosa* and *C. albicans* represent an example of co-infection and are commonly related with chronic and healthcare-associated

infections [11–13]. From a clinical point of view, it is crucial to entirely characterize the biofilm populations (single- and mixed-species) and understand the mechanisms underlying the changes that occur during co-infection as result of the established interactions. The monitoring of polymicrobial communities is routinely performed by culture-dependent approaches that require appropriate selective media and optimal growth conditions and is hindered by the presence of cells in a viable but poorly culturable state or underestimated by the presence of cellular aggregates [14,15]. Under these circumstances, standard microbiology methods are often unsuitable to diagnose polymicrobial biofilm infections. Molecular methodologies, such as quantitative real-time PCR (qPCR) [16,17] or fluorescence in situ hybridization using peptide nucleic acids (PNA FISH) [18–20], have helped in the characterization of members of microbial communities, affording a specific and sensitive quantification of specific species in biofilm samples that are unable to be detected by culturable methods. Nevertheless, these approaches cannot distinguish cellular subpopulations. Thus, flow cytometry (FCM) could present an accurate alternative to the analysis of biofilm cells, since it enables detailed investigation of cellular subpopulations due to its ability to perform multiparametric single-cell analysis [21]. The analysis of biofilm communities by this technique is particularly useful, since it allows a quick achievement of cell counts and provides an overview about the type of cells present in the samples, namely their size and complexity. In addition, this method allows for the discrimination between live and dead cells, also providing information about damaged cells [22]. Based on this, the goal of this study was to explore FCM analysis to characterize and discriminate mixed-species biofilm samples before and after being challenged by antimicrobial treatments.

Accordingly, the aspects that can influence FCM analysis as well as tips to circumvent them are presented through this work.

### **2. Results**

### *2.1. Planktonic versus Biofilm Cells*

Before biofilm analysis by FCM, an optimization of the process is desirable to adjust all sample analysis parameters. Firstly, as bacteria might be in the FCM detection limit, owing to its small size, a preliminary optimization was performed in order to guarantee the accurate detection of *P. aeruginosa* and thus the reliability of the results. Due to the complexity and heterogeneity of biofilms, optimizations using planktonic cells were performed. Since *P. aeruginosa* and *C. albicans* have distinct sizes and the flow cytometer allows for separation by size and complexity, it was assumed that their distinction on mixed communities should be well achieved using the same dyes. As can be observed in the dot plot (Figure 1), it was possible to define gates for each single-species planktonic culture tested.

Before examining the mixed-species biofilms, single-species biofilms of both species were also analyzed. The results obtained for these biological samples showed that cells from *C. albicans* biofilm populations are very different on cell size and complexity from the planktonic cells, leading to gating adjustments (Figure 1). In addition to the typical fungal cell gate previously defined for the planktonic cultures, several fluorescently labeled events up to the demarcated bacterial region were detected (Figure 2). To check whether this feature was particular for the *C. albicans* SC5314 strain, two more strains of each studied species (*P. aeruginosa* and *C. albicans*) were analyzed both on planktonic and biofilm states. Although this effect was not observed for *P. aeruginosa* PAO1, two more *P. aeruginosa* strains were also analyzed to strengthen the absence of that behavior. While this effect was not observed for any of the *P. aeruginosa* strains (Supplementary Figure S1), it was consistent on all *C. albicans* strains tested (Supplementary Figure S2).

**Figure 1.** Representative dot plots obtained for planktonic cells and single biofilms of *P. aeruginosa* PAO1 and *C. albicans* SC5314 by flow cytometry (FCM).

**Figure 2.** Representative dot plots obtained for *C. albicans* SC5314 biofilm by FCM. In 'all data points', the dot plots SS (Side Scatter) × FS (Forward Scatter) are represented, as acquired in the logarithm. Dot plots of *P. aeruginosa* and *C. albicans* represent the areas delineated to represent bacteria and fungi, respectively.

### *2.2. E*ff*ect of Hyphae and Biofilm Matrix on Flow Cytometry Analysis*

In an attempt to understand the factors generating the differences observed in the *C. albicans* biofilm populations in comparison with planktonic cultures, two possibilities were inspected, the impact of hyphal growth and the effect of the biofilm matrix. Since *C. albicans* biofilm cells present different phases of growth, it is expected that cells with distinct hyphal lengths are also present. Additionally, the biofilms were grown in Roswell Park Memorial Institute (RPMI) 1640, a medium that favors hyphal growth; therefore, more elongated hyphae could appear in these samples compared with planktonic cultures [23]. Based on this knowledge, the possibility of hyphal growth introducing some heterogeneity in the population, with a consequent interference in the results, was raised. To evaluate this hypothesis, planktonic cultures of *C. albicans* SC5314 grown in Sabouraud Dextrose Broth (SDB) and in RPMI supplemented with serum 2% (*v*/*v*), to induce hyphal cell growth, were analyzed. It was noticed that the fungal population which had hyphae induction became more heterogenous (Supplementary Figure S3), thus allowing us to verify that the presence of hyphae could account for the observed

differences. However, a closer inspection of the results for *C. albicans* 547096, a strain that does not have the ability to form hyphae in both planktonic and biofilm conditions, allowed us to notice a change between both modes of growth (Supplementary Figure S2). Despite it being clear that hyphal growth has some influence on the results, this did not entirely explain what happened in biofilm samples.

To deeply understand what influenced the results obtained in the biofilm samples, the possible interference of the biofilm matrix was also evaluated. The EPS matrix is composed of large amounts of eDNA [3,24] that can be bounded by the used fluorochromes, but, due to its small size, is only detected by the cytometer when it is attached to other matrix components. Therefore, efforts were done to extract the biofilm matrix, then the biofilm cells with and without the matrix were analyzed on the cytometer. The FCM of the biofilm matrix samples revealed a similar pattern to that obtained previously for *C. albicans* biofilms (Figure 3). Although this alteration between planktonic and biofilm samples was not previously clearly detected for *P. aeruginosa* strains, the analyses of the biofilm matrix samples allowed us to notice that the EPS matrix also play a role in the bacterial biofilm cell counts (Figure 3).

**Figure 3.** Representative dot plots obtained for biofilm matrix of *C. albicans* 547096 and *P. aeruginosa* PAO1 by FCM. In 'all data points', the dot plots SS (Side Scatter) × FS (Forward Scatter) are represented, as acquired in the logarithm. The dot plot of *P. aeruginosa* represents the areas delineated to represent bacteria.

The results indicate that both hyphal growth and the biofilm matrix account for altering the typical gate observed in *C. albicans* planktonic cultures. However, the presence of the EPS matrix has been shown to have greater impact on the FCM analysis. Although the *P. aeruginosa* gate does not differ between planktonic and biofilm, it was also evident that the matrix influences FCM analysis.

Overall, the data clearly highlighted that biofilm matrix extraction, before FCM analysis, is a requirement for this kind of methodology. Otherwise, it will not be possible to accurately analyze biofilm populations, due to the risk of populations not being properly differentiated, and consequently causing cell counts to be severely influenced by positively marked matrix events.

### *2.3. Influence of Sonication on Biofilm Cell Viability*

As this study has mixed-species biofilms as the main focus, a sonication procedure was used based on a previous optimized protocol. The sonication time was evaluated to ensure that this procedure did not result in cell lysis (Supplementary information Figure S4). The results indicated that the matrix extraction with 30 s of sonication at 30% amplitude was not harmful to any of the strains tested (*p* < 0.05). However, these sonication conditions were not effective in removing the entire matrix of *C. albicans* SC5314 and 324LA/94 strains (Supplementary Figure S5). As can be observed in Figure 4, although some matrix has been removed from *C. albicans* 324LA/94 samples, the cellular suspensions still present some matrix traces, which are being gated where *P. aeruginosa* is usually gated, making the study of mixed-species biofilms containing this *C. albicans* strain unfeasible.

**Figure 4.** Representative dot plots obtained by FCM for *C. albicans* 324LA/94 biofilms before (**A**) and after (**B**) extraction of biofilm matrix.

### *2.4. Analysis of Mixed-Species Biofilms*

To focus and frame the study, the number of strains used in the mixed biofilm formation were narrowed to one strain for each organism. In the case of *P. aeruginosa*, as no differences were observed between the strains tested in the FCM data analysis, the PAO1 strain was selected, because it is one of the most commonly used strains for biofilm research. Regarding *C. albicans*, to minimize the effects of the biofilm matrix and to eliminate hyphae influence, 547096 was the strain selected. The sonication parameters were tested in these mixed-species biofilms, and the results show that the matrix was effectively extracted (Figure 5).

**Figure 5.** Representative dot plots obtained for *P. aeruginosa* PAO1 and *C. albicans* 547096 mixed-species biofilms by FCM. In 'all data points', the dot plots SS (Side Scatter) × FS (Forward Scatter) are represented, as acquired in the logarithm. Dot plots of *P. aeruginosa* and *C. albicans* represent the areas delineated to represent bacteria and fungi, respectively.

Indeed, the results showed a clear and distinct separation between bacterial and fungi populations, making the evaluation of this consortium by FCM possible. Since the *C. albicans* matrix, which was gated where *P. aeruginosa* is usually gated, does not appear because it was removed, bacterial cells will be more feasible counted.

### *2.5. Antimicrobial E*ff*ect on Mixed-Species Biofilms*

Pre-established polymicrobial biofilms were treated with ciprofloxacin or linalool and their antimicrobial effect was evaluated by FCM (Figure 6) and colony-forming units (CFU) were counted (Table 1).

**Table 1.** FCM counting and colony-forming units (CFU) enumeration of pre-established mixed biofilms treated with two different concentrations of ciprofloxacin and linalool. Both 24 and 48 h-old untreated biofilms are included for comparison purposes. Log10 values represent means ± standard deviations (sd).


\* Significantly different compared with 24 h-old biofilm control (*p* < 0.05). # Significantly different compared with 48 h-old untreated biofilm (*p* < 0.05).

**Figure 6.** Representative all data points (**A**) and specific dot plots (**B**) obtained for *P. aeruginosa* PAO1 and *C. albicans* 547096 mixed-species biofilms by FCM. Dot plots of *P. aeruginosa* and *C. albicans* represent the areas delineated to represent bacteria and fungi, respectively. The arrows indicate alterations in the core of the population compared to the controls.

To understand the effect of the antimicrobials tested, it was considered important to verify the behavior of the total population in terms of size (FS) and complexity (SS). Figure 6 displays the graphs of all data points obtained for 24 and 48 h-old biofilms, and for the populations after antimicrobial treatments. The analysis of the dot plots indicated that there are no evident changes when comparing the 24 h-old biofilms of *P. aeruginosa* and *C. albicans* with the corresponding 48 h. The core of the fungal population presented a slight increase in the SYTO BC mean fluorescence intensity (MFI) from 2189 to 4208. Regarding the effect of the antimicrobials, the data revealed that, when using both ciprofloxacin concentrations, the core of the *P. aeruginosa* population is altered compared with the controls. In terms of viability, ciprofloxacin treatment does not appear to affect *P. aeruginosa* biofilm cells. In the case of *C. albicans*, no notable differences in terms of size and complexity were observed. Nevertheless, the yeast population seemed to be slightly destabilized, with the observation of two sub-populations, as well as a decrease in the SYTO BC uptake and an increase in the propidium iodide (PI) uptake compared with the 24 h-old biofilm control. These differences were more pronounced with the lowest concentration of ciprofloxacin ranging from 2182 to 1072 for SYTO BC MFI, and from 580 to 692 for PI MFI.

Using linalool, *P. aeruginosa* and *C. albicans* populations appeared altered, being more dispersed, suggesting a more heterogeneous population. With both linalool concentrations, *C. albicans* cells presented a reduced size and complexity (Figure 6). A population displacement along the PI axis was also observed, meaning that part of the cells is double stained. Furthermore, we observed a decrease in the SYTO BC uptake compared with the control, whereas SYTO BC MFI ranged from 2189 to 1604 and 1715 with the lowest and the highest concentration, respectively. Concerning *P. aeruginosa* populations, there was no effect observed in terms of viability and only a slight decrease in SYTO BC MFI was detected (from 117 in 24 h-old biofilm to 84 after biofilm treatment).

The cellular quantifications of the untreated and treated mixed biofilms are gathered in Table 1. Overall, FCM counts were somewhat lower than those detected by CFU, except when mixed biofilms were challenged with linalool, wherein *C. albicans* CFU counts were very low or even null. However, it must be taken into consideration that the standard deviation values associated with the CFU counts are higher than those observed for FCM, which might suggest more data variability between experiments.

Regarding the FCM results (Table 1) for *P. aeruginosa*, no significant differences were observed between ciprofloxacin-treated biofilms and the 24 h-old biofilm control. Furthermore, around 2 and 2.5-log reductions were achieved with the lowest and highest concentrations of ciprofloxacin, respectively, compared to the 48 h-old untreated biofilm. Taking into consideration CFU counts, the application of ciprofloxacin gave rise to some reductions in *P. aeruginosa* cells, notably when the higher concentration was used (about 2 and 3.5-log reductions compared with 24 and 48 h-old untreated biofilm, respectively). Ciprofloxacin treatment had no relevant effect on *C. albicans* as cell numbers remain unchanged for both concentrations, whatever the method used for biofilm cell counting.

The use of linalool had no relevant effect on *P. aeruginosa* viability and culturability compared with the 24 h-old biofilms. However, when compared with 48 h-old biofilms, although not statistically significant, about a 1-log reduction was observed when using either FCM or CFU. Concerning the fungus population in the mixed consortia, the use of FCM or CFU to count biofilm cells challenged by linalool gave rise to very distinct scenarios (*p* < 0.05). Indeed, through FCM, no significant alterations in the number of cells were noticeable in comparison with the respective controls. However, CFU counts revealed a clear negative effect of linalool in *C. albicans* culturability, achieving reductions of more than 4 or 5-log compared to the 24 or 48 h-old untreated biofilms, and even total eradication with the highest linalool concentration.

### **3. Discussion**

The polymicrobial nature of most infections [5] leads to the growing need for the study of these complex communities. Currently, there are several techniques that can be employed to perform biofilm analysis that are often dependent on the investigation purpose. Although FCM has been essentially applied for studying planktonic cultures [25], there are some studies using biofilms [26–28]. Pan et al. (2014) compared three methods (FCM, CFU and a spectrophotometry method of optical density measurement) for the quantification of bacterial cells after exposure to nanoparticles and

found that FCM measurement was the quickest and most accurate method for bacterial detection [29]. This methodology has also been successfully applied in the study of single-species biofilms and to investigate bacterial physiological responses [26,27]. FCM was also successfully applied in the study of cell viability in planktonic mixed cultures, using Gram-specific fluorescent staining [30].

Since the study of polymicrobial biofilm communities using FCM is less common, the rationale behind this study was to exploit and fruitfully apply this technique in the characterization and quantification of mixed-species biofilms while paying attention to the eventual hitches that this technique could engender and trying and to find solutions to surpass them.

Biofilms are complex microbial communities in which cells are embedded in an EPS matrix that cements cells together and provides heterogeneous microenvironments, leading to cells adopting different physiological states. Thus, biofilm cells differ phenotypically from their planktonic counterparts [31]. The results obtained in this study have shown that the analysis of biofilm cell viability using FCM is not straightforward, essentially due to the presence of cells with different morphologies and the biofilm matrix. *C. albicans* biofilms are typically formed by a mixture of vegetative cells, pseudohyphae and hyphae [32], with this mixture of morphologies having been detected by FCM. Due to the size and complexity of fungal cells, it was noticed that the counts made by FCM were influenced by the presence of hyphae, since they can be counted as more than a single event (Supplementary Figure S3).

The EPS matrix represents a significant part of the biofilms, playing an important role in their development and cohesion [33–35]. The matrix is suggested as the biofilms' house [36] because it protects biofilm cells from physical, chemical and biological adversities. The biofilm matrix contains several constituents, including eDNA [3,24], which, once attached to other matrix components, may be counted as positive events when passed in the flow cytometer, as highlighted by Figure 3. When the biofilm matrix was analyzed, several fluorescently labeled events, from the fungal cell gate previously defined up to the demarcated bacterial region, were detected. This squeezing of the events on the FS axis might be due to the heterogeneity of biofilm matrix components. These aspects may explain the impossibility of distinguishing the bacterial and fungal populations present in mixed cultures. Thus, matrix extraction is a crucial step before biofilm analysis by FCM. There are different methods that have been described for the extraction of EPS from single- and mixed-species biofilms, including centrifugation, filtration, heating, blending, sonication and treatments with agents or resins [37]. The EPS isolation method selected should be adapted accordingly to the type of biofilm under investigation. In the case of mixed-species biofilms, the selection and optimization of the extraction procedure was even more difficult. Moreover, it has to be taken into account that if the process becomes too complex or time consuming, the advantage of using FCM is lost. Sample sonication was the chosen methodology, but the matrix was not effectively extracted for the three *C. albicans* strains tested (Supplementary Figure S5) meaning that the success of the matrix extraction was strain dependent. These findings require longer sonication times and/or amplitudes to be tested in order to increase the amount of matrix extracted. Although it is known that sonication is an extraction method whose parameter optimization is microorganism dependent, in the case of mixed biofilms, there must be a commitment to reconcile the best possible parameters for all species present in the consortia. Thus, a reasonable time of sonication for both microbes present in the mixed cultures must be chosen, once the desirable target is achieved wherein the process does not affect the cellular viability of any of the tested species.

Based on previous studies, it can be assumed that, when using the same concentrations of antimicrobials agents, a greater antimicrobial effect occurs on planktonic cells, when comparing to the effect observed on biofilm cells, namely for ciprofloxacin [38] and linalool [39] towards *P. aeruginosa* and *C. albicans*, respectively. It is expected that, following an antimicrobial treatment, the composition and state of microbial populations might be altered [40]. Although, in some cases, there were no evident changes in the number of cells, it is important to emphasize some variations in SYTO BC and/or PI MFI, as well as variations in the core of the population (Figure 6). Ciprofloxacin is a broad-spectrum antibiotic of the fluoroquinolone class that acts on DNA gyrase (topoisomerase II) and topoisomerase IV, resulting in the inhibition of DNA replication, recombination and transcription, and thus causing bacterial death [41]. Ciprofloxacin affected the size and complexity of *P. aeruginosa* (Figure 6) and, regarding the results obtained by FCM, both concentrations of ciprofloxacin showed a bacteriostatic effect (Table 1) as the number of cells was lower after treatment compared with the 48 h-old untreated biofilm. Concerning CFU results, in addition to its bacteriostatic activity, a bactericidal activity was also observed to be more pronounced with the highest concentration of the antibiotic. These discrepancies between different methods are in accordance with the results observed by other authors [42]. Though ciprofloxacin is a bactericidal antibiotic according to classical testing, which uses planktonic cells, it is well known that cells from biofilm are more tolerant to the actions of antibiotics and harder to eradicate. Biofilm cells usually require higher doses to be eradicated by an antimicrobial [43]. The bacteriostatic effect observed here is due to the fact that cells were grown embedded in a biofilm. Both bacteriostatic and bactericidal effects of ciprofloxacin were previously reported in *Escherichia coli* [42]. Taking the 24 h-old biofilms as the reference, the application of ciprofloxacin did not alter the number of *C. albicans* cells; however, when the comparison is made with the 48 h-old untreated biofilm, a decrease in the number of cells was observed. These results indicate that ciprofloxacin might also have a fungistatic effect. Moreover, it was observed that the *C. albicans* population was divided into two sub-populations, shifted to the left with a reduced uptake of SYTO BC (Figure 6), meaning that a diminished metabolism of viable cells after ciprofloxacin treatment may have occurred. Indeed, a decrease in staining intensity was previously associated as an indicative characteristic of decreased metabolic state [26,44]. Although fluoroquinolones have no intrinsic antifungal growth-inhibitory activity, topoisomerase I and II are found in pathogenic fungi [45–47], which suggests that ciprofloxacin might be a strong candidate to interact with antifungal agents [48–50]. Linalool is a terpene alcohol commonly found as a component of the essential oils of aromatic plants. This compound has been reported as having antifungal activity, specifically against *C. albicans* [51,52]. A more recent study showed that linalool suppressed the expression of several virulence-related genes [53]. The effect of linalool on *C. albicans* cells showed marked differences between FCM and CFU counts. Although *C. albicans* did not suffer reductions in terms of FCM counts, the number of CFU was diminished after linalool application, meaning that their growth ability was affected. Even though no reduction in the FCM counts was observed, a clear alteration in population diversity, as well as a shift in the fungus population along the PI axis, was observed (Figure 6), meaning that cells are double stained with SYTO BC and PI. This increase in the number of cells being double stained suggests that linalool damages the cell membrane, allowing PI to enter into the cells. Thus, it can be assumed that part of the cell is in an intermediate physiological state between life and cell death, which may explain the reduction in the ability of fungi to grow on solid media. A double-stained population is considered injured, as discussed by Léonard et al. [54]. Therefore, the discrepancy between CFU and FCM counts may be explained by the emergence of viable but non-culturable cells; nonetheless, an additional non-culture-based method would be valuable to unambiguously assure this hypothesis. Indeed, the exposure of cell populations to antimicrobial action can lead to the appearance of different cell subpopulations, particularly an increase in the number of viable but non-culturable cells [55,56]. The differences here observed reflect the problem associated with CFU counting, since it does not allow for the detection of viable but non-culturable cells. Differences between culturable and FCM counts were previously reported by other authors [26,44]. The abovementioned linalool membrane effect was also observed in other studies, where it was reported that this antimicrobial agent acts by causing the disruption of the membrane integrity and interrupting the cell cycle [51,57]. Concerning *P. aeruginosa*, the application of both linalool concentrations did not alter the number of cells when compared to the 24 h-old biofilm control, but caused a decrease when the 48 h-old untreated biofilms were used for comparison, meaning that that linalool might have bacteriostatic activity. Similar bacteriostatic effects of linalool were previously reported for other bacteria, namely

*Enterobacter cloacae*, *E. coli*, *Proteus mirabilis*, *Salmonella enteritidis*, *Salmonella typhimurium, Staphylococcus epidermidis* and *Listeria monocytogenes* [58].

In addition to the discrepancies between FCM and CFU counts already mentioned, others were also observed (Table 1). In most cases, the number of culturable cells was higher than the FCM counts. This fact can be explained by the presence of injured cells that can reverse their state on fresh culture media and thus recover their growth ability. Moreover, in general, CFU results have higher variability between experiments, which leads to less accurate results.

Overall, this study showed that FCM can be a reliable methodology for the study of mixed-species biofilms, allowing for the discrimination of the stakeholders. However, for each consortium, previous optimization procedures must be followed, namely biofilm matrix extraction methodologies. Moreover, it was demonstrated that, when FCM is applied to scrutinize the mode of action of antimicrobial treatments, new insights can be provided due to the multiparametric analysis this technique allows.

A great step forward in the present research would be the clinical implementation of this methodology. Once the protocol is well developed and it is routinely applied, it is expected that it would be applied to real biofilm communities in clinical settings using blood samples or other samples where microorganisms are founded. Although FCM has been employed much more in the field of hematology, it has already been studied in different contexts. Microbial detection by FCM has already been proven to be possible using blood samples, as demonstrated in a previous work [59]. Clinical microbiology has undergone important changes during the last few years. Indeed, in recent years, microbiological techniques used in laboratories have been increasingly complemented by cutting-edge technologies such as FCM. The use of these practices presents several advantages to others, such as culture-dependent methods or microscopic approaches, since it provides multiparametric single-cell analysis very rapidly.

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

### *4.1. Microorganisms and Culture Conditions*

Three reference strains of *P. aeruginosa*, two non-mucoid strains (PAO1 and UCBPP-PA14 (PA14)) and a mucoid strain (ATCC 39324), were used throughout this work. In addition, two clinical isolates of *C. albicans* (324LA/94, an oral isolate obtained from the culture collection of Cardiff Dental School (Cardiff, UK) and 547096, a urinary isolate obtained from the culture collection of the Biofilm Group of the Centre of Biological Engineering (Braga, Portugal)), and a reference strain, SC5314, were tested.

Prior to each assay, *P. aeruginosa* and *C. albicans* strains were subcultured from the frozen stock preparations onto Tryptic Soy Agar (TSA) and Sabouraud Dextrose Agar (SDA) plates, respectively. TSA and SDA were prepared from Tryptic Soy Broth (TSB; Liofilchem S.r.l., Roseto, Italy) or SDB (Liofilchem) supplemented with 1.2% (*w*/*v*) agar (Liofilchem). The plates were then incubated aerobically at 37 ◦C for 18–24 h.

Pure liquid cultures (pre-inocula) of *P. aeruginosa* were grown overnight in TSB, whereas *C. albicans* was maintained in SDB. For biofilm assays, 0.22 μm of filter-sterilized RPMI 1640 medium (Gibco® by Life TechnologiesTM, Grand Island, NY, USA) at pH 7.0 was used.

### *4.2. Biofilm Formation*

Biofilm assays were performed as previously described [60], with some modifications. Briefly, the initial cell suspension (pre-inocula) was centrifuged (3000× *g*, 4 ◦C, 10 min) and the pellet resuspended in RPMI 1640 to achieve a concentration of ~1 × 10<sup>7</sup> CFU per mL. Bacterial concentration was estimated using an ELISA microtiter plate reader at an optical density of 640 nm (OD640 nm) (Sunrise-Basic Tecan, Männedorf, Switzerland), while yeast cells were enumerated by microscopy using a Neubauer counting chamber. For mixed-species cultures, a combination of 50% of the suspended inoculum of each species was used. Cellular suspensions were further transferred to 24-well plates

(Orange Scientific, Braine-l'Alleud, Belgium). Plates were then incubated aerobically for 24 h on a horizontal shaker at 120 rpm and 37 ◦C.

### *4.3. Hyphal Induction*

In order to promote hyphal growth of *C. albicans* cells, planktonic cultures of *C. albicans* S5314 were grown overnight in RPMI supplemented with 2% (*v*/*v*) fetal bovine serum (Biochrom AG, Berlin, Germany) at 120 rpm and 37 ◦C. Cells were then analyzed by FCM.

### *4.4. Biofilm Quantification*

### 4.4.1. Determination of Culturable Cells

After biofilm formation, wells were washed twice with sterile water after discarding the planktonic fraction. Afterwards, 500 μL of phosphate-buffered saline (PBS; 10 mM potassium phosphate, 150 mM NaCl; pH 7.0) was added to each well and the biofilms were scraped. In order to ensure the reproducibility of the scraping method, the conditions were strictly followed in all experiments by using a pipette tip and scraping each well about 1 min. To remove any aggregates, biofilm suspensions were vigorously vortexed (V1-Plus Biosan, Riga, Latvia) for 30 s. The resulting biofilm suspensions were then serially diluted in sterile water and plated onto agar plates (TSA for *P. aeruginosa* and SDA for *C. albicans*) for single-species biofilms. For mixed-species biofilms, *Pseudomonas* Isolation Agar (PIA; Sigma-Aldrich, St. Louis, MO, USA) and SDA supplemented with 30 mg/L gentamycin (Sigma-Aldrich, St. Louis, MO, USA) (to suppress the growth of *P. aeruginosa*) were used for the specific isolation of *P. aeruginosa* and *C. albicans*, respectively. Agar plates were incubated aerobically at 37 ◦C for 24–48 h for culturable cell counting. Values of culturable sessile cells were expressed as log10 CFU per mL and represent the average of the triplicates for each strain. At least two independent experiments were carried out in duplicate.

### 4.4.2. Extraction of Biofilm Matrix

For the extraction of the biofilm matrix, a previously described protocol was followed [61]. In brief, after washing and scraping the biofilm, all suspensions were sonicated for 30 s at 30% amplitude in a sonicator (Cole-Parmer 750-Watt Ultrasonic Homogenizer, Vernon Hills, IL, USA). The cells were then separated from the matrix by centrifugation at 3000× *g* for 5 min at 4 ◦C. The supernatant was filtered with a membrane pore size of 0.2 μm. The pellet, which corresponds to the cells of a biofilm without matrix, was resuspended in 1 mL of PBS to be analyzed further.

### 4.4.3. Flow Cytometry Assay

Biofilm cell viability was also determined by FCM. In brief, pre-formed biofilms were washed twice, scraped in 1 mL of PBS, vortexed at maximum speed (30 s) and analyzed by cytometry. In addition, the biofilm cells (without the EPS matrix) and the EPS matrix itself (after matrix extraction procedure) were also analyzed. Lastly, 0.5–2 μM of SYTO BC (Invitrogen™, Carlsbad, CA, USA) and 15 μM of PI (Invitrogen™, Carlsbad, CA, USA) were added to the tested suspensions. Samples were incubated in the dark for 20 min, at room temperature, and were analyzed further in an EC800TM flow cytometer (SANYO, Osaka, Japan). SYTO BC fluorescence was detected on the FL1 channel (PMT = 5) while PI fluorescence was detected on the FL4 channel (PMT = 3). SYTO BC absorbs at 485–487 nm and emits at 500–504 nm while PI excitation occurs at 535 nm and emission at 617 nm.

For all detected parameters, amplification was carried out using logarithmic scales. The cellular concentration was determined by acquiring the counts by the equipment. Multi-parametric analyses were performed on the scattering signals (forward scatter, FSC and side scatter, SSC), as well as on the FL1 (green fluorescence) and FL4 (red fluorescence) channels. When appropriate, a slight adjustment of the gate was made to guarantee the inclusion of the total population in the FCM analysis. For all assays, at least two independent experiments were carried out in duplicate.

### *4.5. Influence of Sonication on Biofilm Cell Viability*

The biofilm suspensions obtained were pooled and sonicated for 30 s at 30% amplitude in a sonicator. A non-sonicated sample was included as a control. After this, for each sample, the values of biofilm-culturable cells were determined by CFU counting. On these samples, total cell counting was performed in duplicate for both species.

### *4.6. Antimicrobial E*ff*ect on Mixed-Species Biofilms*

The antimicrobial effects of the antibiotic ciprofloxacin (Sigma-Aldrich) and the naturally occurring terpene alcohol, linalool (Sigma-Aldrich), were evaluated in mixed-species biofilms of *P. aeruginosa* PAO1 and *C. albicans* 547096. For this, 24 h-old pre-established mixed biofilms were exposed to defined concentrations of each antimicrobial: 0.3 or 1.2% *v*/*v* for linalool and 0.25 or 8 mg/L for ciprofloxacin. The rationale behind the use of these concentrations was based on the assumption that the lower concentrations had already been reported as inhibitory for planktonic culture; then, these concentrations were gradually boosted until we found one that had an inhibitory effect on biofilm without causing total cell eradication. Briefly, after biofilm formation, 500 μL of cell suspension was replaced by the antimicrobial solutions prepared at 2 times the desired concentration. Plates were then incubated aerobically at 37 ◦C for another 24 h. The results were assessed for biofilm cell culturability through CFU enumeration using selective growth media, as previously described, and by FCM. Both 24 and 48 h-old untreated biofilms were used to infer whether the antimicrobial agents demonstrated a bacteriostatic/bactericidal or fungistatic/fungicidal activity. At least two independent experiments were carried out.

### *4.7. Statistical Analysis*

Data were analyzed using the Prism software package (GraphPad Software version 6.01). One-way ANOVA tests were performed, and means were compared by applying Tukey's multiple comparison test. The statistical analyses performed were considered significant when *p* < 0.05.

### **5. Conclusions**

Overall, the obtained data allow us to strengthen the belief that FCM is a versatile and accurate technique to analyze biofilms; however, it is crucial to take into account some technical aspects to avoid erroneous interpretations. FCM analysis is strain dependent and, as the biofilm matrix is variable for each of microorganism, either grown in isolation or in polymicrobial consortia, the method of extraction (regardless of the one chosen for this purpose) must be personalized for each case. The use of FCM to analyze mixed biofilms challenged by the application of an antimicrobial can provide important insights, explaining the alterations in the behavior of the microbial community and suggesting antimicrobial modes of action in biofilm populations. These outcomes strengthen FCM as a promising technique to study heterogeneous biofilms and evaluate the efficacy of therapeutic approaches.

Throughout this work, the pitfalls related to FCM analysis of inter-kingdom polymicrobial biofilms were fully addressed and efforts were made to circumvent them in order to reliably characterize complex biofilm communities, which are still poorly explored using this approach. This work highlights the fact that this technique can be fruitfully used in the understanding of antimicrobial studies as long as specific and tailored optimization is carried out.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/11/741/s1. Figure S1: Representative dot plots obtained for planktonic and biofilm cells of *P. aeruginosa*, Figure S2: Representative dot plots obtained for planktonic and biofilm cells of *C. albicans*, Figure S3: Representative dot plots obtained by FCM of planktonic cells of *C. albicans* SC5314 (A) and hyphal growth induction (B), Figure S4: Effect of sonication process on biofilm cell viability. CFU enumeration of *C. albicans* (A) and *P. aeruginosa* (B) biofilms before and after 30 s of sonication at 30% amplitude, Figure S5: Representative dot plots obtained for *P. aeruginosa* and *C. albicans* biofilms (A) and biofilm cells after matrix extraction (B) by FCM.

**Author Contributions:** Conceptualization, T.G., A.P.M., L.D.R.M. and M.O.P.; methodology, T.G., A.P.M. and L.D.R.M.; validation, T.G., M.O.P. and L.D.R.M.; formal analysis, T.G., A.P.M. and L.D.R.M.; investigation, T.G., A.P.M. and L.D.R.M.; data curation, T.G.; writing—original draft preparation, T.G.; writing—review and editing, M.O.P., L.D.R.M.; supervision, M.O.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Portuguese Foundation for Science and Technology (FCT) under the scope of the strategic funding of the UID/BIO/04469/2020 unit and BioTecNorte operation (NORTE-01-0145-FEDER-000004), which was funded by the European Regional Development Fund under the scope of Norte2020–Programa Operacional Regional do Norte. The authors also acknowledge COMPETE2020 and FCT under the project POCI-01-0145-FEDER-029841 and FCT for the PhD grant to Tânia Grainha (grant number SFRH/BD/136544/2018).

**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/).

### *Article* **Analysing the Initial Bacterial Adhesion to Evaluate the Performance of Antifouling Surfaces**

### **Patrícia Alves 1, Joana Maria Moreira 1, João Mário Miranda 2,\* and Filipe José Mergulhão 1,\***


Received: 8 June 2020; Accepted: 15 July 2020; Published: 17 July 2020

**Abstract:** The aim of this work was to study the initial events of *Escherichia coli* adhesion to polydimethylsiloxane, which is critical for the development of antifouling surfaces. A parallel plate flow cell was used to perform the initial adhesion experiments under controlled hydrodynamic conditions (shear rates ranging between 8 and 100/s), mimicking biomedical scenarios. Initial adhesion studies capture more accurately the cell-surface interactions as in later stages, incoming cells may interact with the surface but also with already adhered cells. Adhesion rates were calculated and results shown that after some time (between 5 and 9 min), these rates decreased (by 55% on average), from the initial values for all tested conditions. The common explanation for this decrease is the occurrence of hydrodynamic blocking, where the area behind each adhered cell is screened from incoming cells. This was investigated using a pair correlation map from which two-dimensional histograms showing the density probability function were constructed. The results highlighted a lower density probability (below 4.0 × 10<sup>−</sup>4) of the presence of cells around a given cell under different shear rates irrespectively of the radial direction. A shadowing area behind the already adhered cells was not observed, indicating that hydrodynamic blocking was not occurring and therefore it could not be the cause for the decreases in cell adhesion rates. Afterward, cell transport rates from the bulk solution to the surface were estimated using the Smoluchowski-Levich approximation and values in the range of 80–170 cells/cm2.s were obtained. The drag forces that adhered cells have to withstand were also estimated and values in the range of 3–50 × 10−<sup>14</sup> N were determined. Although mass transport increases with the flow rate, drag forces also increase and the relative importance of these factors may change in different conditions. This work demonstrates that adjustment of operational parameters in initial adhesion experiments may be required to avoid hydrodynamic blocking, in order to obtain reliable data about cell-surface interactions that can be used in the development of more efficient antifouling surfaces.

**Keywords:** bacterial adhesion; blocking effect; hydrodynamics; parallel plate flow cell

### **1. Introduction**

Bacterial adhesion to a surface triggers a series of events that may lead to biofilm formation and fouling of that surface. Biofilms are bacterial cell communities that are embedded in a self-produced and highly hydrated matrix of extracellular polymeric substances (EPS). Living in biofilms is the most common state for bacteria in natural environments [1], and biofilms can be composed of single or multiple species that interact with each other [2]. Biofilms can cause deterioration of industrial equipment, food spoilage and disease [3,4], but they are also used in wastewater treatment systems and have been investigated for the production of valuable molecules, including recombinant proteins [5,6].

The first step in biofilm formation is the surface adsorption of molecules from the surrounding medium. This originates a conditioning film that can affect subsequent bacterial adhesion [7]. Initially, bacterial adhesion is reversible, but then the adhered organisms start to produce EPS and to anchor themselves irreversibly, leading to the development of the biofilm structure. Then, biofilms mature and release bacteria, often leading to serious bacterial transmission issues [1]. Microorganisms that adhere first have a pivotal role in linking the biofilm to the surface and their retention is crucial to maintain the biofilm on that surface when it is challenged by shear forces [8]. Thus, a better understanding of the initial adhesion process may provide clues to the development of antifouling surfaces.

One of the most promising strategies to prevent or delay biofilm formation on a given surface is to use a coating. In both medical and industrial settings, different types of antifouling coatings have been developed to prevent adhesion, which operate by contact killing or that release biocidal agents [9]. Release systems may promote the development of different types of resistance, and contact killing surfaces may originate a layer of dead cells and cellular debris that can serve as anchoring points for subsequent cell adhesion [7]. In that case, newly adhered cells can be protected from the biocidal agent by the layer formed by dead cells and debris. Anti-adhesion systems are, therefore, an attractive way of preventing or delaying biofilm formation [10]. The development of anti-adhesion coatings is an intense area of research, but these coatings have to be tested in environmental conditions that mimic their application scenario. Since medical and industrial biofilms often develop in areas where significant fluid motion exists, fluid displacement systems have been used to study these initial bacterial–surface interactions [11]. One of the most commonly used platforms for adhesion studies is the parallel plate flow cell (PPFC), which enables real-time monitoring of bacterial adhesion when the system is mounted on a microscope stage coupled to an image acquisition system [11].

A typical initial adhesion experiment performed in a PPFC with continuous monitoring of the number of adhered cells generates a pattern where a linear trajectory can be identified first, followed by a decrease in slope where adhesion seems to be leveling off [8]. It has been proposed that during the linear phase, cells arriving at the surface interact solely with the surface and that the rate at which organisms adhere in this phase is truly representative of the affinity of that organism to that surface [8]. It is also believed that at later stages, when the surface is partially covered, an arriving cell will interact with the surface but also with other adhered cells and that the observed adhesion rates level off due to hydrodynamic blocking [8]. As a consequence, these second adhesion rates would not be truly representative of the interaction between a single cell and the surface, as data would reflect contributions from different interactions that are hard to separate [8].

It has been demonstrated that hydrodynamic blocking can reduce the adhesion of cells by screening the surface behind already adhered cells [12]. If the surface is entirely free of cells, an incoming cell can freely attach as long as it can withstand the shear forces. However, when cells start to attach irreversibly, an incoming cell can no longer attach immediately behind an already adhered cell because that area is effectively screened. This creates a shadow where cell adhesion is prohibited (Figure 1). During colloidal particle deposition, it has been shown that the area blocked by one particle can represent 8 to 675 times the cross-sectional area of that particle [13,14]. In general, blocking is more likely to occur at high surface coverages, but local flow effects can introduce anisotropy in cell adhesion [12].

A detailed analysis of the hydrodynamic blocking has been presented in several works by Adamczyk and co-workers [15–17], but this approach is not straightforward to many researchers, and this may explain why a blocking analysis is absent from many studies dealing with adhesion under flow conditions. In a study by van Loenhout et al. [12], the hydrodynamic blocking effect on particle adsorption was assessed by performing experiments in laminar flow and Monte Carlo simulations to evaluate the effect of the hydrodynamic shadow on particle distribution. The spatial distribution and anisotropy were analysed through a two-dimensional (2D) map of a pair correlation function that was obtained from image analysis. This enabled the observation of the exclusion zone originating from the

hydrodynamic blocking. The position of the adhered particles could be calculated and revealed the density of particles adhering around a given particle when compared to the overall density.

**Figure 1.** The hydrodynamic blocking effect. The area blocked by an adhered cell where further cell adhesion is prohibited is represented by the shadow.

The arrival of cells to the surface is dictated by mass transport, and in flow systems, this transport is achieved by convection and diffusion. Although solutions for the convective–diffusion equation can be obtained by complicated mathematical procedures, there are approximate solutions such as the Smoluchowski-Levich (SL) approximation that assumes that all microorganisms sufficiently close to a surface will adhere irreversibly [8]. On the other hand, adhered cells have to withstand hydrodynamic forces that may cause cell detachment, and therefore the adhesion rates observed in initial adhesion experiments are a balance between all these effects. It has also been shown that initial adhesion experiments can produce valuable data for the development of antifouling surfaces [10,18], and therefore the correct interpretation of that data is critical.

In this study, we have monitored in real-time the initial adhesion of *Escherichia coli* to a polydimethylsiloxane (PDMS) coating. *E. coli* was chosen as a model organism due to its relevance in both clinical and industrial settings, and PDMS is a very versatile polymer commonly used in both scenarios [19]. An interpretation of the events unfolding during initial adhesion is provided, showcasing the relative importance of cell transport to the surface, hydrodynamic blocking, and detachment forces.

### **2. Results and Discussion**

Bacterial adhesion experiments were performed at different flow rates yielding a range of shear rates between 8 and 100/s, similar to relevant biomedical scenarios (Table S1) [20,21].

Figure 2 shows the number of cells adhered to the PDMS surface during the experimental time at different flow rates. In all tested conditions, an initial adhesion rate was determined by linear regression of the first experimental points (Figure 2, blue dashed line), and it was observed that after some time (between 5 and 9 min), the experimental points did not fit this initial regression. Thus, a second linear regression was made with the remaining points (Figure 2, red dashed line) so that a second adhesion rate could be determined. It was found that this second adhesion rate was lower (on average 55%) than the value determined with the first regression (in blue) (Figure 3a), but the most significant reduction was observed for the lower and higher flow rates (on average 76%; *p* < 0.05, Figure 3a). The flow rate of 2 mL/s presents the lowest reduction in the adhesion rate with a decrease of 20% (Figure 3a). Additionally, the adhesion rates were statistically different when comparing the results for the lower and the higher flow rates (1 and 2 mL/s with 8 and 10 mL/s, respectively) (*p* < 0.05).

**Figure 2.** Number of *Escherichia coli* cells adhered to polydimethylsiloxane (PDMS) as a function of time. The dashed line indicates the best fit to a linear function for the initial adhesion rate (blue) and final adhesion rate (red) of the experiments. The adhesion assays were performed during 30 min at different flow rates: (**a**) 1 mL/s, (**b**) 2 mL/s, (**c**) 4 mL/s (**d**) 6 mL/s, (**e**) 8 mL/s, and (**f**) 10 mL/s corresponding to shear rates of 7.5/s, 15.0/s, 33.7/s, 51.6/s, 80.3/s, and 100.8/s, respectively, in a parallel plate flow cell (PPFC). Error bars indicate the standard deviation from three independent experiments.

**Figure 3.** (**a**) Initial adhesion rates determined by linear regression of the first experimental points (blue bar) and for the remaining points (red bar) at different flow rates (1 to 10 mL/s). Statistical significance between the adhesion rates and the different flow rates was evaluated by one-way analysis of variance (one-way ANOVA, Tukey's post-hoc test) (*p* < 0.05); (**b**) Area fraction covered by *Escherichia coli* cells at the end of adhesion assay (30 min) on polydimethylsiloxane (PDMS) at different flow rates (1 to 10 mL/s).

It has been hypothesized that the main cause for this reduction in adhesion rate could be due to hydrodynamic blocking as arriving cells would interact with either already adhered cells or with the free surface [8]. Since the blocking effect is commonly evaluated as a function of the surface area coverage [12,22], this value was determined for the time where the slope decreased at each flow rate. Values between 0.8 to 1.8% were obtained, indicating low surface coverage at that time.

In order to assess if hydrodynamic blocking was occurring, we have performed an image analysis for each flow rate at the end of the adhesion assay (30 min), where surface coverage is at its highest value. It is precisely in those situations that blocking would be most likely to occur in our assays [12,22]. The image analysis revealed that the final surface coverage was similar for all tested flow rates (Figure 3b) and was below 4%.

Afterward, a pair correlation map was used to establish if significant areas were blocked by already adhered cells. Then, 2D histograms were created, showing the probability density function of the presence of cells around a given cell (Figure 4). The central cell was excluded from the calculations, which explains the low probability in the center of the image. Cells at a distance larger than 50 pixels (30.5 μm) were also excluded from the analysis as it has been shown that blocking is more effective at distances shorter than the cut-off value used in this work [23]. If hydrodynamic blocking was occurring, the pair correlation map should show a low probability of cell adhesion along the flow direction, as demonstrated in previous studies [12]. However, in our experiments, the pair correlation maps are symmetrical, as the density probability of the presence of a cell around a given cell was uniform (Figure 4) and had no directional bias. This demonstrates that hydrodynamic blocking was not occurring during our experiments, not at the end of the assay and surely not at the time point where the adhesion rates decreased.

In a previous study, it was shown that the size of the blocked area is a function of the dimensionless Péclet number [24]. In the present work, the Péclet number was below 0.4, which is much lower than the scenarios simulated in that study (2 to 100). Indeed, the size of the blocked area was shown to increase at higher shear rates and large particle sizes [12], and this may also explain why a shadow area was not observed in our work. Additionally, the low surface coverage may also have prevented the occurrence of blocking as it has been reported that even at surface coverages of about 10%, the blocking effect may not be significant [22].

Since hydrodynamic blocking was not occurring, other factors may explain the reduction in adhesion rates that we have observed (Figure 2). Higher flow velocities increase the number of contacts between planktonic cells and the surface, but the increased shear forces may also prevent adhesion or promote detachment [25,26]. In order to ascertain the effect of flow rate variation on the transport of cells from the bulk solution to the surface, the SL approximation was used [8]. Figure 5 shows that as the flow rate increases, mass transport is favored (values in the range of 80–170 cells/cm2·s were obtained), and since the SL approximation considers that all cells in close proximity to the surface will adhere, higher adhesion rates are expected at higher flow rates. However, as the flow rate increases, the wall shear stress also increases, and therefore higher drag forces are expected (values in the range of 3–50 × 10−<sup>14</sup> N were determined—Figure 5). It has been shown that sufficiently high shear stresses cause adhering bacteria to slide and roll over a surface, which may lead to detachment [27].

**Figure 4.** 2D histograms representing the probability density function of the presence of a cell around a given cell (reference cell is in the center). Results for different flow rates are shown: (**a**) 1 mL/s, (**b**) 2 mL/s, (**c**) 4 mL/s (**d**) 6 mL/s, (**e**) 8 mL/s, and (**f**) 10 mL/s corresponding to shear rates of 7.5/s, 15.0/s, 33.7/s, 51.6/s, 80.3/s, and 100.8/s, respectively, in a PPFC. The number of independent cells measured for each flow rate was (**a**) 1.53 × 106 ± 1.84 × 10<sup>5</sup> cells/cm2; (**b**) 1.49 × 106 ± 6.14 × 104 cells/cm2; (**c**) 1.13 × 106 ± 8.87 × 10<sup>4</sup> cells/cm2; (**d**) 1.29 × 106 ± 9.84 × 104 cells/cm2; (**e**) 9.41 × 10<sup>5</sup> ± 8.25 × 104 cells/cm2 and (**f**) 1.07 × 106 ± 1.37 × 10<sup>5</sup> cells/cm2. The direction of the flow in all histograms is depicted by an arrow located on panel (a).

**Figure 5.** Variation of the drag force (D) and values obtained with the Smoluchowski-Levich (SL) approximation as a function of the flow rate (1 to 10 mL/s). The full line represents the predicted adhesion rate using the SL approximation and the dashed line represents the drag force.

It is known that the strength of adhesion can depend on the history of the contact between a bacterium and a surface and that factors like the residence time and the shear applied during adhesion are strong modulators [28,29]. Additionally, these forces are strongly depending on chemistry [30] and mechanical properties of the surface [31]. It is likely that for each experimental condition tested, the relative importance of mass transport and detachment changes, and this may induce variations in the observed cell adhesion. In any case, since blocking is not occurring, the experimental values obtained at this second stage are also a reflection of the interaction between single cells and the surface.

### **3. Materials and Methods**

### *3.1. Bacteria and Culture Conditions*

*E. coli* JM109(DE3) from Promega (USA) was selected for this study because it has been used in previous works from our group for the evaluation of initial adhesion in antifouling surfaces [10,18,32,33] and because it was shown to have similar biofilm formation behavior to different clinical isolates, including *E. coli* CECT 434 [21]. The inoculum was prepared as previously described [34]. Briefly, 500 μL of a glycerol stock (kept at −80 ◦C) was added to a total volume of 0.2 L of the inoculation medium composed by 5.5 g/L glucose (Chem-Lab nv, Zedelgem, Belgium), 2.5 g/L peptone (Oxoid, Basingstoke, Hampshire, England), 1.25 g/L yeast extract in phosphate buffer (1.88 g/L KH2PO4 and 2.60 g/L Na2HPO4; Chem-Lab nv, Zedelgem, Belgium) at pH 7.0. The culture was incubated overnight at 37 ◦C, with orbital agitation (160 rpm in a shaker: IKA KS 130 basic, Staufen, Germany). Subsequently, this culture was centrifuged (at 3202 *g* for 10 min at 25 ◦C) to harvest the cells, and these were washed twice with 0.05 M of citrate buffer (composed by citric acid, Scharlau, Barcelona, Spain; pH 5.0) to remove any traces of the culture medium [35]. Cells were again harvested by centrifugation and resuspended in citrate buffer by vortexing in order to reach an optical density of 0.1 (OD610 nm). A calibration curve was used to determine the cell density (7.6 × 10<sup>7</sup> cells/mL). This suspension was used to perform adhesion experiments.

### *3.2. Surface Preparation and Experimental Setup*

Adhesion experiments were performed in a PPFC, as described by Moreira et al. [20]. The PPFC used in the present work has a rectangular cross-section of 0.8 × 1.6 cm and a length of 25.42 cm. Briefly, glass slides (7.6 × 2.6 × 0.1 cm, VWR, Carnaxide, Portugal) were washed with a 0.5% detergent solution (Sonasol Pril, Henkel Ibérica SA, Barcelona, Spain) for 30 min and then the detergent was rinsed with

distilled water. Subsequently, the surfaces were immersed in 3% sodium hypochlorite for 30 min. Finally, the surfaces were rinsed with distilled water and prepared for coating. PDMS (Sylgard 184 Part A, Dow Corning; viscosity = 1.1 cm2/s; specific density = 1.03, Midland, MI, USA) was prepared by performing the following steps: i) the curing agent (Sylgard 184 Part B, Dow Corning, Midland, MI, USA) was added to the PDMS at a 1:10 ratio; ii) the mixture was placed in the vacuum chamber in which the pump was turned on and off periodically thus changing the pressure and collapsing air bubbles that may have formed. Subsequently, the mixture was used to coat glass slides by spin-coating (Spin150 PolosTM, Caribbean, Netherlands) at 4000 rpm for 60 s in order to obtain a thickness of 10 μm. The PPFC was mounted in a microscope (Nikon Eclipse LV100, Tokyo, Japan).

### *3.3. Adhesion Experiments*

The bacterial suspension was introduced in the flow cell for 30 min at flow rates of 1, 2, 4, 6, 8, and 10 mL/s. The flow rates were adjusted with a valve. Adhesion was followed by brightfield microscopy, and three trials were performed for each flow rate. Images were taken at 60 s intervals to enable a more accurate adhesion analysis. Obtained images were processed using ImageJ (version 1.38e) software [36] and for each flow rate tested, the number of adhered cells per unit area was determined as a function of time (Code 1—supplementary material). Initial adhesion rates were obtained by linear regression analysis of initial points. After some time, from 5 to 9 min, the adhesion rate decreased, and the remaining points were subjected to a second linear regression. The difference between the first and second slopes was analysed. Standard deviations on the triplicate sets were calculated for all analysed parameters. Graph production and statistical analysis were performed using GraphPad Prism 6.01 (La Jolla, USA). The differences between the slopes as well as the variation of the adhesion rates with the shear conditions were evaluated using one-way analysis of variance (one-way ANOVA, Tukey's post-hoc test). All statistical analysis used a 95% confidence limit, so that *p* values equal to or greater than 0.05 were not considered statistically significant.

The percentage area fraction covered by *E. coli* cells was determined for different flow rates (1 to 10 mL/s). For this, ImageJ macro scripts were created to convert the images to an 8-bit greyscale file format and thresholds were applied to determine the occupied area (Code 2—supplementary material).

### *3.4. Blocking Analysis*

After each 30 min trial, 5 images were taken in different regions of the surface to obtain a large set of adhered cells. Using ImageJ, the images were inverted, and the maxima detected (Figure S1). Each maximum corresponds to a cell. The respective coordinates were obtained and a pair correlation map was constructed [12]. Results are presented in the form of a 2D density probability function. The probability density function can be used to calculate, by integration, the probability of a cell being found around another cell at a region defined by Δ*x* and Δ*y*, where *x* and *y* are the coordinates centered in the reference cell. The probability density function was calculated using an R language script (see R scripts—supplementary materials). To find the probability density function, a threshold *Rmax* was defined. Pairs at a distance larger than *Rmax* were not considered to construct the probability density function.

### *3.5. Mass Transfer and Drag Force*

The theoretical mass transport was estimated though the *SL* equation (approximate solution) [8] for each flow rate:

$$\Delta SL = 0.538 \frac{D\_{\rm os} \mathcal{C}\_{\rm b}}{R\_{\rm b}} \left( \frac{\text{Pe } h\_0}{\text{x}} \right)^{1/3} \text{ } \tag{1}$$

where *<sup>D</sup>*<sup>∞</sup> is the diffusion coefficient (approximately 4.0 <sup>×</sup> <sup>10</sup>−<sup>13</sup> <sup>m</sup>2/s for *E. coli* [37]), *Cb* is the bacterial concentration (7.6 × 10<sup>13</sup> cell/m3), *h*<sup>0</sup> is the height of the rectangular channel (0.08 m) and *x* is the distance for which an average velocity variation below 15% was determined (m) as detailed in Moreira, et al. [20]. *Rb* corresponds to the microbial radius (4.5 <sup>×</sup> <sup>10</sup>−<sup>7</sup> m) assuming that *E. coli* has a cylindrical shape estimated accordingly to the equation [15]:

$$R\_b = \frac{1}{\ln\left(\frac{L}{b}\right) - 0.11} \left(\frac{L}{2}\right). \tag{2}$$

where *L* and *b* are the length and the diameter of the *E. coli* used in this study, respectively.

The *SL* equation also includes the Péclet number (*Pe*) which represents the ratio between convective and diffusional mass transport, given for the parallel plate configuration as:

$$Pe = \frac{3v\_{\text{av}}R\_b{b}^3}{2\left(\frac{h\_0}{2}\right)^2D\_{\text{ov}}},\tag{3}$$

where *vav* is the average flow velocity (m/s) determined in Moreira et al. [20].

It is assumed that gravity effects and hydrodynamic lift are negligible compared to the drag force [38], which was estimated by the following equation [38]:

$$D = 32.0\tau\_w R\_b^2 + O(\text{Re}\_c),\tag{4}$$

where τ*w* corresponds to the wall shear stress.

The flow within the near-wall region can be characterized using the local Reynolds number (Re*c*), based on the shear rate, γ, [38], as follows:

$$\text{Re}\_{\mathfrak{c}} = \rho \gamma R\_{\mathfrak{h}}^2 / \mu\_r \tag{5}$$

where ρ, is the density (993.37 kg/m3), and μ is the dynamic viscosity of the fluid (0.000694 kg/m.s). The values for τ*<sup>w</sup>* and γ were obtained by Computational Fluid Dynamics as described in [20] and are listed in Table S1 (Supplementary material).

Re*<sup>c</sup>* is always lower than 1, as *Rb h*0, and so the inertial effects are negligible in the near-wall region and terms of higher order, *O*(Re*c*), in Equation (4) are negligible [38].

### **4. Conclusions**

Bacterial adhesion studies are important not only because they provide an understanding of the early stages of biofilm formation but also because they may provide clues for the development of more efficient antifouling surfaces. These studies should be performed in conditions that mimic the real-life scenario not only regarding the surfaces and bacteria under evaluation but also in defined hydrodynamic conditions prevailing on that scenario. When real-time monitoring of initial adhesion is performed, a decrease in the initial adhesion rates is often observed along the experimental time. This decrease is often associated with hydrodynamic blocking, which can occur at significant surface coverage values. For low surface coverage situations, this decrease is most likely caused by cell detachment, which occurs when the force exerted on a single bacterium overcomes the adhesion force between the cell and that surface. Initial adhesion experiments should, therefore, be conducted so that low surface coverage values are obtained (by adapting the test conditions, namely the assay time), and the absence of blocking should be verified so that reliable results can be obtained. This enables the performance evaluation of different coatings so that more efficient antifouling surfaces can be developed.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/7/421/s1, Figure S1: Steps of the method. From left to right: initial image, inverted image and maxima; R scripts—Relative positions script and Histogram script; Code 1—Cell adhesion analysis; Code 2—Surface coverage; Table S1: Shear rate and wall shear stress at the different flow rates tested (determined by Computational Fluid Dynamics) and examples of biomedical scenarios where these shear rates can be found.

**Author Contributions:** Conceptualization, P.A., J.M.M. (Joana Maria Moreira), J.M.M. (João Mário Miranda) and F.J.M.; methodology, P.A., J.M.M. (João Mário Miranda) and F.J.M.; investigation, P.A., J.M.M. (Joana Maria Moreira) and J.M.M. (João Mário Miranda); resources, J.M.M. (João Mário Miranda) and F.J.M.; data curation, P.A., J.M.M. (João Mário Miranda) and F.J.M.; writing—original draft preparation, P.A. and F.J.M.; supervision, J.M.M. (João Mário Miranda) and F.J.M.; funding acquisition, J.M.M. (João Mário Miranda), F.J.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Base Funding—UIDB/00511/2020 of the Laboratory for Process Engineering, Environment, Biotechnology and Energy—LEPABE and Base Funding—UIDB/00532/2020 of the Transport Phenomena Research Center—CEFT—funded by national funds through the FCT/MCTES (PIDDAC). P.A. acknowledges the receipt of a Ph.D. grant from the Portuguese Foundation from Science and Technology (FCT) (PD/BD/114317/2016). The authors would like to acknowledge the support from the EU COST Actions ENIUS (CA16217) and ENBA (CA15216).

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

### **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*
